AI for law and accountancy firms: the complete guide (2026)
Most professional services firms don't have a talent problem or a pricing problem. They have a leakage problem. Time goes unrecorded, work-in-progress sits unbilled, and the same document gets rebuilt from scratch for the hundredth time. AI won't give the advice, but it can now run a lot of the work around it.
A growing law or accountancy firm sells expertise, but a surprising share of its people's week goes on something else: opening matters, checking identity, chasing documents, filling in time sheets, building reports for partners and turning a draft from last year into a draft for this year. In firms that have grown by acquisition, every office often does those things its own way.
1. Why professional services, why now
Firms have always known where the time goes. What has changed is that the work in between, reading, summarising, assembling and checking, is now something software can do reliably.
Two things have changed in the last 18 months:
- The models are good enough. Reading an engagement letter, a set of accounts, a contract, a client's ID documents or a matter file, and acting on it, is now reliable work for AI, not a demo.
- The bottleneck is context, not capability. What's missing is how your firm works: your precedents, your risk rules, your billing arrangements, how each partner likes things done. That knowledge lives in a few senior heads and a lot of folders. Getting it into a form software can use is the real job.
The firms that solve this will grow fee income faster than headcount. The ones that don't will keep adding support staff for every new team.
2. Where the time and money actually leak
The same six leaks show up in almost every firm we look at.
| Leak | What happens | What it costs |
|---|---|---|
| Time never recorded | Calls, emails and short pieces of work are done but never make it onto a time sheet | Fee income lost before it reaches WIP, and a distorted view of which work is profitable |
| Lock-up | WIP sits unbilled and bills sit unpaid while fee earners get round to it | Cash tied up in work already done, often funded by partners or an overdraft |
| Onboarding by email | Conflict checks, client due diligence, ID documents and engagement letters are chased by hand | Slow starts, frustrated clients and compliance risk when a step is missed |
| Documents rebuilt from scratch | Every letter, report or first draft starts from a blank page or someone's old file | Senior time spent on assembly instead of judgement |
| Management information built by hand | Partner and group reports are pulled together from the practice management system, the accounts and spreadsheets each month | Problems seen weeks late, and finance teams that grow with every office added |
| Client updates and chasers | Clients are chased for documents and updated on progress one email at a time | Fee earner time absorbed by admin that the client never sees as value |
None of these is a professional problem in the technical sense. Each is an information problem: the facts exist somewhere in the firm, but not in the right place at the right time.
3. What AI can and cannot do
Be precise about the boundary. Firms are regulated, and overselling AI to them is the fastest way to lose them.
AI can read. Matter files, contracts, accounts, bank statements, ID documents, client emails and correspondence can be read and understood without re-keying.
AI can assemble. A first draft from your precedents and the facts of the matter. A time entry from the emails and calls that happened. A partner report from the practice management system and the accounts.
AI can check. Is this client's due diligence complete? Has this matter gone over its estimate? Is this WIP ready to bill? Does this draft match the firm's standard terms?
AI can chase. Clients for documents, fee earners for time and bills, debtors for payment, sent on schedule, logged and escalated when ignored.
AI cannot give the advice. It should not decide what a client should do, sign off a set of accounts or accept a new client. Those stay with qualified people, who remain responsible for the work. The point is that they spend their time on those calls, not on assembling what feeds them.
The rule we use: AI prepares, checks and chases. A named, qualified person approves anything that goes to a client, commits the firm, or carries professional risk.
4. The matter lifecycle, stage by stage
Map the life of a matter or engagement and the opportunities become obvious. Each stage has admin that repeats every time.
| Stage | Admin that repeats every matter | What AI does | Who still decides |
|---|---|---|---|
| 1. Win | Reading the enquiry, scoping, estimating fees, writing the proposal | Drafts the proposal and estimate from similar past matters | Partner agrees scope and price |
| 2. Onboard | Conflict checks, client due diligence, engagement letter, matter opening | Collects and checks documents against your rules; chases what's missing; drafts the engagement letter | Risk or compliance signs off the client |
| 3. Deliver | Drafting, research notes, client updates, document requests | Produces first drafts from your precedents; keeps the client updated and chased | Fee earner reviews and owns the advice |
| 4. Record and bill | Time recording, WIP review, bill preparation | Suggests time entries from activity; flags WIP ready to bill; drafts the bill and narrative | Partner approves the bill |
| 5. Collect and report | Credit control, matter closure, partner and group reporting | Chases debtors; builds the monthly reports from live data | Managing partner and finance act on what they show |
Underneath every stage sits the same thing: a live view of each matter's profitability and each team's lock-up, updated as the work happens rather than at month end.
5. Why point tools haven't fixed it
Most firms already pay for a practice management system, a document system and a growing number of AI subscriptions. The problem is that each one owns a slice, and the joining-up still happens by hand.
| Approach | What it's good at | Where it falls short | Best for |
|---|---|---|---|
| Spreadsheets and email | Flexible, familiar | Knowledge lives in individual files and inboxes; nothing joins up | Small firms with one office |
| Practice management and time recording | Matters, time, billing, the system of record | Only as good as what people enter; weak at reading documents or chasing anyone | Every firm past a handful of fee earners |
| Document management | Storing and finding files | Holds the knowledge but doesn't act on it | Firms with large document volumes |
| Point AI tools (drafting, research, chat) | One task done well | Another login, another silo; doesn't know your firm's rules or clients | Individual fee earners |
| An AI operating layer (the Saturn approach) | Sits across the tools you already have; learns how your firm works and runs the workflows between them | Needs a structured install, not a self-serve trial | Growing firms and groups where support costs are scaling with fee income |
The last row is not a replacement for your practice management system. It's the layer that makes your systems work together and does the work that currently falls between them.
6. What an operating layer looks like
Saturn works in four steps. None of them asks your fee earners to learn a new system.
- Connect. We plug into what you already run: practice management, document management, accounts, email, shared drives. Nothing is migrated or replaced.
- Structure. We build a living map of how your firm actually works: how a matter is opened, what complete due diligence looks like, which precedents apply, how each partner wants reporting. Each of those becomes an executable skill.
- Act. AI agents run those skills through a permissions-aware gateway. They draft, check and chase, and hand anything that needs judgement to the right person, inside the tools that person already uses.
- Compound. Every correction a fee earner or supervisor makes is captured. The map gets sharper, so the hundredth engagement letter needs less checking than the first.
For a group of firms, that map is also how different offices start working the same way without anyone being forced onto a single system on day one.
7. The economics
The value case is about fee income and cash first, and headcount second. These are the levers we look at with any firm, expressed as proportions so you can apply them to your own numbers.
| Lever | Assumption | Impact |
|---|---|---|
| Time captured | Work that currently goes unrecorded is suggested as time entries for fee earners to confirm | A small rise in recorded hours per fee earner adds up quickly across a firm |
| Lower lock-up | WIP is flagged and billed promptly, and debtors are chased on schedule | Cash released from work already done, with less reliance on borrowing |
| Faster onboarding | Due diligence and engagement letters are assembled and chased automatically | Matters start sooner and fewer compliance steps are missed |
| Drafting time | First drafts are produced from your own precedents | Senior time moved from assembly to review and advice |
| Management information | Partner and group reports produced from live data | Problems seen while they can still be fixed, without growing the finance team |
Your numbers will differ, and the first two levers usually matter most. The point is the scale: a properly scoped install should pay for itself on one lever, not five.
8. Governance, confidentiality and security
Firms hold privileged, confidential and personal information, and they answer to their regulators for how they handle it. An AI layer has to make that picture safer, not wider.
- Permissions by role and matter. An agent working for a fee earner sees what that fee earner can see, and nothing more. Ethical walls and matter restrictions are respected.
- Every action logged. Each document read, draft produced and message sent is recorded against a matter and a person. If a client, insurer or regulator asks, the trail is there.
- Qualified people approve. Anything that goes to a client or carries professional risk is approved by a named person before it leaves.
- Spend controls. AI usage is metered and capped, so costs are predictable.
- Your data stays yours. We connect to your systems where they are. Client data is not used to train shared models.
9. A 12-week rollout
We start with one workflow that moves fee income or cash, prove it, then widen.
- Weeks 1–2: Map. We sit with a practice group and the finance team, connect your systems and pick the first workflow. Usually that's WIP and billing, or client onboarding. We agree what success looks like in numbers.
- Weeks 3–6: First workflow live. The agents run alongside the team on live matters. Every output is reviewed, and every correction is captured.
- Weeks 7–8: Measure. We compare against the baseline: lock-up days, time recorded, onboarding time, hours saved. Gate: carry on only if the numbers have moved.
- Weeks 9–12: Widen. Add the next workflows, or the next office. Review steps are reduced where accuracy has earned it.
After week 12, the system runs on a maintenance subscription, and each new workflow is quicker to add because the map of your firm already exists.
10. FAQ
Do we have to replace our practice management system? No. Saturn sits across what you already run.
Our fee earners won't change how they work. Is that a problem? They don't need to. The agents work in email and the systems they already use.
What about professional responsibility? It stays exactly where it is. A qualified person reviews and approves anything that goes to a client.
Is this just for large firms? The sweet spot is growing firms and groups where support costs are scaling with fee income. Very small practices usually don't have enough repeat work to justify an install.
We've grown by acquisition and every office does things differently. Does that matter? It's one of the best places to start. The map of how each office works is how they begin to work the same way.
How long until we see results? The first workflow is live on real matters within about six weeks, with measured results by week eight.
11. How we work with firms
Saturn was built inside operating businesses, not sold from the outside in. We buy and run companies ourselves, and the same system runs our own portfolio and our clients today, including a company brain over hundreds of documents for a regulated fintech. The work in a firm is different, but the problem is the same one we see everywhere: the way the business really runs is stuck in people's heads and disconnected systems.
If you run or manage a growing firm and support costs are rising with every team you add, we'd like to hear how you're handling it. The first conversation is 30 minutes. We'll show you Saturn running on a live business, and together we'll pick the one workflow where it would move fee income or cash for you.
← All posts