AI for Professional
Services Firms
Many firms still rely on nothing more than an unmanaged browser chatbot open on a few partners' laptops. But we've been embedding workflows powered by artificial intelligence for clients across a number of professions. Our approach starts with understanding which tasks and workflows can be tackled with AI, and which shouldn't be. It also involves aggregating key knowledge into a private, regulator-ready layer of your firm's 'brain', with appropriate permissions in place. And, critically, it's done with human judgement, responsibility, and sign-off firmly retained by you.
Search the firm knowledge
Keep client permissions
Show the source
Staff find useful past work without opening every file.
Why we think this matters
Professional services is rife with unbilled work, which makes it ripe for efficiency gains with AI. But in regulated professions, mistakes carry reputational and legal risks. So it's no surprise that specialist platforms are being developed, including Harvey, CoCounsel, and many more. But many of these solutions are enterprise-level, per-seat platforms with minimum commitments. These can be worthwhile for some, but others can achieve greater benefits from a more tailored approach to AI, one that starts with the work that needs doing, rather than a prescribed approach.
Regulators and professional bodies are paying close attention too, recognising both the efficiency gains AI can bring and the risks of getting it wrong. From the Ayinde judgment, which confirmed that a lawyer is responsible for understanding and checking all cases and propositions relied on, to RICS' new standard on the responsible use of AI, which stipulates that AI should not replace judgement for material surveying tasks, and the FRC's recent guidance, we're entering an era of increased accountability for professional firms using AI.
How we build AI into a firm's workflows
Our work involves embedding AI into your firm's existing ways of working, rather than prescribing new processes. The specifics vary by profession and workflow, but examples of workflows that lend themselves well to AI include:
- Intake and conflicts. This can include capturing enquiry details, collating KYC information into an audit-ready format, and drafting engagement letters that can be checked and signed off by an authorised signatory.
- Drafting. This includes drafting on the basis of your precedents or instructions. It involves the drafter checking the document for accuracy of facts, legal position, etc., as appropriate.
- Document review. Review and/or summarisation of documents can be undertaken by AI, with summaries referencing pages for ease of human review. It's critical to understand both false flags and false negatives.
- Time recording and WIP. This involves agreeing with the team whether AI-generated descriptions are okay, and on what basis, including privacy and monitoring, before agreeing a format and piloting AI-drafted time.
- Building your firm's brain. Aggregation of precedent, firm guidance, manuals, etc., into a single, searchable format that's accessible to the right people. This is a powerful application of AI in professional services, as it surfaces knowledge that can currently only be accessed via a few partners. Access controls are critical, as some precedents will be restricted to certain matters/clients, or require permission before use.
How to get started: the AI Audit
The best way to understand where you should be using AI is via an AI Audit. We help firms prioritise which workflows to introduce AI to by mapping out where it could help, and prioritising that list by value, readiness, and downside. We'll show you where AI could be embedded into workflows, prioritise which workflows to focus on, and flag areas of regulatory or other risk to consider.
From there, we'll:
Agree what data can be used, including advice, precedent, firm manuals, and build a private repository of this, with appropriate permissions
Start introducing AI into individual workflows with citations and room for professional judgement
Agree governance controls before launching each workflow: risk register, supervision, model card, client notice if required
Agree contracts, including: Ownership of your knowledge, our IP, and any new IP; Prompts and audit trails; Data privacy, retention, hosting, subprocessors, data leaks, both incoming and outgoing; Handover plan including access, hosting, maintenance.
This helps minimise hallucinations because AI can only refer to precedent, legislation, contracts, etc., it has been shown, with links to the source, protects client confidentiality by reviewing AI provider contracts alongside your own client contracts to understand data protection, residency, subprocessors, etc. and future-proofs your investment by helping you understand the strengths and weaknesses of the models you are using and whether they need updating over time. We'll also discuss with you how AI could affect training and development plans for juniors, to ensure they're still able to learn the ropes.
Get in touch
Get in touch if you'd like to talk about introducing AI into your firm. We'll point you in the direction of more detailed guidance for law firms, accountancy practices, recruitment businesses, or consultancies, or you can speak to us about other professions.