Keeping Control as AI Does More: Creating a Firm Where Lawyers Stay in the Loop and AI Can Be Trusted to Do More

Most law firms have the same instinct about AI: a lawyer should review everything it produces before it reaches a client. That instinct is sound. Professional responsibility, client confidentiality and the reputational cost of getting something wrong all point the same way. Keep a person firmly in control. For most tasks, that remains the right approach, but we are not talking here just about legal work.

As AI becomes more capable, though, applying the same level of review to every task, from high-stakes client advice to routine document filing, stops being practical. Firms need a way to decide how much review each piece of work needs, and to write that decision down.

📅 Talk to us

AI agents, systems that plan, research, draft, check their own work and take action with little direction, are now practical and can handle multi-step tasks that used to need constant supervision. Moving towards agents, your roadmap needs to cover levels of review as your use of AI develops.

Developing your AI Adoption Roadmap

The roadmap is where a firm sets out how it will use AI, in the legal work and in running the firm. It is a good deal more than an AI policy. A policy tells people what they may and may not do. The roadmap adds why the firm wants AI, who decides, how use cases are chosen and approved, how people are trained and how the technology is kept fit for the purpose. It is a living document, reviewed as the technology and the firm's experience move on, with a named person responsible for keeping it current.

The main sections likely cover:

• Our approach to AI: objectives, guiding principles, the four levels of review and the adoption journey from Assist to Automate
• The traffic light for tools and uses, and the approved tools list
• Use stages, agentic AI and the use case development pathway
• Playbooks and skills
• Client engagement and disclosure
• Professional responsibility, supervision and record-keeping
• Data protection and security
• AI in HR and people management
• Incidents and reporting
• Training by role
• Governance structure, including how lawyers and IT work together
• Technical implementation and applications with embedded AI
• Metrics and reporting
• Vendor requirements, risk management, exceptions and policy review

Appendices can add more detail with some core registers, documents and templates.

Each firm develops and agrees its own approach in these areas to establish a platform to work from that should be constantly kept under review and developed further as people learn more.

Four levels of review

We see human oversight at four levels, matched to the work. Most people are currently at Level 1 and will be for some time - and that's ok, provided you have that roadmap with a sound starting point and shared plans to develop from there.

• Level 1: Human-led, AI-assisted.  The lawyer does the work. AI drafts, summarises, researches or suggests from what it's been given. The lawyer reviews everything it produces, corrects and takes responsibility for it. This covers most of what people in a firm do with AI today: asking a question, testing an idea, improving a draft, a first version of a letter, a summary of a file, in legal work and in running the firm. Most of it has nothing to do with legal authorities. General tools are unreliable for case law and legislation unless they draw on a dedicated legal research source or the firm's own approved materials, and a firm gets a long way before it goes near that. The rules on confidentiality and approved tools need to be defined and respected from the outset. This can also where standards start to be written: what lawyers keeps correcting should be in the playbook later. For high-judgement, high-risk work it's the right model for the foreseeable.

• Level 2: AI-led, human-supervised.  Lawyers have written down what good looks like for a piece of work: the steps, where the decisions are made, what must be checked, what gets escalated. That's content for the playbook. AI applies it to a defined task or set of documents and the lawyer supervises: reviewing at agreed checkpoints, checking the flags and the cited facts, sampling for quality. The lawyer's role shifts from doing the work to checking it against a standard.

• Level 3: AI-managed, human-governed. AI runs a (perhaps short, for starters) defined workflow or step end-to-end on a path the firm has fixed: trigger, inputs, checks, output. It completes the step and refers anything it can't handle to the legal team. The lawyer governs it: reviews the exceptions, audits samples and signs anything that reaches a client.

• Level 4: Agentic AI, autonomous within boundaries.  An agent plans and runs multi-step work within guardrails the firm has set, choosing its own steps: researching a question, drafting, cross-checking sources, presenting findings for review. The difference from Level 3 is that the firm approves the guardrails and the monitoring, since it hasn't fixed the path. Nothing that acts on a system or a client reaches Level 3 or 4 until the firm's governance group has approved it, use by use.

For everyone in the firm the roadmap puts these into three stages: Assist (Level 1), Assist+ (Level 2) and Automate (Levels 3 and 4). Automate is recognised as Red, not to be used until approved.

The same four levels apply across the firm: client onboarding, complaints handling, billing queries, HR processes, tender and questionnaire responses as much as drafting and advice. Some of the earliest candidates for Level 3 are administrative steps with a fixed timetable and a defined output.

Traffic lights for AI tools and uses

Alongside the levels of review, the roadmap can classify the tools themselves and the uses they're put to. Green means approved: the firm has looked at the tool, the data protection terms and the configuration, and it can be used with client data under the agreed level of review. Amber is in progress, under evaluation or being set up, so no client data until it moves to Green. Grey is on the list but not yet even evaluated. Red is prohibited, which in most firms means the free consumer tools with no data protection and anything the firm has decided against. Anything that isn't on the approved list is Red until the governance group has considered it.

Where law firms should be now

Most firms are at Level 1. Some are moving into Level 2. That's entirely appropriate.

The firms that will be best placed as AI matures are the ones building the structure now: the roadmap agreed, the playbooks written by the people who do the work, a governance group that decides what moves up a level. With that in place a firm can move deliberately as confidence grows. Without it, every move has to be debated ... probably time and again.

The pace of change in AI is accelerating. Firms that build this foundation now will be ready to take advantage of what comes next while keeping the professional standards and governance that clients expect.

From roadmap to everyday practice

The roadmap gives the firm the environment to work in: people who understand what AI can and can't do for their work, the technology set up so it can be used safely, and the agreed way of deciding what gets used, by whom and with how much review. On that footing lawyers and managers bring their own knowledge and judgement to the work, with AI-enabled tools doing the parts they're good at, and the combination develops as the firm's experience grows.

The purpose is to give lawyers more of their time back for the parts of the job that need a human: thinking through the client's problem, advising, negotiating, building the relationship. That delivers more value for the client and makes the work more rewarding, financially and intellectually. The knowledge the firm has built up is written into playbooks, one for each piece of work, that a trainee can follow, a partner can supervise against and an AI application can safely work to.

This is the sequence of our approach: establish the foundations, identify and prioritise the use cases, develop and test them, then extend across the practice. It's set out at AI Adoption for Law Firms.

AI does not create competitive advantage on its own. Competitive advantage comes from the playbooks, judgement and legal knowledge your firm teaches AI to apply.

Practical questions for managing partners

Does your firm have a way to decide which level of review each piece of work needs? Is that decision written down, governed and understood across the firm?

If the answer is "everyone uses their own judgement", you have the problem most firms have: inconsistency, unmanaged risk and no defensible position when a client, the regulator or your insurer asks how you manage AI.

Where to start?

The AI adoption review gives the partners a clear picture of current AI use across the firm, the governance and policy gaps, the priority opportunities and practical next steps. It stands on its own if the firm goes no further.

If you'd prefer to talk it through, I'm happy to have an initial conversation.

Contact: Allan Carton

Email: acarton@cartonconsultants.com

Mobile: 07779 653105

📅  Book a call here >>

Related reading

• Practice and Case Management System Selection: Updating and replacing practice management and other systems.
• AI Adoption for Law Firms: our approach, from foundations to extending across the practice.
• Microsoft 365 and Copilot for Law Firms: the right licences, SharePoint and Teams set up around matters and Copilot ready for real work.
• Upgrading Technology in Law Firms: A Practical Guide: more detail on selection of systems.
• Legal Technology - Working Smarter: getting more from the systems the firm already has, AI included.

Allan Carton is the founder and Lead Consultant at Carton & Co, specialist consultants to law firms on technology adoption, client relationships and business development. A qualified solicitor with an MBA from Alliance Manchester Business School, Allan has advised law firms on business development and client relationships since 1990 and on technology selection and adoption since 1992. He works directly with managing partners and senior leaders as an independent sounding board - helping firms make confident decisions on AI, legal technology and client strategy.

Connect with Allan on LinkedIn | Carton & Co