BlackBoiler, a company that has spent over a decade building automated redlining technology, this week launched Veris, a new platform that takes its original deterministic editing engine and supercharges it with generative AI and an agentic, chat-based interface.
Running directly inside a Microsoft Word add-in, Veris allows contract-review teams to negotiate and mark up agreements without ever leaving the document.
Alongside the launch, BlackBoiler is rolling out two new subscription tiers aimed at expanding its reach beyond its traditional enterprise base. The Starter tier, designed for solo reviewers, runs $1,250 a year. The Pro tier, built for recurring team reviews, costs $3,000 per user per year.
In a demonstration for LawSites, Daniel Broderick, co-founder and CEO of BlackBoiler, said that Veris was developed as a response to a recurring request from customers. While they valued the precision and consistency of BlackBoiler’s data-driven editing, he said, they wanted it delivered through a more interactive experience and with faster setup.
“They wanted that in a more agentic experience with faster onboarding,” he said.
Combining Two Approaches
BlackBoiler’s original system predates the current generation of large language models (LLMs). Instead of relying purely on predictive text, it uses an organization’s historical data — actual examples of past contract markups — to drive its edits.
Veris retains that deterministic foundation while incorporating LLMs “where necessary,” Broderick said. It includes strict controls over what data is sent to the models, backed by a robust validation layer to fact-check the output.
That validation process is a core featue of Veris. The system statistically analyzes every suggested edit, measuring how dramatically a sentence changes and tracking specific word additions or deletions. It then cross-references those changes against similar edits BlackBoiler has processed in the past.
The goal is to limit hallucinations and to “only edit what needs to be edited and not more than what needs to be edited,” Broderick said.
Robert Moore, BlackBoiler’s director of sales, said the determinism is what distinguishes Veris’s approach from using an LLM alone.
He noted that two people giving the same input to a general-purpose model can receive different outputs, whereas Veris grounds edits in a company’s own standard. A “judge” component validates a suggested edit by tracing it back to the examples a customer provided.
Automating Playbook Setup
The faster onboarding that Veris enables comes largely from its ability to automate the work of building a playbook — the master set of rules that governs how an organization wants contracts edited. Broderick said the company has automated a curation step that previously required involvement from humans employed by BlackBoiler.
In the demonstration, Broderick built a playbook by uploading a single marked-up contract and having the system extract rules from it. Users can also do this by uploading a policy document or a written description of how they want to handle specific risks, or they can simply describe a rule directly.
In the demo, Veris pulled roughly 20 rules from the sample document, displayed matching rules from BlackBoiler’s master rule libraries where they existed, and let the user accept, reject or revise each one.
From there, Veris runs an “enhancement loop” entirely in the background. For each rule, Veris:
Generates a prompt and a corresponding judge.
Searches BlackBoiler’s database for similar clauses.
Applies the prompt to edit those clauses.
Uses the judge to evaluate the results across multiple examples.
Refines both the prompt and the judge automatically.
That approach, Broderick said, removes the human variable from prompt engineering. Because different lawyers would inevitably write different prompts and get different results, Veris relies on the premise that “the data should build the prompts.”
Users can upload up to 20 contracts through the app, he said, with larger volumes handled offline to avoid timeouts. The prompting and judging can be run against BlackBoiler’s data or a customer’s own data.
Two Review Modes
Veris offers two ways to review a document, reflecting the different ways users prefer to work.
A “full review” inserts all suggested edits directly into the contract as tracked changes. Broderick said this suits intake-driven pipelines where a document is routed to an attorney already marked up.
A “quick review” places suggested revisions in the margin for the user to insert one at a time, ordered either by document position or by risk level.
Users can also interact with a document through a chat interface — for example, instructing it to change the governing law to a particular state — and can save such instructions as new playbook rules on the fly. Playbooks can be scoped to an entire organization or to specific users, depending on access.
Bottom Line
When it comes to legal AI adoption, validation remains a major hurdle. BlackBoiler says Veris is designed to squarely address that issue, pairing the creative power of gen AI capabilities with a determinisitic layer to provide constraints and checks.
“Instead of relying on each user’s prompting skill, Veris derives prompting standards from the edits and review behaviors that define how an organization negotiates,” Broderick said.
Because the product uses that same historical foundation to evaluate generated text before it ever becomes a final work product, Broderick believes Veris represents where the industry is heading.
“The next phase of contract AI will be shaped by consistency, governance, and cost-efficient execution,” he said, “not just language generation.”
