The flat-rate AI period won’t last forever

Most law firms adopting AI right now see a simple monthly subscription. A drafting tool might cost a few hundred dollars per user. A document platform may bundle AI features into an annual contract. It feels predictable enough to put in the technology budget and move on.

That pricing model is unlikely to hold as usage rises.

The cost of AI is tied to work done. Every contract reviewed, deposition summarised, discovery batch classified, and long email thread analysed consumes computing capacity. Vendors can absorb some of that cost while adoption is still being encouraged. Once AI is embedded in daily workflows, usage-based charges, tier limits, overages, and premium processing fees tend to appear.

For a law firm, the risk isn’t that AI becomes too expensive to use. The risk is more specific. A document-heavy matter that looked profitable at intake can quietly absorb hundreds or thousands of dollars in AI processing costs that no one assigned to the matter budget.

That matters for firms between $1 million and $25 million in annual revenue. At this size, you usually don’t have a dedicated legal operations team reconciling every software transaction against every matter. Partners are focused on clients. Finance sees vendor invoices after the cost has landed. Associates use the approved tools because they need to move work forward.

The right response is to calculate your likely AI cost per matter before pricing changes force the issue. Then build workflows where AI reduces costly manual effort without creating a new category of unmeasured spend.

See Omni for law firms if you want to identify which legal workflows can deliver that kind of controlled return.

Start with the work, not the vendor invoice

A firm can’t manage AI cost by looking only at its monthly technology bill. You need to understand the units of work being sent through AI systems.

For a commercial litigation practice, those units might include:

• Pages of discovery reviewed and categorised
• Documents extracted from a virtual data room
• Privilege or issue flags generated
• Witness interview notes summarised
• Draft chronology entries created
• Client emails classified and routed
• Draft motions, outlines, or research memos produced

For a property, employment, family, or general commercial practice, the units differ, but the principle stays the same. You might measure contracts, intake calls, attachments, matter updates, due diligence checklists, or correspondence threads.

The first mistake firms make is treating every AI interaction as equally valuable. It isn’t.

Using AI to create a rough internal summary of a three-page email is cheap and often useful. Running first-pass analysis across 80,000 discovery documents may be highly valuable, but it needs a defined matter budget, a quality-control step, and a client billing decision. The scale changes everything.

Ask three questions for each repeated task:

1. How often does this task happen?
2. What does it cost us in human time today?
3. What could it cost in AI usage when processing is charged by volume, complexity, or speed?

You don’t need false precision. Most firms can begin with estimates based on the last 10 to 20 comparable matters.

For example, if junior associates spend two days on a first-pass document review, their nominal cost might be $200 to $400 per hour depending on your market and staffing model. Some of that time is billable, some is written down, and some becomes invisible internal administration. An AI workflow may reduce the first pass dramatically. But if it sends large document sets through a premium model without controls, the cost could be material.

The comparison isn’t AI cost versus zero. It’s AI cost plus human review versus the actual labour cost, write-off risk, turnaround time, and client experience of the current process.

Build a per-matter AI cost model now

You don’t need a complex finance project to get started. Build a working model in a spreadsheet or your existing matter reporting system.

Track each matter type separately. A fixed-fee employment investigation should not be compared with a bet-the-company litigation matter. A standard commercial lease review should not be compared with a distressed M&A transaction.

For each matter category, record:

Measure	What to capture
Matter type	Employment dispute, commercial litigation, conveyancing, M&A, family law, and similar categories
Fee model	Hourly, fixed fee, capped fee, retainer, or blended
Expected document volume	Files, pages, emails, attachments, contracts, or discovery records
Current human effort	Hours by partner, associate, paralegal, and administrative staff
Current write-downs	Time not billed, discounted, or exceeded against a fixed fee
AI-supported task	Intake classification, extraction, first-pass review, summarisation, drafting, or routing
Expected AI volume	Number of calls, documents, pages, or requests
AI budget	A sensible allowance based on current vendor data and a future usage-price range
Review requirement	Who verifies output before it reaches the client or affects legal judgment

The goal is not to forecast a vendor’s exact future pricing. No firm can do that reliably. The goal is to see where your economics are exposed.

A useful first benchmark is to group matters into low, medium, and high AI-volume bands. Low-volume matters may only use AI for intake and administrative summaries. Medium-volume matters may include document extraction and contract review. High-volume matters include discovery, due diligence, investigations, or large-scale file analysis.

Then decide where AI spend belongs.

For hourly matters, some firms may be able to recover approved technology costs where engagement terms and local professional rules allow it. For fixed-fee work, the AI allowance needs to be included in the matter price from the start. For contingency or speculative matters, the firm may set a strict internal limit until the case reaches a defined stage.

The key is that partners should be able to see the cost before a matter becomes unprofitable.

Document-heavy matters are where the pressure will show

Document review is the most obvious exposure because the volumes can change quickly.

A new instruction arrives. The client says there are “a few folders” of material. Two days later, your team receives 15,000 emails, shared-drive exports, spreadsheets, PDFs, chat records, and scanned attachments. A partner needs an early view of risk within a week.

Without a structured process, junior staff begin opening files. They build rough notes. They search manually. They spend days producing a first-pass view that may be partly billable, partly discounted, and partly written off when the client challenges the invoice.

This is where a Document Review Agent can be useful. The agent doesn’t replace legal judgment. It processes the first pass under a defined workflow. It can extract key clauses, group documents by topic, flag missing information, summarise stated positions, identify potential issues, and produce an associate-grade memo for review.

The end-to-end process should look like this:

1. A matter team opens a controlled review request and assigns a matter code.
2. Files enter an approved workspace with access controls and retention rules.
3. The agent identifies file types, duplicates, language issues, unreadable scans, and total volume.
4. The agent applies the agreed review lens, such as change-of-control clauses, termination rights, indemnities, key dates, privilege indicators, or disputed facts.
5. It creates a structured output with links back to source documents.
6. An associate validates the findings, corrects errors, and adds legal interpretation.
7. A partner reviews the final advice where the matter requires it.
8. Finance or legal operations records AI usage against the matter budget.

That final step is often missing. Firms measure the hours that follow the review, but not the AI work that enabled it.

Usage pricing will make that gap harder to ignore. If the system processes 10 times more material than expected, someone needs to know while the work is happening, not at the end of the month.

Intake is also an AI cost decision

AI cost control isn’t only about discovery. It begins before a new matter opens.

Many firms lose valuable enquiries because calls arrive after hours, during lunch, or while reception is managing another call. Industry ranges often suggest that 30% to 40% of after-hours legal intake does not convert when the caller waits too long for a response. The client doesn’t think of this as an intake workflow. They think their legal problem is urgent, and they call the next firm.

An Intake Voice Agent answers those calls, gathers the essential facts, runs an initial conflict-check workflow, captures contact details, and books an appropriate consultation directly into the firm’s calendar.

That process needs boundaries. It should not provide legal advice. It should state that clearly. It should collect only the information your firm needs at that stage. It should escalate urgent matters according to defined rules. It should never represent that a solicitor-client relationship exists before your firm confirms it.

From a cost perspective, intake is a strong place to build AI discipline because the unit economics are clear. You can measure calls handled, consultations booked, qualified matters opened, and cost per converted lead. You can also set simple routing limits so a lengthy, unqualified caller doesn’t trigger unnecessary processing.

The same logic applies to email and form submissions. A Matter Triage Agent can review incoming messages, classify the practice area, score fit against your criteria, identify urgency, and route the matter to the right partner with a one-paragraph brief attached.

This reduces response delays without giving every incoming message unlimited access to expensive processing. You decide what the agent reads, what it stores, what it escalates, and how long it retains the information.

For a practical way to map the intake side of this work, download the AI Client Intake Checklist for Law Firms. It gives your team a worksheet for response rules, conflict checks, handoffs, and measurement. You can also access the direct client intake checklist download if you’re ready to use it with your intake manager.

Put controls around AI spend before you need them

The firms that handle this shift well won’t try to ban usage-based AI. They’ll put commercial controls around it.

Start with matter-level guardrails:

• Set an AI budget at matter opening for document-heavy work.
• Require partner approval when projected usage exceeds the budget.
• Use lower-cost processing for classification and triage where it is fit for purpose.
• Reserve premium processing for work where accuracy, complexity, or turnaround justifies it.
• Separate exploration from production. A lawyer testing an approach should not automatically run a full corpus.
• Track rework. If associates routinely redo AI outputs, the claimed time saving may not be real.
• Keep a human reviewer accountable for legal conclusions and client-facing work.
• Review vendor terms for data handling, retention, training restrictions, and pricing triggers.

There is also a pricing decision. Your engagement letters and estimates need to keep pace with the work model.

If a matter involves predictable high-volume analysis, explain the process and assumptions early. Some clients will value faster issue identification and a more focused legal team. Others will expect the efficiency benefit to lower their fee. Either way, you need a clear position before the work starts.

One trades-business owner in our network described a similar problem with estimating software. The tool was cheap until the business started processing every enquiry, every revision, and every site image through premium features. The issue wasn’t technology adoption. It was that the owner had no cost unit tied to each job.

Law firms face the same management problem, with higher professional obligations and more complicated matter economics.

Find the leakage before adding more tools

AI can reduce the manual burden in a legal practice, but buying another subscription doesn’t automatically solve leakage.

Most firms have existing losses hiding in routine work. Attorneys commonly spend an estimated four to six hours a week on unbilled document handling, internal updates, intake follow-up, and matter administration. Some of that work is necessary. Much of it comes from poor routing, repeated data entry, unclear ownership, and information sitting in inboxes.

For a firm in the $1 million to $25 million range, the broader leakage opportunity can often sit in the $80,000 to $250,000 annual range. That is not a promise of savings from one tool. It is a signal to inspect the workflow carefully. The answer may be automation, better matter scoping, clearer handoffs, or a different staffing model.

Our Omni advisory work starts with the operating reality rather than a software demo. We look at where the work enters, who touches it, what gets delayed, where quality drops, and how costs should be measured.

If you want to prepare for AI usage pricing without guessing, Book a 60-min Omni Audit. It is a working session, not a sales deck.

What you should leave an Omni Audit with

A good audit should produce decisions you can act on.

In 60 minutes, we map one or two workflows that are creating the biggest combination of cost, delay, and client friction. For law firms, that is often intake-to-consultation, document review, matter triage, or the handoff between an associate and a supervising partner.

You leave with three outputs:

1. A workflow map showing the manual steps, handoffs, systems, and likely failure points.
2. A short list of AI agent opportunities, including where the Intake Voice Agent, Matter Triage Agent, or Document Review Agent could fit.
3. A commercial view of potential impact, including the data needed to estimate cost per matter and protect margins under usage pricing.

There is no deck to admire and no vague recommendation to “use more AI.” You get a practical next-step view of what should be automated, what must stay under human control, and what should be measured before you scale.

You can read more about the operating model behind Omni, but the better next step is to look at your own matters and workflows.

The coming AI cost shift doesn’t need to become a shock. Firms that count usage by matter now will have options later. They can change pricing, set limits, select the right tasks for automation, and show clients where speed and quality have improved.

Firms that wait may discover the problem after a high-volume matter has already consumed the margin.

See the AI audit for law firms, then Book my Omni Audit when you’re ready to put real numbers against the work.