N

Nextpoint

The AI is real and peripheral. The main eDiscovery product page does not mention AI, machine learning or technology assisted review. Every feature it names is deterministic or classic text processing: deduplication, email threading, near duplicate detection, OCR, hit reports, coding, Bates stamping, production and privilege log templates, and pattern based auto redaction. Privilege review is done by people, and the log is generated from their coding.

The AI consists of result ranking on models Nextpoint hosts and AI Transcript Summaries on Amazon's managed foundation model service. Without it, what customers pay for remains: unlimited data hosting at a flat per user rate from discovery through trial.

Accuracy is treated honestly at Nextpoint but not measured. Its trust page says every AI feature is tested extensively before release, and that summaries and results are starting points to trust but verify. No accuracy figure, test set, recall or precision statistic, error rate or evaluation is published behind the testing claim. Nothing says whether a transcript summary links back to the passages it rests on, which a reader would need to verify it, though the transcript itself is already in the customer's hands.

A person should review AI output, Nextpoint says, without saying where that happens. Its position is that the AI speeds review rather than replacing judgment, and that generated summaries are starting points to trust but verify. That is a caution, not a control. Nothing on the eDiscovery platform page, the trust and security page or the pricing page, or in the terms of use, describes what happens when a transcript does not support a summary.

No confidence indicator, declining threshold, correction route or in product review step is described. The platform does have tamper resistant per document activity logs and six permission levels, but nothing published ties them to the AI.

More than 600 law firms and legal teams are claimed as customers, and many are named, including McDermott, Foley & Mansfield, Jackson Walker, Seyfarth Shaw, Simmons Hanly Conroy, Clyde & Co, Williams Kastner, Yates Construction and Augusta University. Eight testimonials carry a name, job title and firm. One gives figures: a litigation support services manager at Foley & Mansfield describes migrating 180 databases, running close to 250, and much shorter training time than on the firm's previous platform.

What is missing is a measured outcome. The claims of up to 50 percent lower cost and review time cut in half are not tied to any customer, and no testimonial is dated. A case studies library is also published.

Nextpoint's commitments on confidentiality and training sit outside its contract. Its trust page says AI runs entirely inside its Amazon environment, case data never trains any model, and inputs and outputs are never shared with or kept by an outside model provider. The terms of use describe the service as a platform for privileged data and work product, confirm the customer owns its content, and limit Nextpoint's use to providing the service, with no license to improve products.

Separation is documented: six permission levels down to the document, mandatory multi factor authentication, single sign on, limited staff access with approvals, and confidentiality agreements. What is missing: the training commitment is a website FAQ, not a contract term, the terms have no AI clause, no retention period is stated for prompts or output, and the master services agreement is not published.

Nothing published addresses the advice line. The eDiscovery platform page, the pricing page, the trust and security page and the terms of use (updated June 2026) do not say that output is not legal advice, whether an attorney client relationship arises, or who is responsible for a summary a lawyer relies on. No bar authority, including ABA Formal Opinion 512, and no jurisdiction limit is named. The terms deny that information from the service is accurate or reliable, which allocates risk rather than drawing the line.

Nextpoint does say its AI speeds review rather than replacing judgment and that output should be verified, a supervision statement without a disclaimer. The platform advises on nothing, though transcript summaries are work a lawyer acts on.

An AI governance position comes with a real mechanism at Nextpoint, and bias goes unmentioned. Its trust page answers five AI questions: whether case data trains models, where documents are processed, how this differs from a consumer chatbot, whether output can be trusted, and which compliance framework covers the AI. The answer to the last is that AI features sit inside Nextpoint's SOC 2 Type II scope and its annual IT risk assessment, so model behavior is inside an audited scope.

An ungated document, Security, Compliance, Privacy and AI at Nextpoint, sets out its AI principles and safeguards. What is missing: no accountable owner is named, the stated testing before release has no published result, and nothing addresses bias or uneven output across custodians, document types or languages.

Nextpoint publishes detailed, current practice across most of the ground. Access control is strongest: six permission levels down to the document, mandatory multi factor authentication, single sign on through Okta or Azure AD, IP restricted AWS access with lockout, and staff access limited to a subset of employees with approvals and prompt revocation. Customers can export at any time, deletion follows a Data Archive Form, and retired media are wiped to DoD 5220.22-M or NIST 800-88 or destroyed.

The incident process covers daily log review, notice to affected users as soon as possible without harming the investigation, cooperation duties, and the customer's sole right to decide onward breach notice. A public status page reports incidents. Staff get background checks and annual training, and storage is redundant with eleven nines of stated durability. What is missing: no subprocessor list beyond Amazon, no retention period during the term, and no breach notice deadline.

Liability is handled by a standard limitation package, and nothing runs toward the customer. The terms of use (updated June 2026) cap liability at fees paid in the twelve months before the event, exclude indirect and consequential damages including lost data, and set a one year limit to bring claims, under Illinois law with venue in Chicago. The service is provided as is, with no promise that information from it, a generated summary included, is accurate or reliable.

The customer releases Nextpoint, with a California Civil Code 1542 waiver, and indemnifies it; there is no vendor indemnity. What is missing: no carve outs from the cap, no warranty on output, no insurance position and no AI term. The Subscription Agreement and Master Services Agreement the terms refer to are not published.

Working connections are named with what they do. Evidence can be pulled from OneDrive, Dropbox and Google Drive without manual export. Single sign on runs through Okta or Azure Active Directory, with custom SSO through Client Success, and multi factor authentication is mandatory. No connector to iManage or NetDocuments, matter management, electronic billing, court filing, Word or Outlook is listed. An API exists, mentioned only as an invitation to discuss custom integrations, with no public documentation.

Infrastructure at Nextpoint is described in detail, with residency left unaddressed. It runs on Amazon Web Services, with redundant storage across facilities, eleven nines of stated durability, checksums, survival of two facility losses, Virtual Private Cloud isolation, CloudWatch monitoring and IP restricted staff access. Its own office holds client data in a keyed office in a keyed building, with server access by password over a VPN.

What is missing: no region, residency option or tenancy model is stated, and nothing separates where data is stored from where AI runs, beyond AI running in the same Amazon environment. GovCloud appears once, about Amazon's physical controls, not as an option. The Advanced plan's custom account headquarters feature is not defined.

SOC 2 Type II, described as an annual third party audit covering security, availability, confidentiality and privacy, is held by Nextpoint. SOC 3 reports are available on request, and SOC 2 reports after a signed nondisclosure agreement. It separates its own certification from Amazon's SOC 1, SOC 2, SOC 3, ISO 27001 and FedRAMP for the underlying cloud, without claiming those as its own. A public status page, an annual IT risk assessment and an ungated overview document support it. What is missing: no auditor, report date or coverage period is given, and the SOC 2 report needs a signed agreement.

Generative AI runs on Amazon Bedrock, Amazon's managed foundation model service, which Nextpoint names. Ranking runs on models Nextpoint hosts in its own private cloud, and transcript summaries run on Bedrock in the same environment. Naming the service means its terms can be read. Bedrock does not train on customer data, does not store or log prompts or completions, and does not share inputs or outputs with model providers.

What is missing: which Bedrock model writes a summary, any model version, and a promise to notify customers when the model changes, which Bedrock makes easy.

Nextpoint publishes the full shape of its pricing and no number. Three plans are named with their features. Essential covers smaller review, production and deposition work, Advanced adds analytics, auto redaction for each document, production templates, advanced permissions and activity reporting, and Apex is a custom enterprise plan. Each user costs a flat monthly rate. The page lists what is never charged: storage, processing, hosting, OCR, deduplication, productions, exports and sharing, compared with per gigabyte fees elsewhere.

Annual prepayment discounts and volume pricing are mentioned. No figure appears at any tier, and every plan routes to a call. One discrepancy: the marketing says unlimited data, while the terms of use say plans may be user based or data based, with hosting and processing limits set in an unpublished master services agreement.

Coverage at Nextpoint is documented three ways, with no stated limits. By buyer: law firms, corporations, government agencies and education, each with a page. By specialty: construction litigation, insurance defense, mass tort and IP litigation. By use: data collection, internal investigations, FOIA requests and transcript management. File types are listed, including PST, MSG, EML and MBOX email, Office files, images, audio, video, archives, source code, CAD files and building information models, with text extraction and OCR.

It states it handles millions of documents and multi party MDL matters, reusing processed data across hundreds of related cases. What is missing: no excluded practice, size limit or unsupported jurisdiction is named; the only limit stated is that specialized formats may need specific workflows.