Trial by Error

Ever find yourself playing out a negotiation in your head, trying to predict what the other side will come back with, trying to close the gaps and put out imaginary fires before they happen? With Trial by Error you can do just that but in a safe test environment powered by AI.

Simply upload a REDACTED* version of the agreement or clause; describe who your client is (their industry, risk profile, pain points, dealbreakers, desired end position etc); and describe who the counterparty is, to build the background and context for your project. Then start marking up and submitting your suggested negotiation points. AI will simulate a role-play and respond to your comments/suggested changes with counter-points and pushbacks. AI will also rate the strength of your position and suggest any areas you may wish to do some training or professional development in.

For the seasoned lawyer: Trial by Error is the virtual colleague to bounce ideas off, to be your sense checker and the workmate who keeos you on your toes and your skills on check. For students/trainees/junior lawyers: Trial by Error is a safe training ground for developing and harnessing those negotiation skills.

*Do not upload confidential information or personal data. Parties should be referred to as "Party A and Party B" or "Supplier and Customer" not by their names, company numbers etc and contact details should be taken out. Sensitive information and commercial details such as pricing should be removed or replaced with generic placeholders such as £xxx to preserve confidentiality.

Session Setup: create a practice session by choosing an agreement type, defining client and counterparty profiles, and uploading a redacted contract document.

Clause Markup: user reviews the contract and flags specific clauses they want to negotiate, adding their position and proposed redline.

AI Counterparty: each flagged clause is sent to an AI that plays the opposing party. Responses include legally-grounded counterarguments with English law citations, a 1–5 strength rating for the counterparty's position, and a flag indicating whether the exchange is particularly valuable for training purposes.

Iterative Negotiation: user can respond to each AI reply, driving a back-and-forth negotiation on individual clauses until they accept or abandon a position.

Authentication: access is gated behind Replit Auth; each user signs in with their Replit account and can only interact with their own sessions.

Stack: React + Vite frontend, Express API, PostgreSQL via Drizzle ORM, and Anthropic Claude via the Replit AI integrations proxy for the counterparty intelligence.