Use the matter files
Draft with sources
A lawyer reviews it
Lawyers can check the source before approving the work.
What is happening right now
Fee earners in your firm are using AI to carry out research, drafting, review, etc. Some will be using AI for client work already. They’re using public accounts via their phones. You can’t stop it and shouldn’t try, but what’s missing is oversight.
The Ayinde case
The Ayinde judgment last year, which dealt with authorities that didn’t exist, highlights the need to get ahead of this, and implement controls. The SRA has also warned about the risks of generative AI tools, including around issues of hallucination and confidentiality. The firm will remain responsible for the accuracy of AI-generated content.
Steps for a successful AI project
1) Avoid using public accounts, implement a private stack with a contract with the AI provider that prevents them from training models with your data, along with conflict checking, logging of queries at a matter-level, and citations. The AI should link to primary sources so fee earners can easily validate the AI’s reasoning. The firm should own the data, prompts, outputs, logs, etc. This will provide visibility over fee earners’ usage of AI, you’ll be able to see settings and usage. The firm will be able to evidence its position should the SRA, or indeed a client, raise queries. The COLP wouldn’t want to be caught out.
2) Integrate with DMS and practice management, avoid having to visit a separate website. Embedding AI into existing systems will make adoption easier.
3) Pick the use-cases to automate, repetitive, dull, and/or expensive workstreams are most suited to automation, for example, review/drafting: clustering documents in disclosure, first-pass drafting of agreements/letters using precedent libraries; research: internal, such as finding all contracts that contain X clause, and external with references; intake: initial assessment of claims. Fee earners should always check citations, reasoning, relevance, application to client’s circumstances, missing information. AI is not going to take responsibility, it’s all still on fee earners.
4) Billing models and policies, who’s going to read the policy pack? We’ll help you work through these questions.
Public AI accounts vs private stack
Using a public account could expose the firm to risks including hallucinated case-law referenced in correspondence on its letterhead; unknown contract terms; queries to the AI not being logged; inadvertent release of confidential information. The solution is a private stack as outlined above. There are off-the-shelf AI tools for law firms, such as Harvey or CoCounsel, but Vu provides technical support to deploy a stack. Vu can help you with:
an audit to establish which tools are being used, how, whether they’re approved, who’s using them, whether there are controls around use; how the data moves through the AI, what happens to it once it’s been used; review the contract with AI provider(s); compare firm’s processes against Law Society and SRA guidance, are there gaps?
implementation, in phases. The first phase might be to focus on one workflow and agree expectations, what’s in-scope; fixed fees; when it’s working well; when we should stop.
Next steps
Have a chat with us. Tell us about your current AI tools; the types of matters; who owns the firm’s professional standards. We can advise on next steps and help you to get ahead of AI.