Docketwise vs eImmigration by Cerenade: how they compare in 2026
Docketwise and eImmigration by Cerenade sell the same product to the same firms: immigration case management built around a library of government forms, with AI that reads passports and other documents into those forms. eImmigration sits in the top two bands on nine of fifteen axes and Docketwise on seven of fifteen. The gap is in eImmigration's own paperwork. Its subscription agreement and data processing addendum, both dated 20 February 2026, earn it the higher grade on privilege, liability and deployment: a dedicated database for each customer in the United States, five named subprocessors, breach notice within 24 hours, deletion within 30 days with written certification, and the costs of a data incident carved out of the liability cap. Docketwise answers on the question buyers ask of AI first. The 8am Terms of Service that govern it state that customer content is not used to train or fine tune any model and that model providers may keep it no more than 30 days, while eImmigration's agreement limits client data to performing the service and never mentions training. Both publish full price lists, and neither names the model behind its AI.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The models are a core capability layered on a product that would function without them, which is B, and the vendor's own tier structure demonstrates it rather than leaving it to inference. Docketwise was founded in 2016 and built its position on the forms library: USCIS, EOIR, DOL and DOS forms auto-populated from a single multilingual questionnaire, with case management, deadline and priority-date tracking, USCIS status integration, eSignature, invoicing and trust accounting around it. **The Basic plan at $69 per user per month contains no AI at all** and remains a working immigration practice system; DocketWise IQ appears only on the Pro and Advanced tiers. What the AI adds is real but bounded: a Document Assistant that reads scanned passports and green cards and maps extracted details into forms, and a Writing Assistant that refines and translates correspondence. Recorded on the other side, because the direction is visible: 8am launched enhanced IQ capabilities across the portfolio in June 2026 and describes AI-assisted data capture as central to the value, and the vendor now leads its own home page with auto-filled forms and AI-assisted data capture together. Verified 12 September 2026.
The AI is a layer on a case management system that has been sold for three decades, and the system is the product. Four AI Workmates are Cerenade's own: Summarizer, which digests an uploaded PDF, pasted text or a URL; Extractor, which pulls data from passports, USCIS receipts and other PDFs including handwritten ones and maps it into client and case fields; the Translator behind thirteen-language intake and form translation; and Report.AI, which answers caseload questions in plain language and offers to run the report. Two of the four are standard on every plan, one sits inside a premium feature. Around them sit more than 300 government forms with version control, 120 case templates, e-filing to USCIS, State and Labor, travel-history import from CBP, a client portal, billing and trust accounting, none of which needs a model. A firm that switched every Workmate off would still have the product it bought. Verified 20 September 2026.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Accuracy is asserted without measurement, which is C. The claims are consistent and unquantified: form auto-population produces measurable accuracy and time savings, the Document Assistant reduces manual entry and improves accuracy, and the platform minimises filing errors, described as the number one source of error in immigration practice. No figure, error rate, sample, test set or evaluation is published anywhere, and no failure mode is named. Grounding is real in outline, since extraction runs on the customer's own uploaded identity documents and populates named agency forms rather than generating free text, but nothing describes how a mapped value is traced back to the page it came from or whether the source is shown to the reviewer. R15 governs the authority limbs: the product cites no legal authority, so citator-style checking and linked primary sources do not bite. One thing is expressly not credited here. The 8am group publishes a **Case Assistant** that returns answers with citations traced back to their source, and its own availability table places that assistant in MyCase and not in DocketWise, so it belongs to a sibling product and is not this record's evidence. Verified 12 September 2026.
Accuracy is asserted on the marketing pages and measured nowhere. The AI Workmates page promises error-free, ready-to-file packets and the home page promises to eliminate data entry errors; the Extractor is described as preventing data entry mistakes. No accuracy rate, error rate, evaluation or test set is published for any of the four features, and no failure mode is named. Grounding is inherent rather than documented: each Workmate works on material the firm hands it, so a summary comes from the document uploaded and an extraction from the PDF supplied, but nothing describes how a field is matched, whether a low-confidence extraction is flagged, or how a reader checks a translated form against the English original. The one published acknowledgement that AI can produce errors sits in a resource piece about using AI generally, not in any description of these features. Verified 20 September 2026.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
The AI is scoped narrowly and the oversight around it is implied rather than described, which is C. The scope is genuinely modest and the vendor is straightforward about it: the Document Assistant extracts details from scanned identity documents and maps them into form fields, and the Writing Assistant edits and translates correspondence. Neither drafts a legal position or decides anything, and the workflow that surrounds them is one an attorney drives, completing a questionnaire, reviewing populated forms and filing them under signature. Reverse Autofill and real-time collaboration on questionnaires are published as features, which are editing surfaces of a kind. What is not published is any of the structure this axis asks for. Nothing states that a populated form must be reviewed before filing, no confidence indicator is exposed on an extracted value, nothing describes what happens when a scanned document is illegible or a field cannot be matched, and no threshold or escalation route appears. That matters here because the output is a government form signed under penalty of perjury: a wrongly extracted date of entry or alien registration number is not a drafting inconvenience. Verified 12 September 2026.
The expectation of human review is published, and no control inside the product is described. Vendor resource material tells firms that a person must check anything legal every time and that a paralegal should verify auto-populated forms before anything moves forward, which is a real position, addressed to how a firm should run itself rather than to what this system does. On the product pages the framing is assistive throughout, the word Workmate doing the work: the AI completes intakes, extracts, translates and reports. What is missing is the mechanism. Nothing states whether extracted data lands in a client or case profile before or after someone confirms it, whether a translated form is marked as machine-translated when it reaches the client, or what happens after an extraction is wrong. Intake Review exists as a standard feature, but it reviews the client's questionnaire rather than the model's output. Verified 20 September 2026.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Named customers in quantity without an outcome carrying a method, which is B under R122(2). Five firms are named on the pricing page with logos, including Kurzban Kurzban Tetzeli and Pratt, Davis & Associates, Rushton Hoffmann & Associates and Fogle Law Firm, and three testimonials carry a named attorney, a title and a firm: Saja Raoof of Saja Raoof, Inc. Law Corporation, Shahzad Khan of Shahzad R. Khan Legal, and Anna Ernest of Ernest Law Group. Scale is published, more than 7,000 immigration law firms and 15,000 legal professionals, alongside an average of six hours saved per day footnoted as based on real customer outcomes. **That footnote gestures at a basis without supplying one**, and a productivity claim of tenfold inside a testimonial is a customer's impression rather than a measurement, so neither reaches the A limb. The closest thing to a dated third-party artifact is recorded and weighed rather than credited: the AILA Marketplace Study 2022, reported by Law360, ranked Docketwise the most used immigration case management software. That is an adoption ranking rather than an outcome, and it is four years old. A success-stories page is published and was not opened. Verified 12 September 2026.
Customers are named in quantity and the figures sit apart from them. More than a dozen attributed testimonials carry a name, a role and a firm, among them managing attorneys, DOJ accredited representatives and a chief operating officer, several describing a decade or more of use; the logo wall on the pricing page names RAICES, The Legal Aid Society, Quarles & Brady, CARECEN, Immigrant Defenders Law Center and the YMCA, which is an unusually checkable set for this segment. The published numbers are of a different kind: four times faster case preparation, ten days quicker client intakes, three times faster growth, 17,000 legal professionals. They are averages with no firm attached, no period and no method. The gap that matters for this index is that none of the named evidence is about an AI Workmate; the testimonials are about forms, reporting and support. Verified 20 September 2026.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Confidentiality is addressed through security controls rather than through a data position, which places this at C. The protection side is specific and is graded principally on the stewardship row: AES-256 encryption at rest, HTTPS throughout in transit, role-based access controls configured by administrators at user level, permission groups and multiple admin roles on the Advanced tier, two-factor authentication on all plans, production access limited to authorised employees, staff with access to sensitive data required to use single sign-on or two-factor authentication, background checks and signed confidentiality agreements. Three of the five limbs this axis asks about are unaddressed on the surfaces read. **Nothing states whether customer content is used to train models**, which is the first question a firm should ask of a tool that reads its clients' passports. No model provider is named, so nothing states what any third party may retain. And privilege and work product are addressed nowhere, on a product holding the complete file of a client whose immigration status may be precarious. Recorded rather than resolved: the 8am group Privacy Policy and Terms of Service are published at 8am.com and were not opened on this pass. Verified 12 September 2026.
The commitments are contractual, current and readable before signing, which is the main thing a buyer needs. The Software Subscription Agreement of 20 February 2026 makes each party's confidential information the exclusive property of the disclosing party, usable only for the purposes of the agreement, and it deems the subscriber's database files and backups the subscriber's own confidential information. The Data Processing Addendum of the same date goes further: Cerenade may not retain, use or disclose personal data for any purpose beyond performing the services, may not combine it with data from any other source, may not sell or share it, and must keep staff under written confidentiality agreements with access limited to those performing the services. A data protection officer is named. Two things are absent. Privilege and attorney work product are never addressed in those terms, on a platform holding asylum declarations and family histories. And separation inside a firm is permissions rather than walls. Verified 20 September 2026.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
AMENDED 12 September 2026, from D. The prior grade rested on the finding that nothing was published on who may use the product, and named the 8am Terms of Service as a document not opened on that pass. The agreement has now been read in full and it publishes a real position, so D, whose words are that nothing is published on the advice line, is no longer true of this record. What the agreement and the 8am IQ Usage Policy publish, both adopted by this product's own footer and both naming 8am DocketWise or DocketWise IQ expressly: the platform is for legal, accounting and other client-based professionals located in the United States and Canada; the vendor states that it does not provide legal advice and is not engaged in the practice of law; the customer must use the platform in compliance with any rules of professional conduct applicable to its profession; unauthorized practice of law is named as a prohibited activity in both instruments; and output may not be relied on as a substitute for professional judgment or human review. C rather than B, because C's own words are true of this record and B's are not. The intended audience is left ambiguous, and on this product that ambiguity is the substance rather than a technicality. Docketwise is sold to attorneys and also to legal service providers and nonprofits, in the one United States field where non-lawyer accredited representatives may lawfully practise before the agency, and the phrase legal, accounting and other client-based professionals does not resolve whether a DOJ-accredited representative sits inside or outside it. Nothing published addresses what a non-lawyer representative may do with the drafting tool, and nothing on the vendor's own marketing estate addresses professional responsibility at all. The sibling record CasePeer, governed by the same agreement, sits at B because its audience is unambiguous, so the distinction between the two is the audience rather than the paperwork.
A position on human review is published, and it is about how a firm should use AI rather than about what this product is. Vendor resource material says that if AI assists with research, drafting or document review a person needs to check it every time, that a review step should be built into standing procedures so it is never optional, and that a paralegal should verify auto-populated forms before anything moves forward. Nothing on a product page or in the agreement states what eimmigration is and is not, who inside an organisation may rely on an extracted field or a machine translation, or any jurisdiction limit. This matters more than usual here because the product is sold to non-profit legal service providers, universities and corporate teams alongside law firms, and because DOJ accredited representatives appear in its own testimonials. The nearest thing in the agreement is a promise of reasonable effort rather than a guarantee that published forms are accurate and current. Verified 20 September 2026.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
No governance position for the AI was located on the surfaces read, which is the D band. The vendor publishes an unusually complete security and compliance summary covering attestations, infrastructure, access, operational testing, personnel, governance and incident response, and **not one line of it concerns the AI**: nothing names a framework, identifies who is accountable for AI decisions, describes what is evaluated before an assistant ships, or addresses uneven output. The one governance item in the summary that touches the board, an annual briefing on cybersecurity and privacy risk, is a security governance practice rather than an AI one. Bias is live rather than theoretical on this product and the note records why. The Document Assistant reads scanned passports and green cards, which are identity documents issued in many formats, scripts and transliterations, and it maps what it reads into fields on forms filed under penalty of perjury by applicants whose immigration status may turn on a date or a name rendered correctly. Nothing published addresses accuracy across document types, issuing countries or non-Latin scripts. Named as unopened under the sufficiency discipline: the 8am AI page and the trust centre at trust.8am.com. Verified 12 September 2026.
Nothing is published about how the AI is governed. No one is named as accountable for what the models do, nothing describes what is tested before a change ships, no evaluation results appear, and there is no responsible AI statement or principles page. The question this product invites is left entirely open: intake questionnaires and government forms are machine-translated between English and twelve other languages including Farsi, Pashto, Ukrainian and Vietnamese, and the Extractor reads handwritten PDFs, so whether quality holds across languages, scripts and handwriting is the obvious thing a buyer would want tested. Nothing addresses it. Checked the AI Workmates page, the security capabilities page, the pricing page, the integrations page, the privacy policy, the subscription agreement, the data processing addendum and the subprocessor list on 20 September 2026. Verified 20 September 2026.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
Real commitments are published on most of the question, short of the full set, which is B, and the disclosure is more granular than the size of the product would suggest. Access control is the strongest part: role-based controls with administrators configuring permissions at user level, permission groups and multiple admin roles on Advanced, two-factor authentication available on every plan, production access limited to authorised employees who need it for their role, staff with sensitive-data access required to use single sign-on or two-factor authentication, a documented and enforced password policy, background checks on all new employees and signed confidentiality agreements. Encryption: AES-256 at rest and HTTPS for all data in transit. Resilience: AWS across multiple availability zones with continuous automated backups roughly every five minutes and encrypted snapshot volumes. Operational testing: annual third-party penetration testing with SLA-bound remediation, continuous third-party vulnerability scanning, annual control self-assessments with documented corrective actions. Incident practice: a documented protocol covering escalation, rapid mitigation and post-incident review, with customers notified of critical system changes. Deletion: customer data purged from production environments when customers leave. **The missing limb is subprocessors**, no list appearing on any surface read. Verified 12 September 2026.
All of it is published, current and contractual. Deletion: on termination the customer's data is returned first, then deleted within 30 days of confirmed receipt, to NIST media sanitisation standards, with written certification that it was done. Retention: processing continues until the customer says stop in writing, and personal data is deleted when the customer asks. Access: staff are under written confidentiality agreements, access is limited to those performing the services, and the customer controls roles and permissions in the product. Subprocessors: five are named on a page incorporated into the addendum, each with its purpose and its processing location, with advance notice of any change and a thirty-day objection right that can end the contract. Incidents: notice within 24 hours, with the customer controlling who gets told, and the customer's notification, forensic and legal costs expressly carved out of the liability cap. The agreement adds encryption in transit and at rest, seven-day point-in-time restore and geo-replication. Verified 20 September 2026.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
Liability is not addressed on any surface read beyond the ordinary commercial framing, which places this at C rather than lower, and the note states the retrieval position so a later pass can correct it. Nothing published on the product estate sets out a warranty, an indemnity, a liability cap, a service credit or an insurance position, and nothing distinguishes responsibility for a form populated wrongly by the Document Assistant from responsibility for one typed wrongly by a paralegal. What the pricing page does publish sits adjacent rather than on the point: a seven-day free trial, no setup fees on standard accounts, free data migration on annual Advanced plans, and priority support at the top tier, which are commercial terms rather than an allocation of risk. The group's Terms of Service are published at 8am.com/terms/terms-of-service and the Privacy Policy alongside it; **neither was opened on this pass** under the sufficiency discipline, and they are the documents that would carry the cap and any indemnity, so this row would move on a reading. The question has real content here because the output is filed with a federal agency: an incorrect alien registration number or date of entry on a submitted form is a consequence a firm would want allocated in writing. Verified 12 September 2026.
There is a real published allocation of risk, and the accuracy of what the software produces sits outside it. Cerenade indemnifies the subscriber against patent, trademark and copyright claims arising from the software, with the conditions spelled out. Liability is capped at twelve months of subscription fees, consequential and indirect damages are excluded both ways, and the functionality warranty's sole remedy is that Cerenade corrects the software. One carve-out is genuinely favourable and unusual: the costs of a data incident caused by Cerenade or its subprocessors, including the customer's notification, legal, public relations and forensic costs and any regulatory fines, are expressly not subject to the cap. Against that, the agreement promises only reasonable effort, not a guarantee, that published forms are accurate and free of errors, and nothing anywhere addresses liability for what the AI Workmates produce. No insurance is published. Verified 20 September 2026.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Real integrations, named across the categories an immigration firm actually uses, plus a published API, short of the depth an implementer could work from. The named list is unusually broad for a product this focused: practice management connections to **Clio, MyCase and PracticePanther**, accounting to QuickBooks, storage to Dropbox and OneDrive, email and calendar to Gmail, Outlook and Google Calendar, payments natively through LawPay, and a listing on the Google Workspace Marketplace. Two of those are worth naming as unusual. Integrating with Clio and PracticePanther means connecting to direct competitors of the parent's own MyCase, which the vendor presents as serving firms that want to run both. And an **open API is published with a developers page**, which is more than most records at this size offer. Gmail and Outlook are described as add-ons in the plan comparison, so the shape of those two is at least indicated. What holds it off A is documentation: nothing describes which objects sync, in which direction, what authentication is required or what a firm must configure, and the developers page was not opened on this pass. Verified 12 September 2026.
Twenty-two integrations are named on the integrations page, each with a line saying what moves and in which direction, which is what an implementer actually needs. Files are pulled in from OneDrive, Google Drive, Box and Dropbox; Outlook, Exchange and Gmail send client email and log threads and attachments back to the right case; expense items generated in the product are pushed to QuickBooks; HubSpot records sync so an update in either system shows in both; RingCentral answers calls and logs them to client records; WhatsApp Business messages land in the communications centre; UPS, USPS, FedEx and DHL shipments are tracked from the case. Scheduling runs off the firm's own Acuity or Microsoft Bookings calendar. Underneath sit the connections that matter most in this practice area: filing to USCIS, the State Department and the Labor Department, and travel-history import from Customs and Border Protection. API access is a Complete-plan feature for accounts with ten or more users and the documentation is shared on request rather than published. Verified 20 September 2026.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Cloud delivery is stated with real architectural detail and neither the region nor the tenancy model is given, which is C. The deployment side is described further than most records manage: the platform is hosted on Amazon Web Services, in data centres described as having round-the-clock security, biometric scanning and video surveillance, with infrastructure distributed across multiple AWS availability zones for failover and disaster recovery and continuous automated backups taken roughly every five minutes as encrypted snapshot volumes. That tells a buyer how the service survives a failure. What it does not tell them is where it is. No AWS region is named, no country is identified, no residency option is offered or refused, and nothing distinguishes where documents are stored from where the Document Assistant processes them. No tenancy model is described and no single-tenant or private option is mentioned. The gap has particular weight on this product and the note records why: the customer base is immigration firms, the data is identity documents and immigration histories belonging to non-citizens, and where that material physically sits and under whose jurisdiction is a question those clients have direct reason to ask. Verified 12 September 2026.
Both halves of the question are answered, and the answer is the same everywhere: a single tenancy, in the United States. Each customer's data sits in its own dedicated database rather than a shared one, stated on the security page and borne out by the agreement, which offers a copy of the subscriber's own SQL Server database on request and says plainly that segments or selected records cannot be extracted from it. Hosting is Microsoft Azure, with on-the-fly geo-replication to a second Azure data centre within the United States and a seven-day point-in-time restore window. Every one of the five named subprocessors processes in the United States. What is not published is choice or detail: no region option, nothing that differs between the three plans, nothing about where the AI features run as distinct from where the data sits, and an express right for Cerenade to change the hosting facility at its discretion without notice. Verified 20 September 2026.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
Certification is real, current and scoped, with a public trust centre naming the route to the report, short of the report period and auditor, which is B at the top of the band. **R16 is satisfied in the crediting direction, which is worth setting out because this corpus more often applies it the other way.** The attestation is the parent's, SOC 2 Type 2 held at the 8am LLC level, and the child's own surface makes the connection expressly: Docketwise states that it is operated by 8am LLC, that it maintains an active compliance program documented at 8am's public trust centre, and that the current SOC 2 Type 2 report is available through it. That is the second of R16's two connectors, the child's own surfaces linking the parent's artifact as covering it, so the attestation credits rather than being merely described. The trust centre is public at trust.8am.com and Docketwise also publishes its own security page. Around the attestation sits a real programme: annual third-party penetration testing with SLA-bound remediation, continuous third-party vulnerability scanning, annual control self-assessments with documented corrective actions, and PCI DSS v4.0.1 compliance through the LawPay integration. What holds it off A is that no report period, observation window or auditor is named, and the trust centre was not opened on this pass. Verified 12 September 2026.
One certification is real and checkable, the rest are claims. Cerenade has certified to the U.S. Department of Commerce under the EU-U.S. Data Privacy Framework, the UK Extension and the Swiss-U.S. Data Privacy Framework, names the FTC as the enforcement authority and BBB National Programs as the dispute mechanism, and points to the public register where the certification can be looked up without a sales call. Beyond that the security page claims full compliance with GDPR and PCI DSS and a host of other standards, with no scope, no date, no auditor and no report, and says annual penetration testing is carried out without publishing anything from it. The addendum offers the most recent third-party audits or certifications to a customer on written request. No SOC 2 attestation is claimed anywhere, and the ISO 27001, FedRAMP and SOC certifications named on the security page are Microsoft Azure's, which cover the hosting layer and not this vendor. There is no trust portal. Verified 20 September 2026.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
The vendor refers to its AI by brand without identifying anything underneath it, which is C. The AI is named twice over and inconsistently, as **DocketWise IQ** on the product's own pricing page and as **8am IQ** across the group, and neither name resolves to a model. No model is identified, no version is given, no provider is named, no hosting arrangement for inference is stated, and nothing commits to notifying a customer if any of it changes. The security summary names Amazon Web Services as the infrastructure host, and under the ground rules a cloud provider on an infrastructure statement is infrastructure rather than a model, so it is graded on the deployment row and not here. No subprocessor list exists on any surface read, which would be the ordinary place a model provider surfaced. The gap is material rather than formal on this product: the Document Assistant reads scanned passports and green cards, so identity documents belonging to non-citizens are being processed by a model whose operator the customer is not told, and a firm asked by its client where those images went could not answer. The trust centre was not opened and is named as the likeliest home of a subprocessor list. Verified 12 September 2026.
Four AI features ship and nothing published says what is underneath any of them. No model, version or provider is named on the AI Workmates page, the pricing page, the security page, the privacy policy, the agreement or the addendum; nothing states where inference runs; and there is no commitment to tell customers when a model changes, although there is a detailed one for subprocessor changes. The subprocessor list is worth recording as a fact here rather than as an inference: it was updated on 20 February 2026, is incorporated into the addendum, and names five processors, Microsoft Azure for hosting, HubSpot, SendGrid, Userpilot and Zoom, none of which is identified as a model provider. A buyer reading the estate cannot tell which model reads its clients' passports or drafts its translations. Verified 20 September 2026.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
A buyer can learn what this costs without entering a sales process, which is the A band. Rates are published outright, per user per month, in both billing cadences: **Basic $69, Pro $99 and Advanced $119 on annual billing**, against **Pro $109 and Advanced $129 monthly**. The unit is stated plainly and is the one that matters to a firm, a per-user seat with unlimited cases. What each tier contains is set out twice, in three plan cards and then in a full comparison table running to roughly forty rows across case management, forms, CRM, DocketWise IQ and administration, so a buyer can see exactly where the AI sits, which is Pro and above, and where eFiling, eSignature and permission groups sit. The commercial terms around the price are published too: a seven-day free trial, no setup fees for standard accounts, free data migration on annual Advanced plans, escalated feature requests on annual plans, and onboarding and migration available on request. Purchase is self-serve on the Advanced tier through a free-trial signup, with a demo route offered alongside. Nothing material is withheld: there is no unpriced enterprise tier, and the only variable cost, LawPay transaction processing, is disclosed with fees waived for new subscribers. Verified 12 September 2026.
A firm can work out what this costs without speaking to anyone. Three plans are published with both billing cadences: Starter at $60 per user per month, or $55 billed annually, with one premium feature and 10 GB of storage per user; Essentials at $90, or $82 annually, with four premium features and unlimited storage; Complete at $110, or $100 annually, with all premium features, unlimited storage and API access for accounts of ten or more users. Extra premium features on the lower plans cost $15 each per user per month. The unit is the user and the page states there are no per-case fees, no setup fees and no fees. Implementation is priced too: a standard data import is $1,000, with eighteen source platforms named and a custom migration quoted separately. Four discount routes are described, including 50 per cent off for non-profits, schools and universities on an annual term. Verified 20 September 2026.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Coverage is stated with its limits, which is the A band, and this is the clearest statement of scope located in this pull. The practice area is not merely named but declared exclusive: the product is built exclusively for immigration law, and the vendor makes the boundary the centre of its positioning rather than a caveat, contrasting itself with generalist platforms and stating that its data model, workflows and forms library are built around immigration practice. Firm types are enumerated: solo attorneys, boutique firms, larger immigration teams, legal aid and nonprofit immigration organisations, with named audience segments covering attorneys, paralegals and support staff, firm administrators and consultants. Jurisdictional and substantive coverage is given at the level a practitioner would check, the forms library being defined by agency across USCIS, EOIR, DOL and DOS, with the e-filing subset named form by form and one form flagged as temporarily unavailable pending a version update. What is out of scope is addressed rather than left implied: the vendor states plainly that firms wanting general practice management should run Clio or MyCase alongside, and publishes the integrations to do it. Six common use cases are published. Verified 12 September 2026.
Who this is for is set out with unusual precision and where it stops is not. Seven buyer pages address immigration attorneys, paralegals and caseworkers, legal ops leaders, law firms, non-profit organisations, corporations, and schools and universities, and the feature set is mapped to each. The practice coverage is stated in numbers a buyer can check: more than 300 immigration forms with version control, 120 pre-built case templates, family and employment-based matters, filing to USCIS, the State Department and the Labor Department. Two premium features extend past immigration, one for California Judicial Council and county court forms and one described as non-immigration practice areas such as personal injury. What is never stated is the boundary: which matter types, visa categories or proceedings are not supported, what the system does for work before the Board of Immigration Appeals or the immigration courts, and what the non-immigration module actually covers. Verified 20 September 2026.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
AMENDED 12 September 2026, from silent. The prior value recorded that the 8am Terms of Service were published and unopened and that R43(1) was therefore not discharged; the agreement has now been read in full and it settles the question in the vendor's favor. Terms of Service Part III provides that 8am does not use Customer Content or IQ Content to train or fine-tune any large language model or other generative AI model, whether developed by 8am or a third party, and that the substantive content of inputs and outputs is never used to update model weights or otherwise incorporated into the training data of any AI model.
It binds the model layer as well, the LLMs used to deliver 8am IQ being permitted to retain content for up to thirty days and prohibited from training on it. The agreement names 8am DocketWise expressly as a covered Practice Management Solution and this product's own footer adopts it, so it reaches this record on both R16 connectors. Three qualifications belong on the record. Part I grants a perpetual license over aggregated and anonymized Customer Content, excluding personal information, for internal business use to improve the platform, which under the R28 naming test is not a training right and which Part III controls over in any event.
The vendor reserves aggregated human and automated review of inputs and outputs to monitor accuracy and performance, so a person may read a prompt even though no model trains on it. And the weight of the question on this record is unchanged: what is processed here includes scanned passports and green cards belonging to non-citizen clients.
The word training appears nowhere on this estate, in either direction, and the contract closes the route anyway. The Data Processing Addendum of 20 February 2026, which forms part of the subscription agreement and prevails over it on anything touching personal data, says Cerenade may not retain, use or disclose personal data for any purpose other than performing the services, may not use it for any other commercial purpose, may not sell or share it, and may not combine it with data collected from any other source.
That last limb is the one that matters for a model: material from one firm's cases cannot be pooled with anything else. It is a contractual limit rather than an express prohibition on training, and the four AI features are never mentioned in the agreement, the addendum or the privacy policy.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
AMENDED 12 September 2026, from disclosed-vague. The prior value rested on the vendor's compliance summary, which acknowledges a data lifecycle without stating a period; the 8am Terms of Service, named as unopened on that pass, have now been read and they state a period for exactly the material this signal is about. Terms of Service Part III provides that the large language models used to provide 8am IQ may retain IQ Content for up to thirty days, and that this data will not be used to train any LLM.
That is a bounded, contractual answer to what happens to a prompt after processing, and it is stated about the model provider rather than only about the vendor. The agreement adds that 8am itself will process and store IQ Content in accordance with its privacy policy and retain it as necessary to comply with law, perform the services, enforce the agreement or identify misuse. Two limits are recorded and neither displaces the value.
The thirty-day window governs the model layer, and no period is published for the retention of case content generally, the vendor's own material deferring to a Data Retention policy whose period is not stated and committing only that customer data is deleted or de-identified after the applicable period on departure. And a stricter retention rule that appears in the same section does not reach this product: the bar on deleting an 8am IQ Case Assistant chat history attaches to Case Assistant, which the agreement confines to the 8am MyCase Advanced tier.
The customer sets the window by instruction rather than by a setting. Under the addendum Cerenade processes customer data until the customer notifies it in writing to stop, and it must delete personal data if and when the customer asks. At the end of the contract the data is returned first and then deleted within 30 days of the customer confirming receipt, to NIST media sanitization standards, with written certification that it was done.
What is not addressed separately is the AI: nothing states what happens to a document uploaded to Summarizer, a passport image read by Extractor or a question put to Report.AI once the answer comes back, so those sit under the general customer data terms with no stated processing window of their own.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own documented permission model that the firm must administer, which is this value. The published elements are specific and tiered: role-based access controls with administrators configuring permissions at user level on every plan, and **user permission groups, multiple account admins and multiple branches on the Advanced tier**, so a firm running more than one office can separate them. Two-factor authentication is available on all plans.
On the vendor's side, production access is limited to authorized employees who require it for their role, staff with sensitive-data access must use single sign-on or two-factor authentication, and a documented password policy is enforced. The alignment burden sits with the firm, which is what places this value rather than the tier above: nothing enforces an external system's access model and no default configuration is published.
What is not addressed is separation at the matter level, or what the AI can reach. Nothing states whether the Document Assistant or Writing Assistant operate within a user's permission scope, which is the point at which a configured wall would either hold or not.
Separation between firms is architectural and documented: each customer's data sits in its own dedicated single-tenant database rather than a partition of a shared one, which the agreement corroborates by offering a copy of the subscriber's own database and stating that selected records cannot be pulled out of it. Inside a firm the product runs its own permission model, with role-based visibility, administrator-defined access, branch and location management and a conflict of interest check that scans every client and case profile.
That is a model the firm has to configure and keep aligned, not an inherited one. Nothing published describes a matter-level wall that locks a named user out of a specific case file, which is what a firm with a conflict actually has to build.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
AMENDED 12 September 2026, from not-addressed. The prior value named the 8am Terms of Service as unopened and as the document that would carry the position, and it does carry it. Terms of Service Part I permits either party to disclose the other's confidential information to the extent required by law or by order of a court, regulatory authority or other governmental body, and provides that in such cases the receiving party will provide prompt notice to the disclosing party, to the extent permitted by applicable law, to allow the disclosing party to seek a protective order or other appropriate remedy.
Both limbs this value looks for are therefore present, notice and protective-order cooperation, and no transparency report exists. The agreement names 8am DocketWise expressly as a covered Practice Management Solution. Two qualifications, and the first matters here more than on most records. The commitment sits in a mutual confidentiality clause framed around confidential information, and the agreement provides that the privacy policy governs the processing of personal information and supersedes that clause to the extent of any conflict, so the protection for client data may be narrower than the clause read alone suggests.
And a further consequence of legal process is published: where the vendor must respond to a subpoena or other compulsory process relating to the account, the customer must on request compensate the vendor for its personnel's time and costs. What the prior note said about why this matters on this record stands unchanged: the platform holds complete immigration files for non-citizen clients, including identity documents, alien registration numbers and a direct USCIS status feed, and the person whose interests turn on notice is the applicant rather than the firm.
The addendum commits Cerenade to inform the customer in writing, immediately, of any request for access to or surveillance of customer data received from any government official, and it names courts, data protection agencies and law enforcement, with the usual exception where the law forbids telling. It goes further on the data subject side: Cerenade must not respond directly to a request from an individual except where the law requires it, and the customer keeps sole control of the timing, method and content of any response.
This is a contractual commitment a firm can hold the vendor to, not a policy statement. The privacy policy separately reserves the right to disclose when required by legal process. No transparency report or record of requests received is published.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
The corpus is named by source with a maintenance commitment attached, and no licensing basis or cadence is stated, which is this value. The naming is precise and checkable: the forms library is defined by issuing agency across **USCIS, EOIR, DOL and DOS**, the vendor describes it as the largest immigration forms library in legal practice software and as the core product rather than an add-on, and commits to maintaining it for accuracy as the agencies revise forms.
The second corpus is the customer's own client record, from which every form is auto-populated, and the identity documents the Document Assistant reads. What is not stated is the basis or the cadence. These are United States government forms and therefore not proprietary, but nothing published says so or addresses licensing, and no update schedule, version history or review date is given for any form. That matters because the vendor's own support material records a live instance of the problem: the N-400 e-filing path is flagged as temporarily unavailable while it is updated for compatibility with the most recent version of the form, and a customer review complains that a new naturalization form took over a month to appear. Nothing identifies what sits behind the AI itself.
Nothing identifies a corpus because, on the evidence published, none of the four AI features answers from one. Summarizer digests the document, text or page the user hands it; Extractor reads the PDF supplied; the Translator works on the firm's own questionnaires and the government forms in the library; Report.AI answers from the firm's own case records. No external legal corpus, licensed dataset or research collection is claimed anywhere, and the estate makes no coverage claim that would depend on one. The partner drafting integrations are a different matter and are not part of this record.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history, and on this product class the question arrives in a different form worth recording. The product cites no cases, statutes or regulations. It populates agency forms from a client record, tracks deadlines and priority dates, and pulls case status from USCIS. Nothing it produces is a proposition about the state of the law that a lawyer would check for later treatment.
The analogous currency question does bite and is addressed only partially: what must be current here is the **form version**, since agencies reject filings made on superseded editions, and the vendor commits to maintaining the library for accuracy as forms change without publishing an update cadence, a version indicator or a review date. That half is graded on the corpus provenance row. Priority-date and case-status currency is handled through the direct USCIS integration and automatic notifications, which is a mechanism rather than a treatment check. Recorded so the row states the position rather than leaving a reader to infer it from the product category.
Checked the AI Workmates page, the pricing page, the integrations page and the agreement on 20 September 2026. The product works on forms, filings and case records rather than on legal authority, so no subsequent-history question arises in the ordinary way and nothing describes one. The nearest thing is Summarizer, which is offered for reading and recapping administrative appeals decisions; nothing states whether a decision it summarizes is checked for later treatment, and on this estate nothing suggests it is.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot read or map a value with confidence. The Document Assistant is described in terms of what it achieves, reading data from scanned passports and green cards and automatically importing key details to forms and client profiles, extracting names, dates and other key information and mapping structured data directly into USCIS forms. Nothing describes the other case.
No statement says that a low-confidence extraction is flagged, left blank, highlighted for review or accompanied by a confidence score, and nothing describes behavior on a poor scan, an unfamiliar document format, a non-Latin script or a field the source document does not contain. No evaluation is published. The gap is sharper here than on a drafting product because a silently wrong extraction is harder to catch than a visibly wrong one: a plausible but incorrect alien registration number sitting in a populated field looks exactly like a correct one, and the form is filed under penalty of perjury.
The surfaces read were the product pages, the pricing comparison and the vendor's own compliance and capability summary.
Nothing published describes what any of the four features does when it cannot do the job. No confidence score, no uncertainty flag, no abstention path, and no statement of what happens when the Extractor meets handwriting it cannot read, when a passport scan is poor, or when Report.AI is asked something the data cannot answer. The pages describe success only. The one acknowledgment that AI can be wrong sits in a resource piece about using AI generally and is about the firm's duty to check rather than about the product's behavior.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
Searched on 12 September 2026, on the product name with an immigration-software qualifier and on the parent's name, against published trackers of decisions on AI-generated fabricated citations including coverage of the Damien Charlotin AI Hallucination Cases database and two independent sanctions trackers. None located. This is a statement about the public record on that one subject as of the date shown, and under R119 this signal records fabricated citations and nothing else, so it is not a litigation history.
Two notes for a future reader. The product generates no citations to legal authority at all, so the exposure this signal tracks is not the exposure this product presents; its analogous failure would be a misextracted value on a filed government form, which no tracker records and which would surface as an agency rejection or a malpractice matter rather than as a sanctions order. And the parent group rebranded from AffiniPay to 8am in August 2025, so any search of the historical record needs both corporate names.
Searched on 20 September 2026 for any court order, opinion or disciplinary record naming eimmigration or Cerenade in connection with fabricated or hallucinated citations. None located. The product does not generate legal citations, which is the usual route to such a record. The vendor's own resource material discusses hallucination risk in general terms and cites the running count of decisions involving other tools.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
AMENDED 12 September 2026, from not-addressed. The prior value rested on the marketing estate; the 8am Terms of Service and the 8am IQ Usage Policy, named as unopened on that pass, have now been read and professional responsibility is engaged in general terms in both. The IQ Usage Policy, updated 19 August 2025 and incorporated into the Agreement, prohibits engaging in the unauthorized practice of law or offering tailored legal advice without a qualified person reviewing the information, and adds that the models are not fine-tuned to provide legal advice, that they should not be relied on as a sole source of it, and that all outputs should be reviewed and resources verified.
The Terms of Service require the customer to use the platform in compliance with any rules of professional conduct applicable to its profession, and state that the vendor does not provide legal advice and is not engaged in the practice of law. That is engagement in general terms rather than none, which is this value. What remains absent is any named authority: no bar opinion is cited, ABA Formal Opinion 512 does not appear, and nothing maps a feature to a rule of professional conduct in any jurisdiction.
The gap has a particular edge on this record, and it is why the value does not climb further: immigration is the one United States field in which non-lawyer accredited representatives may lawfully practice, the product is sold to legal service providers and nonprofits as well as to attorneys, and the guidance question that actually arises here, which is what a non-lawyer representative may do with an AI drafting tool, is addressed nowhere.
Professional responsibility is engaged in general terms and no guidance is named. Vendor resource material tells firms that a person must check anything legal every time, that a review step belongs in standing procedures rather than in good intentions, and that a paralegal should verify auto-populated forms before anything moves forward; a companion guide acknowledges that AI tools can produce errors or fabricate outright and notes that ethical commitments are what make lawyers hesitate.
No rule of professional conduct, no ethics opinion and no bar-issued AI guidance is named anywhere located, and nothing maps this product to any of them.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Time savings are claimed throughout and nothing addresses billing or disclosure, on a product sold to firms that bill clients. The savings claims are the commercial argument: reduced manual data entry, minimized filing errors, an average of six hours saved per day, and a customer describing a tenfold productivity increase. The fee relationship is engaged and the platform sits inside it more directly than most, because Docketwise ships the billing system too: time tracking, invoicing, trust accounting and native LawPay payment processing are published features on every tier.
So the same product both compresses the work and raises the invoice for it. Nothing published states whether AI-assisted work is identified on a bill, whether a client is told the tool was used, or how a firm should price form preparation that now takes minutes. The point has a particular edge in immigration practice, where flat-fee billing per petition is the norm rather than the exception, so the saving falls to the firm unless the fee moves; nothing published addresses that either. No cost pass-through statement was located.
The time savings are the headline and the billing consequence is never mentioned. Published claims include getting back two days a week, preparing petitions four times faster, four times faster case preparation and ten days quicker intakes, and the product sits squarely inside the fee relationship: Time Keeper tracks time at client, case and caseworker level, and the billing module issues invoices, runs payment plans and handles trust accounting.
Nothing published addresses what a firm bills for work an AI Workmate did, whether a client is told, or how a flat fee set before these features should change after them. No per-matter record of AI-assisted work is described either.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
No located public material supports a client-side disclosure obligation, though this record comes closer than most and the note records exactly where it stops. The client-facing artifact limb is met: a **public trust center at trust.8am.com** carries a current SOC 2 Type 2 attestation, the child's own surfaces link it as covering this product, and the vendor also publishes its own security page and a detailed compliance summary.
Any of those could be forwarded. The two limbs that fail are the ones this signal turns on. **No subprocessor list appears on any surface read**, and **no model provider is named anywhere**, so a firm asked which third party processed its client's passport scan could not answer from anything published. The value is not on-request either, since nothing indicates such a list sits behind a request process. Recorded as the route by which this row would move: the trust center was not opened on this pass under the sufficiency discipline and is the ordinary home for a subprocessor list, so a reading could lift this to subprocessors-listed or higher.
A current subprocessor list is published and incorporated into the addendum, dated 20 February 2026, naming five processors with the purpose of each, the processing location of each and whether it is required or optional: Microsoft Azure for hosting, HubSpot, SendGrid, Userpilot and Zoom, all processing in the United States. Changes carry advance notice and a thirty-day objection right. That is most of what a firm needs to answer a client's vendor questionnaire, and it can be read without an agreement in place.
What is missing is the other half of the pack: no model or model provider is named anywhere, so a firm asked which AI touches its clients' data cannot answer from what is published.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses disclosure of AI involvement or verification of AI output. Nothing states which values on a form were extracted by the Document Assistant as against entered by a person, no record identifies the model and no model is named anywhere, nothing captures that a human checked an extracted field, and no export, certification form or disclosure template is offered. The product does publish an eSignature capability and a form review step, and neither is presented as a record of machine involvement.
The framing this signal usually takes needs adjusting on this record and the note does it explicitly: the output does not go to a court but to a federal agency, on forms signed under penalty of perjury and accompanied by a Form G-28 identifying the attorney or accredited representative of record. **That is a certification regime, and the person certifying is the one who would need to know which fields a model filled in.** Immigration adjudicators and the agencies have begun addressing AI use in filings, so the question is live rather than hypothetical, and nothing published equips a filer to answer it.
Nothing published addresses disclosure of AI use to a court or an agency, and nothing describes a record that would support one. Case history and audit logging exist as standard features, but they record what users did to a case rather than what a model produced, and no page states that an extraction, a translation or a summary is marked as machine-generated in the file or exportable as such. This matters in a forum that has begun asking: the immigration courts issued generative AI guidance in 2025, which appears on the partner drafting vendor's estate and not on this one. Checked the AI Workmates page, the pricing page, the security page and the agreement on 20 September 2026.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- AI Governance and Bias Disclosure
- Good Law Verification
- Refusal and Uncertainty Behavior
- Court Disclosure Support
Which one fits
Choose Docketwise if
- You want a contract term, not a policy page, keeping client documents out of model training. The 8am Terms of Service that govern Docketwise state that customer content is not used to train or fine tune any model, 8am's or a third party's, and that the models behind DocketWise IQ may retain content for no more than 30 days and may not train on it.
- Your security review requires a SOC 2 report. Docketwise is covered by the SOC 2 Type 2 attestation held at the 8am level, states so on its own pages, and points to a public trust center at trust.8am.com where the current report is available. eImmigration claims no SOC 2 attestation of its own.
- You already run Clio, MyCase or PracticePanther. Docketwise names integrations with all three, alongside QuickBooks, Dropbox, OneDrive, Gmail, Outlook and Google Calendar, and publishes an open API with a developers page, so it can sit beside a general practice system rather than replace it.
Choose eImmigration by Cerenade if
- You need to know where client files sit and what happens to them when you leave. eImmigration keeps each customer in its own dedicated database on Microsoft Azure in the United States, and its agreement returns your data and then deletes it within 30 days of your confirming receipt, to NIST sanitization standards, with written certification.
- You want the vendor to carry the cost if it loses your clients' data. eImmigration's agreement carves the costs of a data incident caused by Cerenade or its subprocessors, including your notification, legal and forensic costs and any regulatory fines, out of the liability cap, and commits to notice within 24 hours.
- Your intake runs in more languages than English and Spanish. eImmigration's Translator moves questionnaires and government forms between English and twelve other languages, including Farsi, Pashto, Ukrainian and Vietnamese, and its Summarizer and Extractor come standard on the $55 Starter plan, where Docketwise's AI starts on the $99 Pro plan.
In summary
Docketwise
Docketwise is immigration case management software for immigration law firms, nonprofits and legal service providers, built around a forms library across USCIS, EOIR, DOL and DOS that auto populates from a single multilingual client questionnaire. It is based in Austin, Texas, and owned by 8am, formerly AffiniPay. The AI Legal Index grades it in the top two bands on seven of fifteen capability axes, with A grades on commercial transparency and on firm and practice coverage: prices are published in full from $69 per user per month on annual billing, and the product is declared exclusive to immigration law. Its AI, DocketWise IQ, reads scanned passports and green cards into forms and is included from the Pro tier. As of 12 September 2026 the index located no named model provider, no subprocessor list and no AI governance position.
eImmigration by Cerenade
eImmigration by Cerenade is immigration case management software sold to immigration law firms, nonprofit legal service providers, corporate legal teams and university international offices, with more than 300 USCIS, State Department and Labor Department forms under version control, electronic filing and billing. The AI Legal Index grades it in the top two bands on nine of fifteen capability axes, with A grades on data stewardship, integration depth and commercial transparency. Its subscription agreement and data processing addendum of 20 February 2026 publish a dedicated database for each customer in the United States, five named subprocessors, breach notice within 24 hours and deletion with written certification. Four AI Workmates summarize, extract, translate and report. As of 20 September 2026 the index located no named model, no accuracy measure and no AI governance position.
Questions buyers ask
Docketwise vs eImmigration: which is better for an immigration firm?
On published evidence eImmigration sits in the top two bands on nine of fifteen AI Legal Index capability axes and Docketwise on seven of fifteen, because eImmigration's own agreement and data processing addendum set out hosting, deletion, subprocessors and a data incident carve out from the liability cap. Docketwise holds a contractual bar on training with customer content and a SOC 2 Type 2 attestation through 8am. Both publish full prices, and the right choice turns on which of those a firm needs in writing.
Does Docketwise train its AI on client data?
No, and the commitment is contractual. The 8am Terms of Service, which name Docketwise as a covered product, state that customer content and AI content are not used to train or fine tune any large language model, whether 8am's or a third party's, and that the models behind DocketWise IQ may retain content for up to 30 days without training on it. The same terms reserve aggregated human and automated review of inputs and outputs to monitor accuracy. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
How much does eImmigration cost?
eImmigration publishes per user prices with no per case fees. Starter is $55 per user per month billed annually or $60 monthly, Essentials $82 or $90, and Complete $100 or $110, with extra premium features at $15 each per user per month. A standard data import costs $1,000, and nonprofits, schools and universities get 50 percent off on an annual term. Docketwise publishes its prices too, from $69 per user per month billed annually, with AI from the $99 Pro tier. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
Which one integrates with Clio, QuickBooks and Outlook?
Both connect to QuickBooks and Outlook. Docketwise also names Clio, MyCase and PracticePanther, Dropbox, OneDrive, Gmail and Google Calendar, and publishes an open API. eImmigration names 22 integrations, each with a line saying what moves and in which direction, including OneDrive, Google Drive, Box, Dropbox, Gmail, HubSpot, RingCentral and WhatsApp Business, plus filing to USCIS, State and Labor and travel history import from CBP. Its API documentation is shared on request with the Complete plan. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
What do Docketwise and eImmigration both leave unpublished?
Which AI model reads a client's passport. Neither names the model or its provider behind document extraction, and neither says what the product does when it cannot read a scan with confidence. Neither publishes an accuracy measure for extraction or translation, or anything on whether quality holds across languages, scripts and handwriting. Neither addresses how a firm charging flat fees per petition should treat work the AI did, or keeps a record of which form fields a model filled in before an attorney signs. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
Two things to weigh. Docketwise's strongest data commitments sit in the 8am Terms of Service, group terms that name Docketwise as a covered product. The Docketwise record states no liability cap or indemnity from those terms, so its liability grade describes the record rather than establishing that no cap exists. eImmigration's agreement never uses the word training; its bar on using client data for anything beyond performing the service closes that route by contract without naming it. Neither vendor names the model that reads identity documents belonging to non citizen clients. Docketwise was verified on 12 September 2026 and eImmigration on 20 September 2026. Neither vendor reviewed this page.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything above comes from public material on the dates shown. How the index grades.