How ALC uses Claude Cowork to deliver better-value legal consultancy

We became interested in Cowork in early 2026. Headlines about sharp falls in legal and accounting software shares made us look more closely at what it could do. The attraction was practical: could it help us produce a useful first draft and organise a workflow, rather than simply answer questions?

We also needed to assess the subscription against our data-security requirements. When we bought the plan we selected, the minimum purchase was five seats. We needed two, so we ended up with three spare. Even with that cost, the value to our workflow has been substantial.

Yes. We started using AI assistance in late 2024, but the results were hit and miss. We made six attempts to build a platform that would automate a substantial amount of in-house counsel work. Each time, we ran into roadblocks around the accuracy we needed.

That experience helped us separate useful assistance from dependable legal work. A document can look polished and still miss something commercially important. Today, we use AI for parts of the process, with a lawyer responsible for checking the output.

We call it the ‘black swan clause’. That is our shorthand for an unexpected obligation that does not fit the services our client is actually providing. It is not a formal legal category.

Many of our clients sell services to larger enterprises. Those customers often insist on their own paperwork, which may have been written to cover many different purchases. A term can be sensible for one supplier and inappropriate for another.

For example, imagine a recruitment agency being asked to accept responsibility for maintaining the customer’s IT systems. That is an illustrative example, but it shows the problem: a broad obligation in the agreement may extend beyond the service being bought.

AI can do a lot of the comparison and initial analysis. We still read every term of every contract and check whether it is appropriate in context. We also look at how clauses interact. A tidy summary is not a substitute for that review.

Playbooks and initial contract drafting are particularly useful. We start with an appropriate boilerplate and interview the client about the business: what it provides, how delivery works, how it charges and where it needs flexibility.

We then ask AI to prepare a first draft using the template and that context. It is draft one. We refine it, check it and adjust it as the commercial decisions become clearer.

In our experience, that first pass can save hours of preparation on suitable matters. Smaller agreements, including routine NDAs, can take considerably less drafting time than before. The benefit for the client is less time spent assembling documents and more capacity for advice, negotiation and decisions.

The saving varies with the document and engagement. It does not remove the need for review or mean every matter has a lower fixed fee.

Multi-level lists in Word. We have had outputs that look beautiful but do not use Word’s native numbering properly. When you edit them, the document becomes frustrating to maintain.

That is our experience with particular outputs, rather than a claim that the tools can never create native numbering. The test is whether the document works when the client or counterparty edits it.

We expect less legal time to be needed for some recurring tasks. The amount will depend on the organisation and the work. Where a business has established playbooks and repeatable processes, AI has a clearer set of rules to work from.

Where every agreement needs bespoke terms, the opportunity to standardise is smaller. There can still be useful savings in preparation, but more of the work depends on commercial judgement and negotiation.

Our advice is to put the legal operations your business actually needs in place. Agree preferred terms, realistic fallbacks and escalation rules, while preserving the flexibility your customers require. A rigid playbook that does not fit your sales model can create its own problems.

They should expect us to understand their business and take responsibility for the work we deliver. ALC is AI-enabled: we use AI as a tool, and every piece of work product is checked and validated by a human.

For contract work, that includes reading every term and assessing it against the service, commercial priorities and agreed risk position. Our clients receive reviewed legal support, not an unchecked AI response.