The Billable Hours Problem
A solicitor earns £250 an hour. She also spends two hours a day on tasks that don't require legal training: answering "what documents do I need to bring?", chasing clients for outstanding paperwork, confirming appointment times, fielding questions about the firm's practice areas — precisely the kind of work an AI agent for law firms exists to absorb.
That's £500 a day in fee-earner time spent on admin. Across a five-person firm, it's £2,500 a day. Every day. Most managing partners we talk to know this number is bad and have largely given up trying to fix it, because the alternative — hiring more support staff — solves it on a different and similarly expensive line.
At a 12-person firm handling personal injury and employment claims, a senior associate tracked her actual time for a month. She was spending 11 hours a week on client communication that required zero legal judgment: confirming court dates, chasing medical records, updating clients on progress. That's more than a quarter of her working hours. At her charge-out rate, it was £13,750 a month in lost billing capacity — for tasks that a well-built agent would handle in seconds.
AI agents don't write contracts. They don't argue cases. They don't replace legal expertise. They handle the admin layer around legal work — instantly, accurately, and at any hour — so your fee earners can bill for the time they're actually paid to spend.
What an AI Agent Does for a Law Firm
Client Intake and Qualification
A potential client contacts the firm. They want to know if you handle their type of case, whether you're taking new clients, and roughly what your fees look like.
An agent answers immediately, collects basic details and the situation, asks the qualifying questions your intake process requires, and routes qualified leads to the right fee earner with a summary already prepared.
For high-volume practice areas — personal injury, family law, employment, conveyancing — this transforms how your firm manages enquiries. Every potential client hears back inside seconds, regardless of when they reach out, and the fee earner who eventually picks up the conversation isn't starting cold.
Consider a mid-sized family law firm that takes 60 new enquiries a month. Without an agent, the receptionist fields each call, takes notes, and routes the message to a solicitor who then does a second qualification call before deciding whether to take the matter. With an agent handling initial intake, that first call is replaced by a structured conversation that captures the client's situation, assets involved, whether children are part of the matter, and urgency level. The solicitor receives a one-page summary and makes a take/decline decision in under three minutes, not twenty.
Appointment Scheduling
A client needs to book an initial consultation. The agent checks availability, offers slots, confirms the booking, sends a calendar invite with any preparation instructions, and follows up with a reminder the day before.
Rescheduling requests run the same way — the client messages, the agent offers alternatives, the calendar updates. No phone tag. No admin overhead.
For firms with a high no-show rate, automated reminders alone can recover meaningful revenue. A 10-appointment week with a 15% no-show rate is 1.5 billable hours lost. Automated day-before and two-hour-before reminders typically reduce no-shows by 40–60% in legal settings. At £250 per hour, a five-person firm can recover £150–£250 per week from this change alone — more than £10,000 a year from a single automated touchpoint.
Document Chasing
Waiting for clients to send documents — ID, financial statements, contracts, correspondence — is one of the most time-consuming parts of legal work. Chasing typically falls to a fee earner or paralegal with better things to do.
An agent handles the chase automatically: polite reminder when a document is due, follow-ups at set intervals if it hasn't arrived, notification to the fee earner only when everything is in or when a client has gone silent for too long.
In conveyancing, where transactions regularly stall because a client hasn't sent proof of funds or ID verification, document chasing is a constant drain on paralegal time. A 30-matter pipeline means someone is manually tracking outstanding documents across all active files — and forgetting some of them. An agent tracks every outstanding item, sends reminders at 48-hour intervals, escalates to the fee earner when a client goes seven days without responding, and logs every contact attempt automatically.
FAQ and Practice Area Queries
"Do you handle employment disputes?" "What's the process for a residential conveyance?" "How long does a divorce typically take?" "What's your hourly rate?"
Same questions, same answers, every time, by whoever picks up the phone. An agent answers them instantly, around the clock, freeing reception and fee earners from intake calls that never quite turn into billable work.
The out-of-hours dimension matters more than most firms expect. A person going through a difficult situation — a redundancy, a relationship breakdown, a property dispute — often decides they need a solicitor at 10pm on a Tuesday. If your website has a live agent, they start the process immediately. If it doesn't, they may have moved on by the time you open.
Status Updates for Active Matters
Clients want to know where their matter stands. "Has the other side responded?" "Has the search come back?" "Is there anything I need to do?"
A well-integrated agent provides status updates from your case management system in plain language — without a fee earner needing to make a call or write an email.
This is particularly impactful in conveyancing and family law, where matters run for months and clients check in frequently. Pulling status from your case management system and translating it into plain language removes a class of interruption that fee earners find genuinely disruptive — the "just checking in" call that arrives mid-document.
What AI Agents Should Not Do in Legal Practice
This matters enormously in a regulated profession.
Agents in legal settings should not give legal advice. They answer procedural questions, provide factual information about the firm's services, and handle administrative tasks. Any question that requires legal judgment — "do I have a case?", "should I accept this offer?", "is this contract enforceable?" — must go to a qualified lawyer.
A well-built legal agent has hard boundaries here. When a query crosses into legal advice territory, it says so explicitly and routes to a qualified person. The firm's professional obligations are protected.
This is a design requirement, not an optional feature. We'd rather lose a build than ship one that can't hold the line.
Compliance Considerations
Law firms handle sensitive client data under strict professional obligations — solicitor-client privilege, SRA regulations, GDPR.
Any agent deployed in a legal context must:
- Store client communications securely with appropriate access controls
- Not share client information across matters or with unauthorised parties
- Maintain audit trails of all interactions
- Allow clients to opt out of AI-handled communications at any point
- Be disclosed to clients as an AI system where required
These requirements shape the architecture from day one. A firm that bolts on compliance at the end of a build will almost always have problems.
For US-based firms, the requirements shift slightly — attorney-client privilege obligations, state bar rules on unauthorized practice of law, and applicable state privacy laws (CCPA in California, for instance) all need to be mapped before a line of code is written. The underlying principle is the same: compliance is architecture, not an afterthought.
Where This Doesn't Fit
For very small firms — one or two solicitors, low enquiry volume, high-touch boutique work — an agent is over-engineering. The cost of building and maintaining doesn't pay back at that scale, and the personal touch of the partner taking the call is part of the product.
The fit is strongest for high-volume practice areas where the intake patterns repeat and the routine queries are a real drag on fee-earner capacity.
Off-the-Shelf vs. Custom-Built Legal Agent
| Off-the-Shelf (e.g. chatbot builders) | Custom-Built Agent | |
|---|---|---|
| Setup time | Days to weeks | 6–8 weeks |
| Cost | £50–£500/month subscription | £8,000–£14,000 one-time build |
| Integration with case management | Rarely possible | Core part of build |
| Compliance controls | Generic, not legal-specific | Designed for SRA/GDPR requirements |
| Escalation logic | Basic keyword routing | Judgment-based routing with legal boundaries |
| Document chasing | Not available | Full automation with audit trail |
| Customisation | Limited templates | Built to your intake process |
| Ongoing cost | Recurring monthly fee | Hosting + maintenance only |
Off-the-shelf tools work for very simple FAQ responses. They break down as soon as you need integration with your case management system, compliant data handling, or escalation logic that understands the difference between a procedural question and a request for legal advice.
What to Expect in Practice
The first four weeks after deployment are a calibration period. The agent will surface edge cases your intake process hadn't formally documented — clients who don't fit neatly into your practice areas, questions that sit on the boundary between procedural and advisory, situations where your escalation rules need refining. This is normal and expected. A good build includes supervised monitoring during this period so the team can correct the agent's behaviour in real time.
By week six, most firms have settled into a rhythm. The agent is handling 60–75% of inbound contacts without human involvement. Fee earners are touching new enquiries only to make the take/decline decision. Document chasing is invisible — it happens in the background, and the paralegal only hears about it when there's a problem.
What changes behaviourally: fee earners stop treating their inbox as a to-do list for client communication. Active matter enquiries are handled before they even see them. This is a bigger shift than most expect — the cognitive overhead of managing client communication is substantial, and removing it frees attention that shows up in other ways.
Common Mistakes When Deploying AI Agents in Legal Settings
The most common mistake is under-specifying the escalation logic. Firms often say "route to a fee earner if the question is complex" without defining what complex means. The agent then either over-escalates (routing things it could handle) or under-escalates (handling things it shouldn't). Define your escalation rules in writing before the build starts. If you can't articulate when a human must take over, the agent won't be able to apply that judgment consistently.
The second common mistake is treating the agent as a finished product at launch. Legal intake patterns change — new practice areas, new compliance requirements, client communication preferences that shift over time. Firms that don't budget for ongoing maintenance find their agent gradually drifts from their actual process, and the problems surface slowly rather than all at once.
The third mistake is inadequate client disclosure. Clients must know they're talking to an AI system. Firms that obscure this — either to appear larger than they are or to avoid the friction of the disclosure — create professional conduct risk. The disclosure is a one-line statement at the start of the conversation; it doesn't meaningfully affect conversion rates and it protects the firm.
The Economics for a Legal Practice
A five-person firm handling 40 new enquiries per month and 80 active matters.
Current state: each new enquiry takes 20–30 minutes of intake time (reception call, qualification, fee earner review). Active matter queries take 10–15 minutes each to field and respond to.
With an agent: new enquiry intake drops to about 5 minutes of fee earner time (reviewing the agent's summary and making a decision). Routine active matter queries are handled automatically.
Time saved per month: roughly 25–35 hours of fee earner and reception time.
At £150–250/hour: £3,750–£8,750 in recovered productive time per month.
Agent build cost: £8,000–£14,000 for a well-integrated legal intake and matter management agent. Payback in 1–3 months for most firms with meaningful enquiry volume.
What a Deployment Looks Like
Legal projects need more careful scoping than most because of the regulated environment, but the technical complexity is manageable.
- Week 1–2: Map your intake process, define qualification criteria, document your FAQ content, agree escalation rules and compliance requirements
- Week 3–4: Build and integrate with your case management system and communication channels
- Week 5: Internal testing with your team — edge case review, compliance check
- Week 6–7: Supervised live operation with fee earner oversight
- Week 8: Full deployment with monitoring period
Eight weeks to a compliant, production-ready legal agent.
Related guides
- AI agents for accounting firms
- AI agents for professional services firms
- AI agents for financial services
- AI agents for appointment booking
- Our AI agent development services
Ready to Give Your Fee Earners More Billable Time?
Every hour your lawyers spend on admin is an hour they're not billing. And every client enquiry that waits until morning is a potential instruction you may be quietly losing.
Talk to us about your business — we build legal agents with professional compliance baked in from the start, and if your firm isn't the right shape for this we'll say so on the call.
Frequently Asked Questions
Can an AI agent give legal advice to clients?
No, and a properly built one won't try. The agent handles procedural questions, practice area information, intake, and admin tasks. Any question requiring legal judgment — whether a client has a claim, what a contract clause means, whether to accept a settlement — is immediately escalated to a qualified lawyer. This boundary is hard-coded into the design, not left to the agent's discretion.
Is client data handled securely when using an AI agent?
It depends entirely on how the agent is built. An agent designed for legal use will store client communications in encrypted form, maintain access controls so staff only see data for their matters, log every interaction for audit purposes, and comply with GDPR and SRA data handling requirements. Off-the-shelf chatbot tools typically don't meet this standard. This is one of the main reasons law firms need a purpose-built solution rather than a generic product.
Do clients need to know they're talking to an AI?
Yes. Under SRA guidance and general consumer protection standards, clients should be informed when they're interacting with an automated system rather than a person. In practice this is a short, clear statement at the start of every conversation: "You're speaking with our automated intake assistant." Firms that skip this disclosure create professional conduct risk and, frankly, client trust problems when the reality becomes obvious.
How does the agent connect to our case management system?
Through an API integration built during the project. Most legal case management systems — Clio, Osprey, Leap, Practice Evolve, and others — have developer APIs that allow a third-party application to read and write data under controlled conditions. The agent queries your case management system for matter status, availability, and document status, and writes back intake records and interaction logs. This integration is scoped and agreed during week 1–2 of the build.
What practice areas benefit most from a legal AI agent?
High-volume areas with repeating intake patterns: personal injury, family law, residential conveyancing, employment law, debt recovery, and immigration. These practice areas share a characteristic — the initial enquiry and intake process follows a fairly predictable path, and the routine client communication is high-volume but low-judgment. Bespoke corporate work, complex litigation, and high-value transactional matters with low enquiry volumes typically don't justify the build at a smaller firm.
How much does it cost to build an AI agent for a law firm?
A well-integrated legal agent — covering intake, scheduling, document chasing, FAQ responses, and case management integration — typically costs £8,000–£14,000 to build. A simpler FAQ-and-intake-only agent with no case management integration runs £4,000–£7,000. Ongoing hosting and maintenance is typically £200–£400 per month depending on volume and complexity. Payback for most firms with 30+ enquiries per month is within three months of deployment.
What happens when the agent can't answer a question?
A correctly built agent doesn't guess or hallucinate an answer when it doesn't know. It tells the client it can't answer that question directly, explains what the firm can do for them, and either offers to take a message or provides contact details for the relevant fee earner. Every unanswered query is logged so the firm can review what gaps exist in the agent's knowledge and update it accordingly. This feedback loop is part of the ongoing maintenance process.