AI Readiness Assessment
When: You want to know where AI would pay off in your practice and where it would add risk.
What you receive: An inventory of current and candidate uses, ranked by value and risk, with a recommended first project.
For Law Firms, Legal Departments, and Service Providers
AI in Your Practice, Built to Hold Up
Any firm can buy AI tools. The hard part is showing a client, a court, or a malpractice carrier how the work was checked. I help legal teams build AI workflows with measured accuracy, verification that leaves a record, and clear points where a lawyer signs off.
Thomas Reichert is an Assistant Professor of Law at SIU Simmons Law School, a registered patent attorney (USPTO Reg. No. 77,326), and the founder of MarkSense Analytics. Reichert has consulted with legal service providers on AI and analytics offerings and chairs the law school’s Ad Hoc AI Committee.
Who This Is For
For Organizations That Answer for Their Work
Law Firms
Partners who want AI in their practice groups and need it to meet the standard they already hold associates to.
Corporate Legal Departments
General counsel and legal operations teams deciding which work to bring in-house, which tools to buy, and how to govern them.
Legal Service Providers
Providers building AI and analytics products for law firms and legal departments, whose output has to survive a lawyer’s review before it ships.
The Method
Four Controls That Make Legal AI Defensible
The four controls come from my own research, where a wrong answer ends up in print under my name. In the DuPont work, I checked the model’s coding against my own hand review of 1,002 factor-level findings; they agreed 97.11% of the time (one reviewer, not blind to the model’s coding).
-
Structure the Inputs
The model works from a closed set of sources you control: the record, the file, the governing authority. Everything it says has to trace back to something in that set.
-
Measure Before Trusting
Before a workflow goes live, it is tested against work your own lawyers have already done, and its accuracy is recorded. That number decides where the workflow can be used.
-
Verify With a Record
Every citation and factual claim is checked against its source, and the check is logged: what was checked, how, and with what result.
-
Assign a Lawyer to Sign Off
Each output has a named lawyer who reviews it and owns it. The depth of that review is set by the risk of the task.
Engagements
Ways to Work Together
The usual starting point is an assessment; the other engagements build on what it finds. Each one is scoped in writing before it begins, and the written engagement agreement sets whether the work includes legal advice.
Workflow & Verification Design
When: A use case is chosen and needs to run safely at volume.
What you receive: A documented workflow with source controls, accuracy targets, a verification log, and defined review and sign-off points.
Tool & Vendor Evaluation
When: You are choosing among legal AI products and every demo looks good.
What you receive: A test built from your own completed work, run against each candidate, with results you can compare side by side.
AI Policy & Governance
When: Lawyers are already using AI, and the organization needs rules that match how the work is actually done.
What you receive: A usage policy, an approval process, and a client-disclosure approach mapped to ABA Formal Opinion 512 and the rules of your jurisdiction, coordinated with your own counsel in states outside Illinois and Missouri.
Training for Lawyers & Staff
When: A rollout needs people who know where the tools are strong and where they fail.
What you receive: Sessions built around your own workflows, with exercises on catching errors. For CLE programs and keynotes, see Speaking.
Custom Data & Model Work
When: Your organization holds decisions, filings, or matter records that could answer questions about outcomes, pricing, or risk.
What you receive: A structured dataset built from that material, validation against hand-coded examples, and fine-tuned models where the work justifies them.
Built From the Research
Alongside advisory work, MarkSense Analytics, the company I founded, offers a scoring platform built from the data behind my research. It measures the similarity of trademarks and scores likelihood of confusion against the record of decided cases. Scores describe how comparable decided cases came out; they do not predict how any particular matter will be decided.
The same method can be applied to other multifactor tests, such as fair use and patent obviousness.
Ask About MarkSense- Grounded in published research. Built on the DuPont study’s dataset of about 4,000 Trademark Trial and Appeal Board decisions.
Background
Why Organizations Bring Me In
Research Tested in Public
My AI-driven study of about 4,000 trademark decisions, Doctrine, Data, and the Death of DuPont (opens in new tab), appears at 36 Fordham Intell. Prop. Media & Ent. L.J. 678 (2026). MLex, World Trademark Review, and The Fashion Law covered it, and the cert petition in World Champ Tech v. Peloton, No. 25-736, cited it. I am also the sole author of an amicus brief in RiseandShine Corp. v. PepsiCo, Inc., No. 24-1016, filed July 20, 2026, after the Supreme Court granted certiorari.
Practice Experience
I practiced IP law at Kutak Rock (2018 to 2022) and as a patent attorney at Bookoff McAndrews (2023 to 2024), advising clients from startups to Fortune 5 companies.
A Grounded Tool, Measured
I built The Fair Use Database (opens in new tab), which codes every substantive federal fair use opinion. Its research assistant, Folsom, answers only from the coded corpus and cites its evidence, and its outcome coding agrees with 453 human-coded analyses 97.4% of the time.
Talk Through Your AI Plans
Tell me what your organization wants from AI and where it worries you. I typically reply within one to two business days.
Start a ConversationOr email tom@reichert.law.