Gavel Exec Launches Deep Reasoning Mode – Artificial Lawyer

Gavel has launched Deep Reasoning Mode in Gavel Exec, its Microsoft Word add-in for AI-powered contract redlining, drafting, and negotiation. It uses multiple genAI models, including GPT-5.

Los Angeles-based CEO, Dorna Moini, told Artificial Lawyer: ‘Deep Reasoning Mode combines models trained for legal documents with Gavel’s proprietary legal AI. Trained by corporate and real estate transactional lawyers, Gavel Exec has a deep understanding of how lawyers redline and those practice areas, boasting an 80% redline acceptance rate.’

AL then asked about these models. Moini explained: ‘Gavel Exec uses a network of AI agents to respond to user prompts, each using different models for specific tasks, like understanding document structure, context and style, planning work, generating precise redlines, drafting new sections, answering questions.

‘We have licensed lawyers continuously benchmarking models for task-specific applications. Currently, we’re using some of our own proprietary models for document and context understanding, along with GPT 4.1, GPT 5, and Claude for other tasks. It’s important to note that many of these models have tradeoffs (such as context window, reasoning capabilities, speed, factual accuracy), and they do not individually provide the best quality when used for contract analysis and redlining directly inside Microsoft Word.

‘Specifically, we found that GPT-5 has better logical reasoning and higher factual accuracy, but it is limited in its ability to work with multiple tools and has a relatively small context window, so you cannot just expect to write a ‘GPT-5 wrapper’ and expect it to work out of the box. Instead, we swap models for each task – like redlining, comments, high-level reasoning and planning, and running playbooks based on the performance we see from each model at each task, using our benchmarks created by expert lawyers. This allows Gavel Exec to adopt the latest models quickly for the applications where they yield the best outcomes for customers.’

And in terms of Deep Reasoning’s context window – how large is this? 

‘While each underlying model still has its own context window limit (for example, GPT-5’s is smaller than some others), with Gavel, there’s no practical limit on the size of the project or the number of documents you can include. We’ve built Gavel Exec to break up the work intelligently and understand the full context of even the longest documents behind the scenes, so you can run reasoning across hundreds of pages (with reference to many additional documents outside of the one you’re negotiating) without being constrained by an individual model’s context window.’

And here’s an overview of what Deep Reasoning Mode delivers:

‘Surgical Redlines with Better Jurisdiction-Specific Edits: Context-sensitive clause edits that balance legal accuracy with negotiation strategy, avoiding overbroad or irrelevant changes.

Customizable AI with Large Context Windows: Gavel Exec was the first out-of-the-box AI platforms to allow firms to train and customize their own models. Deep Reasoning now provides bigger context windows, allowing lawyers to feed it a large set of precedent agreements and reference documents to create consistent AI for the entire firm.

Contract Analysis Across Lengthy Documents and Multiple Documents: Has an even better understanding of the entire document when suggesting changes and allows you to upload other documents to understand risks, deviations from market standards, and negotiation priorities across a set of documents.

Multi-Step Reasoning: Handles complex instructions that require sequencing multiple actions, such as benchmarking a clause, suggesting revisions, and then generating a client-ready explanation of the change.

Smarter Summarization: Tailors summaries of the document or the existing redlines in a document to the specific use case, whether preparing a client summary or internal negotiation noted.

Moini concluded: ‘Lawyers don’t want to spend time figuring out which AI model is right for which task, whether that’s GPT-5 or another model. They just want accurate, high-quality work product, and that’s what we help determine for them.’

—

Legal Innovators Conferences in New York and London – Both In November ’25

If you’d like to stay ahead of the legal AI curve….then come along to Legal Innovators New York, Nov 19 + 20, where the brightest minds will be sharing their insights on where we are now and where we are heading. 

And also, Legal Innovators UK – Nov 4 + 5 + 6

Both events, as always, are organised by the awesome Cosmonauts team! 

Please get in contact with them if you’d like to take part.