I

Inlet

The models do the core work on top of a capture layer that would still run without them. A desktop application tracks work across the tools an attorney already uses, with no change to workflow, and records which applications and matters the attorney is working in. AI turns that activity into draft time entries with descriptions, assigned to the right client, in real time; Inlet's FAQ says it turns the emails, documents, calls and meetings attorneys already handle into draft entries. Without the models, a log of activity would remain, but the drafted entries the product is sold on would not.

Accuracy is the whole claim, and none of it is measured. The product writes billing descriptions and assigns time to clients from captured activity, so what matters is whether an entry describes what the attorney actually did and lands on the right matter. Customers quoted on Inlet's home page say it produces strong billing descriptions and captures their day better than they could, and Ross Albers, CEO of Albers & Associates, says that over a year Inlet captured close to a million dollars.

No accuracy rate, error rate or test method is published, and nothing describes how an entry traces back to the activity it was built from.

The system drafts and the attorney approves, and the gate is stated plainly. Inlet's home page says it drafts time entries in real time for release to Clio, MyCase, PracticePanther and others in one click. Inlet says it does not bill clients automatically: the attorney reviews and edits each draft, and nothing is sent to the billing system until the attorney approves it. Administrators see approved time and firm level reporting, not a live view of what attorneys are doing.

No limit the system places on itself is published: what happens to activity it cannot attribute to a matter, what it treats as billable, or whether entries can be approved in bulk.

Inlet names customers with figures and no method. Its home page names 19 firms: Miller Law Group, WhitsonLaw PLLC, Albers & Associates, Provinziano & Associates, Az Family Law Attorneys, Ryder & Phelps, McEvoy & Stuntz, Mansur Law Group, Merhab Robinson & Clarkson, Baker Law Group, EPGD Business Law, Youngman Reitshtein, Holtey Brown Newsom, Hera Law Group, Brick Jones, Jayaram Law, Florida Women's Law Group, Dunning Vallejo & MacDonald and Hook Law.

Nine carry extra billable hours per user per day and extra revenue per user per month, with a named leader quoted at each: Albers & Associates 0.80 hours and $5,281, Youngman Reitshtein 0.51 and $7,904, Baker Law Group 0.57 and $4,610, Dunning Vallejo 0.26 and $4,361, Florida Women's Law Group 0.45 and $4,037, Mansur Law Group 0.69 and $3,714, EPGD Business Law 0.28 and $2,644, WhitsonLaw 0.36 and $2,517, and Merhab Robinson 0.34 and $2,436.

How those figures were calculated, over what period and against what baseline is not stated, so they cannot be checked. A Y Combinator job post describes the company as past product market fit and growing fast, without figures.

Inlet publishes real commitments on collection and training; privilege itself is not addressed. The privacy policy (version 1.3, last updated 1/7/2025) says Inlet does not record or store the content of what a user types. It collects calendar event metadata (titles, times, attendees), email metadata (subjects, senders, timestamps) and call and meeting metadata (participants, times), deletes activity data automatically once time entries are created, and does not store data read from integrations after entries are made.

It says the AI providers work under zero data retention agreements and do not use the data to train models. Inlet's FAQ says customer data is not used to train AI models, and administrators see only approved time. Two limits apply. Nothing addresses privilege or work product, although email subjects and meeting titles, which the policy names as collected, often reveal privileged matter. And the home page's promise that data is shared with no one but the customer sits beside a privacy policy and terms that describe sharing it with AI providers.

The product gives no legal advice, so the unauthorized practice question largely does not arise; the professional responsibility question is billing, and it is not addressed. The central promise is that firms find 15 percent more billable hours, and Inlet says firms see an increase in billable hours within days of turning it on. That raises whether time surfaced from an activity stream was properly billable to that client, and whether descriptions a machine wrote accurately describe the work.

Inlet's home page says it operates under strict compliance with legal ethics guidelines and carries a Legal Ethics Compliant badge, but names no rule, opinion or guidance and says nothing about fee reasonableness. The website terms carry no advice disclaimer or eligibility rule.

Inlet claims compliance without a mechanism anyone could audit. It says it follows SOC 2 security controls, operates in strict compliance with legal ethics guidelines, and does not train on customer data. Nothing names who is accountable for the AI's output, how entry quality is tested before release, or how accuracy is checked across practice areas and firm sizes. The company is small: its Y Combinator profile lists a team of 2.

Inlet describes its own data handling, short of a subprocessor list, an incident commitment and a consistent account of retention. The privacy policy says the application collects activity signals and email, calendar and call metadata only as needed to create time entries, does not record typed content, deletes activity data once entries are created and does not store integration data after processing. The home page adds end to end encryption, a Zero Data Sharing Policy badge, SOC 2 controls and enterprise grade security.

The AI providers are not named, no subprocessor list is published, and no incident notice term appears. The website terms say entered text may be shared with third party APIs that keep it for 30 days and then delete it, and that no vendor keeps it longer, which does not match the privacy policy's zero data retention account. The privacy policy also allows personal information to pass to a buyer in a merger, asset sale, financing or acquisition.

Nothing published says who bears the loss when a drafted entry is wrong. The published terms, effective 25 November 2025, are website terms from Inlet, Inc. covering conditions of use, copyright, communications, reviews, site access, accounts and data use, under United States law. They contain no warranty, limitation of liability or indemnity for the service, and no customer agreement is published. Inlet reserves the right to terminate accounts, edit or remove content and cancel orders at its sole discretion.

Inlet names real integrations in both directions, short of documentation. Approved entries are released in one click to practice management systems: Inlet's home page names Clio, MyCase and PracticePanther, and Inlet also lists Filevine and Aderant. The application captures work in Outlook, Gmail, Word, Acrobat, Google Calendar, Zoom, Microsoft Teams and Slack, reading calendar, email and call metadata. No integration documentation or help center is published, and nothing says how matters are matched to the practice management system, what each connection reads and writes, or what an administrator must configure.

Inlet runs as a desktop application with a web sign in at app.inletai.com, and nothing is stated about where data is processed. The privacy policy says information is processed at the company's offices and wherever the parties involved are located, and may be transferred outside the user's jurisdiction. No hosting provider, region or tenancy model is named.

Certification is claimed without detail. Inlet describes itself as SOC 2 certified and following SOC 2 controls, and its home page shows SOC 2 Security, End to End Encryption and Zero Data Sharing Policy badges and promises that every piece of data is protected end to end. No report type, auditor or period is named, and there is no trust center or route to the report.

Inlet discloses that it uses outside AI providers and names none. The privacy policy says activity information is shared with AI service providers under zero data retention agreements, solely to create time entries, and the website terms refer to third party APIs that keep entered text for 30 days. No provider, model or hosting arrangement is identified, and nothing commits to notice of change.

No pricing is published at any level, not even the unit of charge. Inlet offers a 14 day free trial, started by booking a demo, and every route leads to a demo booking. The website terms mention canceling orders but set no fees.

Inlet describes its scope concretely, short of stated limits. The product serves law firms that bill by the hour, captures work in eight named applications and releases entries to five named practice management and billing systems, which in practice bounds who can use it. Y Combinator's profile and job post describe the customers as small and mid sized law firms, though Inlet's own site states no firm size. Nothing says which operating systems or billing arrangements it supports, or what it does for flat fee or contingency work.