Expert-led analysis, faster reports
Report Builder loads the record, runs the data and drafts the prose. Your expert makes every analytical decision, in their own words, and every number reruns from the data.
Our goal: 50% lower expert costs.
The expert decides. The builder does the rest
Most of a report’s cost is assembly: data, first-pass models, drafting. Report Builder takes that on. The judgment stays with the person who signs.
The record, loaded
Productions with Bates ranges, pleadings, counsel’s stated assumptions, and live data through your own licences.
BuilderEvery decision, the expert’s
Windows, benchmarks, controls, market definitions. The builder sets out options with their assumptions and limits; the expert chooses, changes or replaces them and says why.
ExpertProse drafted from decisions
Each sentence shows who wrote it. The expert’s reasoning goes in as written, never paraphrased.
Builder, expert approvesChecked, then signed
Every figure reruns and is recalculated independently. Signing stays locked until every decision is the expert’s.
ExpertTry the expert’s workspace
Two illustrative matters: a securities event study and an antitrust overcharge regression. Pick an option, write your reasoning, attach support and record the decision: the draft, the ledger and the sign-off checks update as you go.
Testifying expert
Your analysis
- 1Record and assumptionsReviewed by you
- 2Market efficiency4 decisions made by you
- 3Event studyNeeds your decisions
- 4Corrective disclosuresNot started
- 5Inflation per shareNot started
- 6Damaged sharesNot started
- 7Robustness and validationRuns as decisions are made
- 8Report and sign-offLocked until all decisions are yours
Record from counsel
Counsel adds documents and states assumptions. Only you change the analysis.
Event study specification
This step measures how much of each price move is specific to the company. The AI proposes a specification and alternatives, with their assumptions and limitations. You accept, change or replace any of them.
Decision 3.1 · Estimation window
Accept, change or replaceWhy this length; company events inside the window; the alternatives you considered and why you set them aside.
Support for this choiceData
Daily prices through your firm’s own market-data licence, saved as a dated snapshot.
Decision ledger
Kept for Rule 26| Decision | AI proposal | Your choice | Results seen first | Your reasoning |
|---|---|---|---|---|
| 2.1 Market efficiency | All five factors met | Agreed | Not applicable | Weekly volume above 2%; 14 analysts… · 3 sources |
| 2.4 Thin-trading stretch | Exclude August weeks | Changed: kept them | No | Volume was low but spreads stayed tight… · 2 sources |
| 3.1 Estimation window | 120 trading days before the class period | Waiting for you |
Report draft · section 4.2
I estimated the stock’s expected return using a market model fitted over 120 trading days before the class period, with a broad market and a sector index as controls. Your reasoning appears here, as you wrote it. The result sentence is drafted once you record the decision.
Every sentence links to the data, the code and the decision it rests on.
Sign-off checks
Opens when every check passes. Signing is the expert’s statement that the opinions are theirs.
Before your deposition
Practise explaining this step. A questioning agent probes your window and index choices; you answer in your own words.
Built for Rule 26 and Rule 702
Courts ask whether the expert prepared the report and whether the method was reliably applied. The builder keeps the record that answers both, and checks each choice against the ways expert analyses have failed before.
Decision ledger
Each key decision: the options, the AI’s proposal, the expert’s choice, their reasoning, sources and time.
Rule 26(a)(2)(B): prepared and signed by the witnessChecked against past exclusions
Each choice is compared with the reasons courts have excluded similar analyses before, so weak points surface while they can still be fixed.
Built on our library of Rule 702 rulingsResults hidden until decided
The expert can choose before seeing which option moves the number. The ledger records whether they looked.
Answers the outcome-driven attackReproduced and validated
Dated data snapshots and versioned code rerun every figure; a second implementation checks it.
Rule 702: reliable methods, reliably appliedAuthorship on every sentence
Written by the expert, or drafted by AI and approved. Nothing in between.
Clear answers at depositionCounsel stays outside the analysis
Counsel adds documents and states assumptions, labelled as theirs. Only the expert changes the analysis.
Keeps the opinions the expert’sA report that has already answered the other side
When the builder knows most of what the opposing expert will argue, your expert can test their likely specification, explain why it falls short, and close the gaps before the report is served. Plaintiff side, close certification fight, the same matters as on Expected Gains.
Securities class action. $490M plaintiff-style damages, close certification fight.
| Your expert’s opening report | Without prediction | With Supreme Mind today | Ceiling |
|---|---|---|---|
| Opposing expert’s key attacks answered in advance | 1 of 6 | 3.5 of 6 | 4.7 of 6 |
| Chance of exclusion in whole or part, if challenged | |||
| Contested points the court decides your way | 52% | 56% | 58% |
| Reply report drafted before theirs arrives | Little | About 60% | About 80% |
| Expected recovery | $22.5M | $24.9M | $26.1M |
The stronger report is how the Expected Gains value is delivered, not money on top of it. What Report Builder adds on top is the cost goal: half the cost of preparing the report.
Modeled estimates, not measured outcomes. Attacks answered: the share of the opposing expert’s key findings anticipated, measured on our Benchmark for securities and estimated for antitrust. Exclusion rates without prediction: PwC Daubert study (financial experts, 2022) and Peruzzi (antitrust economists, 2024). We assume an attack answered in the opening report carries half the exclusion risk of one that is not, and that a court sides with you 10 points more often on a point you answered. Pilots will measure both and replace these figures. No result is guaranteed, and no report can make exclusion impossible.
Where the time goes
The saving comes from support work, not from the expert’s judgment. The pilot measures it on real matters, against comparable ones.
| Work on a report | Today | With Report Builder |
|---|---|---|
| Data assembly and cleaning | Consulting team | Mostly done by the builder |
| First-pass models and runs | Consulting team | Proposed by the builder |
| Key decisions and reasons | Expert | Expert |
| Drafting | Team and expert | Drafted from the expert’s decisions |
| Review and sign-off | Expert | Expert, with independent checks |
Questions firms ask
Does AI write the report?
AI drafts prose from decisions the expert has made and proposes methods the expert can take, change or reject. The analysis and the opinions are the expert’s.
Which practice areas?
Securities event studies and antitrust overcharge analysis first; damages, valuation and forensic accounting follow.
Who can see what?
Counsel sees the record and the draft; the expert’s workspace is theirs. Matters are walled, and nothing trains our models.
What does a pilot involve?
One matter, one expert, a few months. We measure hours and cost against a comparable matter and share the result with you.
Pilot it on one matter
We are enrolling a small number of firms and experts. You get the builder at pilot terms; we measure what it saves, with you.
Report Builder is in a pilot programme. The 50% figure is our goal, not a measured result; what a firm saves will depend on the matter and the expert. Supreme Mind does not give legal advice or act as an expert witness. See our Terms.