Anthropic’s reported $11.5 billion quarter is a marker worth paying attention to, even if your firm has no immediate plan to buy an enterprise AI platform.

The story behind that number is not just consumer curiosity about chatbots. Enterprise organisations are spending real money on Claude, particularly for coding and knowledge work that used to require experienced people moving through complex information. The original reporting frames it as AI becoming serious business infrastructure.

Large law firms are part of that market. They are testing and deploying tools for drafting, research, document review, matter management, and internal knowledge retrieval. Their clients will see the output. Some will soon ask why a routine contract review, first-pass discovery exercise, or legal research task still takes as long as it did five years ago.

For a law firm doing $1 million to $25 million in revenue, that does not mean copying the buying behaviour of a global firm. It means identifying the manual work that is costing you money now, then testing AI with clear controls and a narrow commercial purpose.

The sensible starting point is not “replace lawyers with AI.” It is removing the repeatable work that consumes associate hours, delays new matters, and rarely creates the client value your firm wants to be known for.

Why Claude’s enterprise growth matters to smaller firms

Claude coding tools are central to the enterprise story because coding is structured work. A developer has a defined task, a codebase, rules, tests, and someone qualified to check the result. AI can accelerate the first draft, explain unfamiliar code, identify issues, and help teams move faster without handing over final responsibility.

Legal work has more in common with that model than many partners first assume.

A commercial contract has defined clauses, fallback positions, governing law considerations, risk thresholds, and a client’s known preferences. Discovery has custodians, date ranges, privilege rules, issue tags, and review protocols. A client intake has practice areas, conflicts, urgency, budget signals, and routing rules.

None of this makes legal judgment automatic. It does make portions of legal operations suitable for controlled AI assistance.

The important signal from Anthropic’s growth is that enterprise buyers are no longer treating AI as an experiment run by one innovation person. They are funding it where it can handle high-volume knowledge work with governance around access, review, and accuracy.

Smaller practices have an advantage here. You do not need a 12-month procurement cycle to improve a lead response process or build a first-pass contract review workflow. You can select one workflow, define guardrails, measure the outcome over 30 days, and decide whether it deserves a wider rollout.

That is a more practical route than buying a broad legal AI subscription, asking everyone to “use it where useful,” and then wondering six months later what changed.

The work that is quietly leaking margin

Most law firm owners do not have a shortage of legal skill. They have a shortage of protected time.

The leak often begins before a matter is opened. A prospective client calls at 6:20 pm. The receptionist has left, the call goes to voicemail, and a competitor responds first. A web inquiry lands in a general inbox overnight. By the next morning, someone has to read it, work out the practice area, check whether it sounds commercially viable, and forward it to the right partner.

That is not glamorous work. It is still revenue-critical.

For firms of this size, we usually see a meaningful portion of after-hours inquiries fail to convert simply because nobody responds promptly. The 30% to 40% range is a useful warning sign for firms that rely on high-intent calls from people facing an urgent employment, family, property, litigation, or criminal matter.

Then there is the work after engagement.

An associate spends an afternoon locating clauses across a supplier agreement, comparing them against a client playbook, and preparing an email summary. Another spends two days on initial discovery review, sorting documents into issue categories and noting possible privilege concerns. A paralegal pulls together emails, attachments, and intake notes into a matter brief before the partner can decide what needs attention.

Much of that work is necessary. Some of it is billable. Yet it can still create a margin problem.

Junior associate time commonly sits in the $200 to $400 per hour range once you look at the actual commercial value of the role. If a review takes 15 hours where a controlled first pass could reduce it to 6 or 7 hours, the firm has a decision to make. It can deliver the matter faster, improve the quality of senior review, take on more work, or pass some savings to the client in a competitive fixed-fee engagement.

There is also billable-hour leakage. Many firms estimate that each attorney loses four to six hours a week on unbilled document handling, intake follow-up, matter updates, and internal administration. That does not always appear as a line item on a P&L. It appears as a partner working late, delayed invoices, and a team that says it is busy while capacity remains hard to measure.

Across a year, that mix of missed matters, non-billable effort, delayed handoffs, and slow first-pass review can reasonably sit in an $80,000 to $250,000 leakage band for a firm in this segment.

The point is not to chase a generic efficiency number. The point is to locate the specific hours and missed opportunities that you can fix without compromising professional standards.

Start with controlled drafting and research

Claude’s growth through enterprise coding tools is relevant because it demonstrates a useful model for law firms. Give the system a bounded task. Provide the permitted source material. Require a structured output. Keep qualified human review as the final control.

For contract drafting, a first test could be narrow:

• Use approved precedent documents and a client-specific clause playbook.
• Ask the model to produce an issue list rather than a final legal opinion.
• Require clause references, quoted source language, and identified assumptions.
• Keep the associate or supervising lawyer responsible for verification and advice.
• Exclude highly sensitive or restricted files until your data handling requirements are confirmed.

The output might be a contract summary that flags limitation of liability, termination, governing law, indemnities, data processing, assignment, and unusual payment terms. It could compare those provisions against a playbook and produce a table of proposed fallback language.

That is useful because it changes where the lawyer spends time. The associate is no longer locating every clause from scratch. They are checking the material issues, applying legal and commercial judgment, and shaping advice for the client.

Legal research can follow the same pattern, with tighter controls. An AI system can help turn a research question into a search plan, summarise the authorities you provide, extract competing principles, and draft a memo structure. It should not be treated as a source of authority in itself. Hallucinated citations, stale law, and missing jurisdictional context are real risks.

Your workflow should require the lawyer to verify every authority in your recognised legal research sources and confirm the reasoning before it reaches a client or court. AI is the first-pass analyst, not the final legal researcher of record.

If your firm needs help setting those controls, Omni Ops is where we design agents and workflows around actual operating rules rather than a loose prompt library.

What an AI agent workflow looks like in practice

The highest-return opportunity is often not a standalone chatbot. It is an agent that follows a defined process from trigger to handoff.

Take a new employment law inquiry received through a website form at 9:45 pm. The client says they have been dismissed, have a hearing in two weeks, and want advice on a potential unfair dismissal claim.

The Matter Triage Agent receives the form submission and associated email. It identifies employment as the practice area, extracts the dates and urgency markers, checks the information against the firm’s qualification rules, and prepares a short brief for the relevant partner or intake team.

That brief might contain:

• The client’s contact details and preferred contact time
• A concise summary of the legal issue
• Key dates and time-sensitive risks
• Questions that still need to be answered
• A fit score based on the firm’s defined criteria
• Suggested next action and nominated owner

It does not decide whether the firm can act. It makes sure the human decision-maker receives the relevant information in a usable form, rather than a raw email that has sat unread for half a day.

Now add the Intake Voice Agent. It answers a call after hours, asks approved intake questions, captures the matter details, and can collect information needed for an initial conflict screen. It can then book a consultation directly into the firm calendar within rules you define.

That does not mean it gives legal advice. It should clearly state its role, avoid legal conclusions, obtain consent where required, and escalate immediately when a caller is distressed, unsafe, or dealing with a time-critical deadline. The objective is a fast human-quality response and a clean handoff, not an automated law practice.

You can see where Omni Voice fits in a broader intake operation, especially for firms that lose calls at lunch, after hours, or during court appearances.

For existing matters, the Document Review Agent can perform first-pass review on contract packs, discovery batches, correspondence, and matter files. It can group documents by topic, identify key clauses, flag potential issues for review, summarise parties’ positions, and draft an associate-grade memo using your template.

A litigation team might give it a defined set of discovery materials and ask it to identify documents relating to representations made before a transaction. The agent can produce document references, excerpts, issue tags, and a summary of potential relevance. The supervising lawyer then assesses context, privilege, admissibility, and strategy.

This is how the work needs to be designed. The agent handles the repeatable pass. The lawyer owns the legal conclusion.

Put professional responsibility ahead of novelty

Law firms cannot use AI carelessly. Client confidentiality, privilege, conflicts, retention, supervision, and data residency all matter. The fact that enterprise vendors are selling at scale does not remove your professional obligations.

Before you load client information into any model or connect it to a document system, get clear answers to basic questions:

1. What data will the tool receive, store, and retain?
2. Is customer data used to train the underlying model?
3. Who can access prompts, source files, and outputs?
4. Can you restrict the tool to approved matter folders?
5. What audit trail exists for inputs, outputs, and human approval?
6. Which tasks require mandatory lawyer verification?
7. How will your team handle errors, complaints, or a potentially privileged document?

Your policy does not need to be 40 pages to start. It needs to be specific enough that a junior lawyer, paralegal, and administrator know what they can use AI for, what they cannot use it for, and when they must escalate.

You should also tell clients what your firm’s approach is when it is relevant to their engagement. Some clients will expect efficiency. Some will have strict data requirements. A clear position beats vague assurances.

For a practical starting point on the front-end process, download the AI Client Intake Checklist for Law Firms. It is designed as a working checklist for mapping response times, conflict checks, routing rules, escalation points, and the information your intake team needs before a consultation.

If you want the direct worksheet version, you can also download the intake checklist here.

A 30-day test that produces a real answer

Do not start by asking which AI tool is best. Start by choosing one workflow where delay, volume, and repeatability are visible.

For many practices, client intake is the best first test because you can measure it cleanly. Set a 30-day baseline for incoming calls and forms. Track response time, consultation bookings, source of inquiry, qualified matter rate, and matters that were lost because the prospect engaged someone else first.

Then design the agent workflow with actual firm rules.

Define what the Intake Voice Agent can ask. Define what triggers an escalation. Define how a potential conflict is recorded. Define which calendars it can book. Define the exact disclosure language it uses. Review a sample of calls and intake summaries every week.

For document review, start with a repeatable document type. A standard commercial agreement, a known discovery category, or a recurring due diligence pack is better than a complex litigation file with no consistent structure. Measure first-pass hours, rework, missed issues, turnaround time, and partner satisfaction with the resulting memo.

You are looking for evidence, not a dramatic demonstration.

If a workflow saves six hours a week but creates three hours of correction work, it is not ready. If it cuts first response time from six hours to six minutes, increases booked consultations, and gives your team better information, you have a strong case to expand it.

This is also why the broader Omni platform matters. The value is not in one prompt. It is in connecting voice, operations, approvals, data sources, reporting, and people into a process your firm can manage.

The commercial question is simpler than it looks

Anthropic’s reported quarter tells you that enterprise leaders have decided AI is worth a budget line. Your firm’s decision should be smaller and more concrete.

Ask where you are spending expensive human time on predictable work. Ask which client-facing delays are costing you matters. Ask which tasks are causing lawyers to spend evenings on administration that nobody will pay for.

Then work from a number.

If your firm is losing 10 worthwhile consultations a year because after-hours callers do not get a response, calculate the average first-year matter value. If associates are spending 12 hours each week on first-pass document work that can be reduced with proper review controls, calculate the capacity value. If partners are manually triaging every inquiry, quantify what happens when that work reaches them as a one-paragraph brief instead.

Those numbers will show where the $80,000 to $250,000 leakage band is hiding in your practice.

You do not need to be the first firm in your market to use Claude or any other model. You do need to avoid becoming the firm that still treats rapid response, structured research, and efficient first-pass review as optional when clients see those standards elsewhere.

See Omni for law firms to understand the workflows we assess and the controls we build around them.

Turn the signal into a firm-specific plan

The next step is not a vendor demo. It is an honest view of how work enters, moves through, and leaks out of your firm.

A 60-minute Omni Audit identifies the process gaps that are worth fixing first. You leave with three outputs: a map of the leakage points, a prioritised set of AI agent opportunities, and a practical implementation path. There is no deck full of generic use cases.

Book a 60-min Omni Audit if you want to assess intake, document review, matter triage, or research workflows against the economics of your own practice.

Anthropic’s enterprise growth is a signal, not an instruction to rush. The firms that benefit will be the ones that use it to improve specific work, retain lawyer oversight, and measure commercial results.

For a clearer view of the process, review the AI audit for law firms, then Book my Omni Audit.