M

MyCase

MyCase is a fifteen year old practice management system with an AI layer added, and the vendor is open about which is which. Cases, contacts, calendars, documents, time entry, invoicing, trust accounting and the client portal are conventional software and make up the whole Basic tier, which has no AI at all. 8am IQ appears at Pro (Writing and Document Assistants) and Advanced (adding Case and Discovery Assistants), priced into the tier rather than sold separately. Remove the models and the system a firm runs its practice on is untouched.

Grounding is real, documented and described in mechanical terms. Case Assistant answers questions across a case's documents, notes, calendar events, invoices, transactions, tasks, case fields, portal messages, texts and call logs, and the vendor says every answer includes a source you can trace and verify before acting. The Terms set out the retrieval method: a case's documents and data are converted into a vector database by an embedding process, case by case, and then retrieved by the language model.

The vendor says it runs proprietary metrics to measure accuracy and detect hallucinations, and limits the surface to preset prompts for each task. None of that measurement is published. The metrics are proprietary, no figure, test set or error rate appears, and the Terms say output may contain errors, omissions or inaccuracies with no warranty of accuracy.

Nothing runs unattended, and the vendor says so directly: data is never stored or analyzed passively and is processed only when a user asks a question. Control sits at two levels. At firm level, the account owner manages a firm wide AI setting, and turning it off removes access for every user; the Terms make the firm responsible for setting it in line with its own policies and professional obligations. At user level, 8am IQ follows existing permissions, so a user cannot reach anything through it they could not already open.

Review is built into the output through source citations, and the Terms require human review, saying output may not replace professional judgment. No thresholds or escalation path are published, but the product is a query tool a user invokes, not an agent, so routing back to a human hardly arises.

A named firm with a figure, and portfolio metrics with a stated method. Vi Nanthaveth of Nanthaveth & Associates, PLLC is on record describing a move from meeting three or four people a day to fifteen. The headline metrics carry a method footnote. The figure of 64 billable hours recovered per year comes from an analysis of anonymized MyCase customer data for firms using the Smart Time Finder feature, averaging 5.3 hours a month.

The ROI calculator's revenue figure is built from that number plus the user's own inputs and is labeled an estimate, not a guarantee. Nothing is dated, the other testimonials use first names only and read like review site extracts, and the 37% caseload figure has no stated basis.

The training question is answered in the agreement, not on a marketing page. Part III says 8am does not use Customer Content or IQ Content to train or fine tune any large language model or other generative AI model, its own or a third party's, and that the substance of inputs and outputs is never used to update model weights. Retention is specific: the models used for 8am IQ may keep IQ Content for up to thirty days, and case data in the vector database is deleted automatically after thirty days of inactivity.

Separation follows the platform's existing permissions: 8am IQ cannot reach any record the user cannot already open, and the vendor says safeguards prevent exposure between customers. What is missing is any express treatment of privilege and work product, for a system that indexes a firm's whole case file.

A full treatment, and all of it sits in the agreement rather than a footer. The vendor states plainly what it is not. Part I says 8am does not give legal or accounting advice and is not engaged in the practice of law. Part III says 8am IQ does not give legal, accounting or other professional advice and that output may not replace professional judgment or human review. Use is scoped by jurisdiction: the Platform is for legal, accounting and other client based professionals in the United States and Canada.

Competence and supervision are addressed. Using 8am IQ to practice law without authorization, or to give legal advice without a qualified person reviewing it, is expressly prohibited. The customer must use the Platform in line with the professional conduct rules of its profession, and the firm wide AI setting gives the account owner a way to enforce that. The disclaimer also recommends consulting qualified counsel on compliance. Disclosure to clients does not arise, since 8am IQ is an internal tool for firm staff, not an agent that faces clients.

A published framework with real substance, and no results behind it. A dedicated 8am IQ Usage Policy is incorporated by reference into the Agreement, a governance document rather than a marketing page, and the vendor publishes principles for building AI in legal technology. The mechanism is described: proprietary evaluation metrics to measure and monitor accuracy and detect hallucinations, preset prompts for each task to make accurate answers more likely, and a commitment to continued testing.

The firm wide AI setting gives governance a control. What is missing: no person or committee is named as accountable for the AI, no testing before release is described in terms a buyer could audit, no evaluation result is disclosed, and nothing addresses uneven output across matter types, practice areas or populations.

The policy is substantive across most of the ground; the subprocessor list is the notable gap. Retention is stated in numbers: thirty days at the language model layer, and thirty days of inactivity before embedded case data is deleted automatically, with the embedding run again on next use. Encryption is 128 bit SSL in transit and 256 bit AES at rest. Access control works through permissions each user inherits. A security and privacy whitepaper is published at an open link, and a trust center exists at trust.8am.com.

Incident practice appears in the Terms, but one sided: the customer must report a data incident within twenty four hours to a named security address, while 8am keeps sole say over whether and how individuals or regulators are notified. Not found: any subprocessor list, or any statement of where data is hosted.

The terms are specific and readable, and point entirely one way. Total liability is capped at the net fees 8am earned in the three months before the event behind the claim, and all indirect, incidental, special, consequential and punitive damages are excluded. The Platform is provided as is and with all faults, and 8am expressly refuses to warrant that information, data or AI output will be accurate, complete, reliable or compliant with law.

Part III adds an AI exclusion: 8am is not liable for any damages from errors or inaccuracies in output or from relying on it. The only indemnity runs from the customer to 8am, and claims must be brought within one year. There is no indemnity for the customer, no warranty on output and no insurance position.

There are real integrations, named partners and public developer documentation, short of described depth. The pricing matrix lists more than 70 integrations at Pro and an Open API at Advanced, and names connected tools in the feature comparison: Google, Outlook, Microsoft Office, QuickBooks, Zapier, LawToolBox for court rules calendaring, and Smith.ai. Payments run through LawPay in the same platform, and Desktop Drive syncs files both ways between a user's desktop and MyCase.

A public developer site at developers.8am.com has a quickstart guide and API reference that an implementer can use. What each integration moves, in which direction and what a firm must set up is not described on any page read. The dedicated integrations page was not opened and may say more.

The service is clearly delivered from the cloud, but neither tenancy nor residency is published. No tenancy model is stated, and no data residency region is offered or named. The nearest thing to a location statement describes who the product is for, not where it is hosted: the Terms say the Platform is for professionals in the United States and Canada. Nothing separates where data is processed from where it is stored, and no hosting provider is named on any page read, though the force majeure clause refers generally to failures of cloud hosting providers. The trust center at trust.8am.com was not opened and may say more.

Certification is stated and a trust center exists, without the details that would let a buyer test it. 8am says on the 8am IQ page that it is SOC 2 Type II certified and links the claim to the AICPA. It runs a trust center at trust.8am.com, linked from the footers of mycase.com and 8am.com, and publishes a security and privacy whitepaper at an open Google Drive link with no form, NDA or sales call. The Terms also commit to PCI-DSS compliance for payments.

What is missing is the attestation detail: no auditor, trust services criteria, scope, coverage period or report date appears on any readable page. The trust center was not opened, so whether reports can be downloaded there is not known.

The architecture is described in detail, but the providers are never named. The Terms say 8am IQ runs on large language models and describe the pipeline: case documents and data are converted by an embedding process into a vector database, case by case, for retrieval by the model. They also say what the model layer does with the data: the models used for 8am IQ may keep IQ Content for up to thirty days, and that data will not be used to train any model.

That is a real statement about what the underlying provider may retain. But no model or provider is identified. Claude appears in the Terms only to be excluded, in a clause saying third party services including Claude are not 8am AI features and are governed by the customer's own agreement with them. No subprocessor list and no commitment to notify customers of changes was found.

A buyer can price this for a whole firm without talking to anyone. All three tiers have published rates at both billing frequencies: Basic $50 annual or $60 monthly, Pro $100 or $120, and Advanced $130 or $150, all in US dollars per user per month, with the annual saving shown for each. The unit is explicit, and the FAQ repeats that the rate depends on the number of active users. A full feature comparison shows what each tier includes, including which AI assistants sit where.

Implementation is covered: guided implementation, training and support carry no extra cost, and there is no long term contract and no setup fee. Add-ons are priced too: MyCase Accounting at $39 per user per month and LawPay with no monthly fee. Even AI overage has a rate: $30 for 5,000 more Discovery Assistant pages, against an allowance of 5,000 pages per user per month pooled across the firm. A ten day trial needs no card.

Segment and practice coverage are described in substance, and the limits are only partly drawn. Six practice areas have their own pages (criminal defense, family, immigration, bankruptcy, personal injury, and trust and estate) with a further index behind them, and the product is aimed throughout at small and midsized firms. One real limit is stated, and it is jurisdictional: the Terms scope the Platform to legal, accounting and other client based professionals in the United States and Canada, and a separate clause covers use by the US Government.

Practice depth shows in features such as conflict check tracking, statute of limitations deadlines, trust balances by case and court rules calendaring. No firm size segments are published, in house and government legal departments are not addressed as buyers, and nothing says which practice types the product handles poorly.