By Katya, Head of Product, Avvoka 

Most law firms already know which documents they'd automate if they had the time. The list sits with the knowledge team and it's usually far longer than the list of templates that are live. 

That gap has little to do with ambition. Building an automated template – the questions, the logic behind them, the clauses each answer controls takes skilled people a long time and there are never as many of them as a firm needs. 

This month Avvoka opened early access to headless templating, which lets AI assistants build legal templates as well as draft from them. On paper it's a narrow change. For partners, it moves the limit on how much of the firm's approved work can be reused. 

Why do most law firms automate so few of their documents? 

Template building is slow, specialist work. Each one needs someone to set up the questions, the conditional logic and the clauses each answer controls. In most firms that falls to a small knowledge or legal engineering team, so automation grows only as fast as that team can build. 

The most-used agreements get automated first. The long tail waits: the variant lease, the jurisdiction-specific NDA, the side letter a practice group uses thirty times a year. Associates keep drafting those from the last deal, with the copy-and-paste risk that comes with it. 

From a partner's seat, this shows up as a question about return. The firm paid for automation and only a fraction of its documents use it. 

Why doesn't drafting straight from an AI assistant fix it? 

A general AI assistant drafts from everything it has ingested, so the same request can produce a different document each time. That's fast, but it isn't the firm's approved wording. Connected to a firm's templates, the assistant drafts from the firm's best work instead and the same request produces the same document. 

That's why Avvoka connected its templates to the assistants lawyers already use: Claude and Microsoft 365 Copilot through integrations, Harvey and Legora through partnerships. Lawyers stay in the tool they draft in. The template supplies the firm's judgement and house style. 

But that only helps for documents that have a template. Connecting the assistant made templates more useful. It didn't make them any quicker to build. 

Can AI build a legal template? 

Yes, with a lawyer reviewing the result. With headless templating, a lawyer or knowledge manager gives the assistant a Word document or an existing template and describes how it should work. The assistant sets up the questions and logic, tests the template against sample answers and saves it as a draft for the team to review. 

The set-up covers the parts that usually take the time. The assistant places each question in the text and ties it to the clauses it controls, including conditions, repeating sections and calculated values. 

In practice, the requests look like this. A team asks for its Retail Lease to be built to match its Commercial Lease, with the same question names and conditional structure. Or it hands over a standard mutual NDA as a Word file and asks for the non-solicit clause to appear only on US deals. Each time, the assistant lists what it built so the team can check it. 

We call this headless because no Avvoka screen is involved. The work happens inside the assistant. It's available first in Claude, with Copilot, Harvey and Legora to follow. 

Can you watch the template being built? 

Yes. A knowledge manager can keep their AI assistant open in one browser tab and the Avvoka template in another and see each change appear as the assistant makes it: questions added to the questionnaire, square brackets turned into placeholders and conditional wording, tables turned into repeating sections. If something looks wrong, they can redirect the assistant in the chat. 

When we tested this on a share purchase agreement, the assistant built about 70 questions, ran two sample deals and saved a draft in around 20 minutes. By hand, that template would take someone senior at least a week. 

Who stays in control of the template? 

The firm's lawyers do. The assistant's work is saved as a draft and the version lawyers draft from doesn't change until someone at the firm reviews and publishes it. The assistant does the build. A lawyer decides what the template says and when it goes live. 

For partners, this matters more than the speed. A template carries the firm's position on a clause into every document drafted from it and nobody wants that position set without sign-off. 

Eliot Benzecrit, Avvoka's co-founder, said as much when the feature was announced: “The lawyer still decides what the template says and when it goes live.” Headless templating also uses the same integration and access model as the existing connection and an administrator approves it before anyone at the firm can use it.  

What changes for a firm's template library? 

More of it gets automated, because more people can build. Template creation no longer depends only on a specialist team's capacity and a template built to the firm's standard can become the pattern for the next one. Every assistant drafting from the firm's library then has more approved work to draw on. 

The last row is the one that stays the same. Avvoka is used by almost a quarter of the Am Law 100 and in those firms a person decides what goes live. Headless templating speeds up the work before that point. 

Firms have always had more good precedent than time to automate it. What that work was worth depended on the size of the team building templates. Removing that limit doesn't change what makes a template good. It changes how many of them a firm can have. 

See it in action 

Headless templating is in early access for Avvoka customers already using the integration with Claude and Copilot. Legora and Harvey integrations are also underway for customers of both, who are using Avvoka.  

Find out more about Avvoka's integrations at avvoka.com/headless-drafting or get a walkthrough from our team at avvoka.com/demo  

About the author: Katya Hansom is Head of Product at Avvoka, where she leads the development of the core Avvoka experience, Curate, and integrations. Before Avvoka, she was Head of Product for CreateiQ, Linklaters’ CLM platform and has scaled compliance tech, fintech and capital markets products for over a decade. She also coaches women in product management.