Unity ELM
Unity ELM is an enterprise legal management platform for corporate legal departments, covering matter management, outside counsel spend, billing guideline enforcement, budgets and approvals in a single system of record. Released in March 2026 on Onit's unified Unity platform, it replaces the configuration-heavy approach of older ELM systems with a set of operational foundations a team can start from and extend. Two AI capabilities sit inside it. Spend Agent, generally available since late 2025, reviews invoices automatically against a department's billing guidelines, identifies errors and discrepancies, and engages with law firms to adjust invoices with limited manual involvement. Ask Unity, offered as a beta, answers plain-language questions across matter and spend data, so a team can look at quarterly spend, vendors on a matter, rejected timekeepers, accruals against budget or line items for court fees without building a report. The platform is designed to work alongside the rest of the Unity suite, including intake and the forthcoming RFP module. Unity ELM is a product of Onit, Inc., headquartered in Atlanta, Georgia, which sells a range of legal technology under the Onit, SimpleLegal, BusyLamp, Bodhala, AXDRAFT, LawBase, Legal Files and SecureDocs names and states it supports more than 3,000 customers worldwide.
Capability grades
All 15 axes, graded from public sources on the date shown. 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.
Marketed as AI-native and structurally an ELM system with two AI capabilities inside it. The operational core is conventional enterprise legal management: matter records, spend tracking, budgets, approvals and billing guideline configuration, described by the vendor as the foundations a team starts with. The AI sits on top in two named pieces, Spend Agent for automated invoice review and Ask Unity for natural-language querying of matter and spend data. Remove both and a working system of record remains, which is what places this at B rather than higher. The vendor's own framing supports the reading: Unity ELM is presented as the automate layer, with Ask Unity extending it. Pages read 1 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 addressed squarely in the agreement and never measured. Services Agreement clause 12.3 is unusually direct for this corpus: Output is machine-generated and may be incomplete, inaccurate, biased or unsuited to the customer's purpose, the customer is solely responsible for review and reliance, and the vendor states it cannot guarantee that Outputs will be 100% accurate or free from hallucinations, or that they will accurately or completely reflect facts. Acknowledging hallucination risk in a contract rather than a footnote is worth something. What is absent is any measurement: no accuracy figure, evaluation, test set or error rate appears on any surface read on 1 September 2026, and clause 12.5 provides the AI Features and Output as is with all warranties of accuracy disclaimed. Grounding is asserted at marketing level, with Ask Unity described as producing insight grounded in the governed system of record, without a described retrieval method. C.
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 product runs a genuinely autonomous function and publishes almost nothing about how it is supervised. Spend Agent is described as automatically identifying invoice errors, discrepancies and departures from billing guidelines, and as interacting with legal providers to adjust invoices with minimal human intervention. That is an agent transacting with a firm's outside counsel, which is the most consequential autonomy claim in this lane. Against it, no page read on 1 September 2026 sets out what runs unattended versus what a person approves, what thresholds trigger escalation, where a reviewer sees the agent's decisions, or how a disputed adjustment is reversed. The agreement supplies only a general allocation, clause 12.3 making the customer solely responsible for review and reliance. Ask Unity is a user-invoked query tool and raises the question less sharply. C: an approval step is implied by the phrase minimal human intervention without being described.
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.
No named customer of this product and no figure attached to it. The March 2026 release announcement states that several clients have successfully implemented Spend Agent with additional organisations preparing to deploy, which is deployment evidence without a single organisation named, a date of deployment or a measured result. Onit publishes a corporate figure of more than 3,000 customers globally across its whole portfolio, which spans eight brands and does not speak to this product. A case studies index exists on the property and was not opened, which is the rebuttal route. As at 1 September 2026, a buyer evaluating Unity ELM cannot identify one department running it or one number describing what changed. D.
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.
A real confidentiality regime and a training position that stops short. Services Agreement clause 8 runs mutual confidentiality with customer data expressly designated the customer's confidential information, and clause 8.5 requires reasonable advance notice before any legally compelled disclosure. Clause 5.1 leaves the customer owning its data. The AI terms are where it thins: clause 12.1 restricts third-party model providers from using Input to train their models for the benefit of others, but carves out de-identified or aggregated form necessary to operate the service, and says nothing about whether Onit trains its own models on customer content, while clause 5.2 separately reserves aggregated anonymised customer data for product improvement and benchmarking. Nothing addresses privilege or work product, and no matter-level or tenant segregation is documented. Clause 4.10 offers only commercially reasonable safeguards. C.
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.
A clear position on advice and nothing beyond it. Services Agreement clause 4.14 is titled No Legal Advice and states in capitals that Onit is not a law firm and does not offer legal advice, extending to employees, agents and the services, and disclaiming legal advice, recommendations, opinions, representation, referrals and counselling. It is a survival clause. What the higher bands ask for is absent: no jurisdiction limits are named, nothing addresses the buyer's own supervision or competence obligations, and no bar or ethics guidance is referenced anywhere. One clause does adjacent work and is worth naming, clause 12.7(d), which prohibits using the AI Features in any way that makes or substantially contributes to a decision with a material legal or similarly significant effect on an individual. That is a high-risk-use restriction of the kind the EU AI Act frames, not a professional responsibility position. C.
AI Governance and Bias Disclosure
Published governance over model behaviour: 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 published. No responsible AI page, no statement of principles, no named accountable owner, no pre-release testing regime and no ISO 42001 or equivalent was located across the release announcements, the Ask Unity material, the security page and the full Services Agreement on 1 September 2026. Bias is mentioned once and only as a disclaimer, clause 12.3 warning that Output may be biased, which allocates the risk to the customer rather than describing anything the vendor does about it. Clause 12.4 permits Onit to improve, replace or deprecate models or prompts at will provided core functionality is not materially reduced, which is a change right without an accompanying governance statement. The prohibition on high-risk use in clause 12.7(d) is a contractual restriction on the customer, not a governance mechanism, and is graded on UPL rather than spent twice.
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.
Strong on subprocessor governance, thin on everything the band asks about. The standout is clause 4.5: a subprocessor list at a published URL, a mandatory email update service the customer must subscribe to, a fifteen-day window to object to any newly added subprocessor, an obligation on both parties to work toward an alternative, and a right to terminate the affected services without penalty if none is agreed. That is a better change-control mechanism than most of this corpus offers. Against it, retention is unquantified: clause 6.5 gives a thirty-day export window and then deletion in accordance with Onit's then-current retention policy, a policy not published. Clause 4.10 promises only commercially reasonable safeguards. No incident or breach notification obligation was located anywhere in the agreement, which for an enterprise platform is a notable omission. The security page describes encryption, daily backup and datacentre controls. C.
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.
The general agreement is balanced and the AI is carved out of the balance. For the platform generally the position is respectable: clause 10.2 caps liability at fees paid or payable in the twelve months preceding the claim, clause 10.3 lifts indemnification obligations out of that cap, clause 11.1 gives the customer an intellectual property indemnity with itemised carve-outs, and clause 9.2 carries a performance warranty that the services will operate substantially as documented with a refund remedy. Then clause 12.5 removes the AI from all of it, providing AI Features and Output as is and as available with every warranty disclaimed including accuracy, and clause 12.3 states Onit is not liable for decisions or actions the customer takes based on Output. Clause 12.8 adds that no service level or uptime commitment applies to AI Features at all. So the thing the buyer is being sold on carries materially less recourse than the platform beneath it. C.
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.
Integration is asserted at platform level and not documented at product level. The release material makes cross-product integration the headline, describing products, agents and workflows including LSR and the forthcoming Unity RFP operating together on the unified platform so data and actions flow without additional configuration, and Spend Agent is described as fully integrated with Unity ELM. Those are internal Onit connections rather than the practice systems this axis asks about. On external systems, clause 4.4 establishes that Onit APIs exist and are licensed subject to fees and restrictions set in an Order, and clause 4.13 addresses third-party applications by disclaiming them entirely. What was not located on 1 September 2026 is any named integration with an e-billing feed, matter management system, document management system or finance platform, or any public developer documentation. C.
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 evident and neither dimension is published. Unity ELM is a subscription web application and the security page describes an enterprise-class secured hosting environment with dedicated servers, redundant power and network connections and biometric access control, which describes the data centre rather than the tenancy model. No tenancy architecture is stated, no region option is offered or named, and nothing distinguishes where data is processed from where it is stored. The agreement is silent on location: clause 4.5 permits subprocessors to host and access customer data without stating where, and clause 13.9 addresses export control rather than residency. For a platform sold to multinational legal departments across pharmaceutical, financial and insurance sectors, the absence of a stated residency option is material. C.
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.
Certifications are named with their standards and none of the particulars. The security page states Onit audits to SSAE-18 SOC 2 Type 2, SOC 1 Type 2 and HIPAA standards, verified by an independent auditing firm, and separately that physical security is audited by an independent firm, with the data centre described as SSAE-18 Type II SOC II. A trust centre exists at trust.onit.com with a controls section. What is missing is everything that would let a buyer test the claim: no auditor is named, no trust services criteria or scope are given, no coverage period or report date appears, and no route to obtain a report is stated in the agreement or on the page. The security page also reads as materially older than the AI it now covers, describing 256-bit SSL and datacentre controls rather than anything about the model layer. The trust centre was not opened on 1 September 2026 and is the rebuttal route.
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 confirms third-party models are involved and identifies none of them. Services Agreement clause 12.9 states plainly that Onit may use third-party model providers or infrastructure, and clause 12.1 addresses what those providers may do with Input, so a buyer knows external models sit underneath. No provider, model family, model or inference location is named on any surface read on 1 September 2026, and the release material refers only to Spend Agent using our model technology. Clause 12.4 cuts against the top band directly, permitting Onit to improve, replace or deprecate models or prompts without any notice obligation provided core functionality is not materially reduced. Set against that, the subprocessor mechanism in clause 4.5 does provide notification and an objection right, so if the model providers appear on that list a customer would learn of a change; the list itself sits at a published URL that was not opened, and it is the rebuttal route toward B.
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.
No pricing information published at any level. No pricing page exists on the property, every commercial route is a demo booking, and no tier, package, unit or figure was located on 1 September 2026. The agreement establishes that a charging structure exists without disclosing any of it: clause 3.1 refers to a Subscription Fee and Professional Fee set in an Order and states all fees are subject to annual increase, clause 4.7 describes usage-based tiers with automatic realignment to a higher tier and invoicing for the difference when usage exceeds a tier maximum, and clause 12.6 permits overage fees at then-current rates for sustained AI overuse after notice. A buyer therefore knows they may be moved up a tier and billed for it, and cannot learn what any tier costs. No pricing row written: there is no published structure to record, only a contractual description of mechanics.
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.
The buyer is identified clearly and the boundaries are not. Onit publishes role pages for legal operations, general counsel and contract managers, and industry pages for pharmaceutical, financial institutions, insurance, technology and government, so the intended purchaser of an ELM platform is unmistakable and the sectors are named. Unity ELM itself is positioned for corporate legal departments managing matters and outside counsel spend. What is absent is the rest of what the band asks: no department or company size segmentation is published, nothing addresses law firm use, and nothing states where the product stops. One real limit does exist and it is contractual rather than editorial, clause 12.7(d) prohibiting use of the AI Features in decisions with a material legal or significant effect on individuals. C.
Legal Signals
What each signal meansA signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The published agreement expressly reserves a right to train on customer content, with no opt out located. Any de identification, anonymisation or aggregation qualifier is recorded in the summary.
The agreement restricts third-party training and then permits a qualified form of it. Services Agreement clause 12.1 commits that Onit will not permit third-party model providers to use Input to train their models for the benefit of others, except in de-identified or aggregated form necessary to operate the Subscription Services. The carve-out is the operative part: de-identified aggregated training is permitted. Clause 5.2 separately allows aggregated and anonymised customer data to be used for product improvement, analysis and benchmarking, and clause 4.6 reserves usage data for lawful business purposes. Two gaps belong on the record as named gaps rather than as the value: nothing states whether Onit trains its own models on customer content, and clause 12.2 records that Onit retains all intellectual property rights in the AI Features, models and derivative works.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Retention is acknowledged in public materials with no stated period.
Retention is acknowledged and the period is deferred to an unpublished document. Services Agreement clause 6.5 gives the customer a thirty-day export window after termination and then commits Onit to delete customer data in accordance with its then-current retention policy, which is not published anywhere located. Clause 6.4 preserves backup and archival copies from the destruction obligation. On AI output specifically, clause 12.2 states Onit has no obligation to retain or recreate Output unless the service expressly provides retention and it is enabled, which addresses whether output persists without stating for how long anything does. No period is published for prompts, Output or platform data during the term.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
No located public material addresses walls or matter level segregation.
No located public material addresses segregation between matters, departments or tenants. The Services Agreement handles access at the level of user credentials, with clause 4.1 making the customer responsible for use under its users' logins and clause 4.3 prohibiting shared credentials, which is authentication rather than a segregation model. Nothing describes how Ask Unity bounds retrieval when it queries across matter and spend data, which is the question that matters for a tool whose stated purpose is analytical access across a department's whole book. The security page states data within Onit remains the property of the project owner and cannot be accessed by any other user, an assertion about the platform generally with no published enforcement detail. Checked the release material, the Ask Unity page, the security page and the full agreement on 1 September 2026.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Terms commit to notice where lawfully permitted. No transparency report located.
Notice is committed in the customer agreement. Services Agreement clause 8.5 permits disclosure of confidential information to the extent required by law and provides that, where permitted by law, the receiving party will give the disclosing party reasonable advance notice of the required disclosure and reasonably cooperate, at the disclosing party's expense, to obtain confidential treatment for the information, with the disclosed material continuing to be treated as confidential between the parties. The obligation is mutual and survives termination. Customer data is expressly the customer's confidential information under clause 1, so the clause reaches the material a legal department would care about. Recorded at the committed tier because no transparency report or disclosure statistics were located on 1 September 2026.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
No located public material identifies the corpus behind the product’s answers.
The product does not retrieve primary law, so there is no legal corpus to source. Ask Unity queries the department's own matter and spend records and Spend Agent reads invoices against the customer's own billing guidelines; neither returns statute or case law. No public material identifies any legal source, licence basis or update cadence, checked across the release announcements, the Ask Unity material, the security page and the Services Agreement on 1 September 2026. Recorded as not addressed because the question does not arise for this product class.
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.
No citator, and none would apply. The product analyses spend and matter data and reviews invoices against billing guidelines; it does not cite legal authority whose subsequent history could be checked. Nothing on any surface read on 1 September 2026 addresses primary law.
Refusal and Uncertainty Behaviour
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 ground an answer.
No located public material describes what the product does when it cannot answer reliably. The agreement addresses the risk by allocating it rather than by describing behaviour: clause 12.3 warns that Output may be incomplete, inaccurate or biased and that hallucination-free output cannot be guaranteed, and makes the customer solely responsible for review. That is a disclaimer, not an abstention path. No confidence signal, no documented refusal behaviour and no description of how Spend Agent handles an invoice it cannot assess was found across the release material, the Ask Unity page and the full agreement on 1 September 2026.
Fabricated Citation Record
Does a public court record exist involving output from this product?
No court order, opinion or disciplinary record naming this product has been located as of the date shown. This is a statement about the public record, not a finding about the product.
No court order, opinion or disciplinary record naming this product has been located. The AI Hallucination Cases database maintained by Damien Charlotin was searched on 1 September 2026 on the product name, the Ask Unity feature name and the parent company name, alongside general sanctions coverage, and nothing naming the product was found. This is a statement about the public record rather than a finding about the product. The product analyses spend and matter data rather than producing citations to legal authority, so the failure mode this signal tracks is not one it exhibits.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
No located public material engages with bar or ethics guidance. ABA Formal Opinion 512 is not named and no state bar opinion appears across the release announcements, the Ask Unity material, the security page or the Services Agreement, checked 1 September 2026. Services Agreement clause 4.14 states that Onit is not a law firm and offers no legal advice, which is a disclaimer rather than engagement with the professional responsibility framework its buyers work under. The buyer here is an in-house legal department rather than a firm, which attenuates the question without removing it, since in-house counsel remain bound by the rules of their admitting jurisdictions.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure.
This product sits closer to the question than most and still does not answer it. Spend Agent's entire function is billing governance on the receiving side: it reviews outside counsel invoices against the department's guidelines, flags errors and discrepancies, and engages firms to adjust invoices, and the release material frames that as replacing manual review. But under the standing treatment of this signal, outside counsel billing is a different object from a record of AI-assisted work, and nothing published describes an audit record of which invoice decisions the agent made or how they would be evidenced in a fee dispute. The published claims are efficiency claims about reducing tedious manual review. Same directional point recorded on PERSUIT applies: the buyer here is the payer, not the biller.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current subprocessor or model provider list is published.
A subprocessor list exists at a published URL and the mechanism around it is stronger than the disclosure itself. Services Agreement clause 4.5 identifies the list at onit.com/sub-processors, requires the customer to subscribe to an email update service, gives fifteen days to object on reasonable grounds to any newly added subprocessor, obliges both parties to seek an alternative and permits termination of the affected services without penalty if none is agreed. A data processing agreement is published separately. What was not established on 1 September 2026 is whether the list names the third-party model providers clause 12.9 confirms are used; the list itself was not opened and is the rebuttal route toward the top value. Recorded at the listed tier on the existence of the published list and the change mechanism around it.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses court disclosure or verification certification.
No located public material addresses producing a record of AI involvement. Nothing describes an exportable log identifying which invoice adjustments Spend Agent made or proposed, which model produced a given output, or who reviewed it, and clause 12.2 states Onit has no obligation to retain Output unless the service expressly provides retention and it is enabled, which cuts against a record existing by default. Clause 6.5 gives a thirty-day window to export customer data using built-in functionality, which is general data portability rather than a disclosure record. The product does not generate court filings, so a standing order would rarely reach it, though a fee dispute with a firm could raise the same question. Checked 1 September 2026.