Managing the Budget and Cost of Legal Research

What does legal research cost today?

Legal research today carries three costs. The largest is the researcher's time, billed by the hour, so every authority read along the way becomes a line on the invoice. The second is the subscription the firm pays for its research platforms. The third is the research time clients decline to pay for, which the firm absorbs as a write-off.

The hours dominate. A legal question of any depth means hours of reading, billed at the researcher's rate before a word of the memorandum is written. An inquiry that reaches dozens of authorities consumes days of billed time. The platform subscription is a fixed cost at the firm level, negotiated by contract and paid whether a given matter uses it heavily or not at all. Firms recover it in different ways, some as a billed disbursement, many as overhead built into rates.

The write-off is the least visible cost. Many outside counsel guidelines cap research time, require advance approval for it, or treat online research charges as the firm's overhead. Research the client will not pay for still has to be done, so the difference is written off. The true cost of legal research on a matter is therefore larger than its invoice line, with part of it carried by the firm.

The pressure on all three costs comes from the client. Legal departments review invoices line by line, so a research entry has to justify itself in terms the client can see. An entry reading three hours of research on a preemption question shows the client the hours, never what the hours found. A memorandum the client can read justifies its price better than any time entry, a point the fixed-fee questions below return to.

Why does the billable hour limit legal research depth?

The billable hour limits legal research depth because every hour of reading is an hour the client pays for. The litigation budget caps those hours. Every additional authority read is an additional charge, so thoroughness itself has a price. Research therefore stops where the budget stops, not where the inquiry is complete.

The cap does its work quietly, in the researcher's choices. Queries are framed narrowly to keep the reading manageable. Promising trails go unfollowed because each one costs another authority, another hour. The marginal case is skipped on a judgment that is really a budget decision. Chapter 3 documents the same constraint across the search era. The budget, not the query technology, decides how much of what a search returns ever gets read.

Thorough research has always been a luxury, priced in hours beyond what most matters could bear. The same legal question receives an afternoon on a small matter and a week when the exposure warrants more. The difference is never in the law. It is in the budget.

The professional risk hides in the skipped reading. A controlling case the researcher never reached does not announce itself. It surfaces in the opposing brief, at oral argument, or in the order denying the motion. The budget decision that excluded it was invisible when it was made, because no one can name the case an unread search result contains.

How do AI legal research tools charge?

AI legal research tools charge in three ways. Subscription pricing adds a per-seat license to the firm's technology costs. Consumption pricing meters the work as it runs, by the task or through credits the firm draws down. Fixed-fee pricing sets the price of a defined research task before the work begins.

For a firm managing matter budgets, the difference among the models is whether the cost of a task is known before the task starts. A subscription makes the firm's total predictable while leaving the cost of any single task a matter of allocation. Consumption pricing ties cost to use but leaves the total open until the work is done. A fixed fee attaches the price to the task itself, so the number can sit on a matter budget before the research runs.

The models reflect the technology each was priced for. The per-seat subscription was built for the chatbot era, when a user's AI use meant a single prompt or a short exchange around one. The seat price rested on an assumption of limited use, with light users subsidizing heavy ones across the firm's licenses. Agentic workflows break that assumption. Agents that plan the inquiry, run the searches, read every authority they uncover, and write the memorandum use many times the model capacity of a chatbot exchange. The cost of serving a seat no longer fits under a flat license. Vendors are answering with consumption pricing metered in credits or tokens. That model moves the cost uncertainty onto the firm, where a research-heavy month arrives on the bill only after the work has run.

Servient prices agentic legal research the third way, as a fixed fee per research task, stated in virtual hours. The fixed fee makes the budget firm, a number accepted before the research runs. It also places the cost of the agents' depth on Servient rather than the firm, because a task whose inquiry runs deep carries the same price as one that resolves quickly. The virtual hour makes the value legible. A virtual hour is the estimated time a lawyer would take to perform the task. Lawyers and clients already value legal work by the hours a task takes, so a price stated in virtual hours shows the client what the agent's work is worth in the profession's own measure.

How does agentic legal research change the cost of a research task?

Agentic legal research changes the cost of a research task by taking the hours out of its largest component, the reading. The agents read every authority they uncover and write the legal research memorandum, the work that once accumulated billed time. The lawyer's time on the task narrows to the initial guidance and the review of the finished work product.

The deeper change is that depth stops driving cost. Under hourly research, thoroughness and price rise together. That is why the budget has always set the depth. An agent that reads the complete body of authority it uncovers does not cost more for having read all of it. Complete research no longer carries a premium price.

A firm can verify two effects in its own numbers. The first is the time a matter spends on research, which compresses when the memorandum arrives written and validated rather than accumulating researcher hours. Review stays efficient because the work product arrives pre-audited, with each legal proposition beside its supporting passage, the validation audit Chapter 8 describes. The second is consistency. Every matter's research starts from the same reading depth over the same curated law, so quality no longer varies with who ran the search.

Agentic legal research also ends the write-off. Under hourly research, the reading a client capped or declined to pay for still had to be done, so the firm carried it as unbilled time. An agent's research consumes no researcher hours, leaving nothing to write off. The hours a researcher once spent reading return to the firm as billing capacity, deployable to work the client pays for in full. Every hour redeployed to billed work raises the firm's realization rate. The margin gain comes from using the billing resources the firm already has more effectively, without raising a rate or hiring a lawyer.

The change also moves research earlier in the matter. Under hourly economics, deep research waits for the moment that justifies its cost, often summary judgment. A research task priced as a fixed number can run at intake, when the strength of each claim still determines the pleading, the staffing, and the value of the matter.

Fixed-fee pricing completes the change by making the cost a term rather than an estimate. A research task quoted before it runs can be accepted, budgeted, and passed to the client as a firm number, which no hourly research task has ever offered.

Can a law firm bill clients for AI legal research?

Yes. A law firm can bill clients for AI legal research, within Rule 1.5's requirement that fees and expenses be reasonable. What the firm may charge depends on the fee arrangement. An hourly firm bills the time actually spent, including the lawyer's review of the work product, never the hours the AI saved. A fixed fee prices the work product itself.

The hourly answer is the settled one. In Formal Opinion 93-379, issued in 1993, the ABA's ethics committee stated that a lawyer who has undertaken to bill on an hourly basis "is never justified in charging a client for hours not actually expended." Opinion 512 carries that reading to AI-assisted work. A lawyer who completes a research task in one hour with an AI tool bills that hour, together with the review time the work product requires. The efficiency itself belongs to the client. That economic logic pushes AI-assisted legal research toward the fixed fee.

The fixed fee is where agentic legal research fits the fee rules best. Rule 1.5 has never limited a fee to the hours worked. Its reasonableness factors weigh the skill the work demands, the lawyer's experience and ability, and the results obtained alongside the time and labor involved. The lawyer's expertise in directing the technology, the guidance that frames the inquiry, and the judgment applied in reviewing the memorandum all support a fixed fee for the research task that is not a function of hours, provided the fee remains reasonable.

The technology cost follows the same fork. Under Opinion 512's reading, an AI charge incurred for a particular matter may be passed through to the client at actual cost, disclosed in advance, while a general subscription stays overhead, because a monthly seat cost cannot be allocated to any one matter. Servient's pricing is built for the pass-through side of that line. A fixed fee per research task, stated in virtual hours, attaches the technology cost to a single matter by design, so the firm can pass it through or fold it into the fixed fee it quotes for the task.

The gains run to both sides of the engagement. The firm passes the technology cost through instead of carrying it as overhead, no longer writes off research a matter needed but its budget would not bear, and earns a margin on the work product that hourly billing never supported. The client keeps a firm budget for the research task, pays less than the same research would cost conducted manually, and receives a depth of research that no longer varies with the budget attributable to the case. Whatever the arrangement, settle it in the engagement letter, and check the client's outside counsel guidelines first, since many now speak to AI charges directly.

What does ABA Formal Opinion 512 say about fees for AI-assisted work?

ABA Formal Opinion 512 applies Rule 1.5's reasonableness requirement to fees for AI-assisted legal work. A firm billing by time charges for the time actually spent, not the hours the AI saved. The out-of-pocket cost of an AI tool used on a matter may be passed through at actual cost, disclosed in advance, while a general-use subscription is overhead.

On time, the opinion distinguishes work from savings. Lawyers who charge an hourly rate, the opinion states, "must bill for their actual time." The lawyer's review of the AI's work product is time actually worked, billable like any hour of review. Time the AI saved was never worked, so it is never billed. Time spent gaining general competence with AI tools is the firm's own investment in its duty of technological competence, a cost the opinion assigns to the firm rather than the client.

On expenses, the analysis follows how the tool is bought. A charge incurred for a particular matter may be itemized as a disbursement, disclosed to the client in advance, at actual cost without markup. A subscription the firm buys for general use is overhead of the kind rates already recover, so it is not itemized unless the engagement letter says otherwise.

The opinion states that the reasonableness requirement governs alternative arrangements with equal force. A flat fee for AI-assisted work must bear a reasonable relationship to the value of the work performed, judged by the Rule 1.5(a) factors, among them the time and labor required, the difficulty of the question, and the fee customarily charged for similar work. A fixed price for a defined legal research task sits comfortably inside that frame, because the deliverable states the value the fee is measured against.

How should a law firm budget legal research on a matter?

A law firm should budget legal research on a matter by task, not by an undifferentiated pool of hours. Identify the points where the matter will need research, set a cost for each, and revisit the list as the matter develops. Fixed-fee pricing makes each number firm. Hourly research makes it an estimate that needs a cap.

Research arrives at predictable points in a litigation matter. Intake calls for a claims analysis of the fact pattern. Motions practice and the disputes that surface in discovery each raise defined questions. Summary judgment demands the deepest research of the matter. A budget built on those points anticipates the research instead of absorbing it.

For any task still run by the hour, scope it in writing before it starts. State the question, set the hours, and decide in advance what finding would justify going deeper, so the decision to extend is made on the merits rather than discovered on the invoice.

Keep the client inside the number. A research budget the client has seen is a fee conversation that never becomes a fee dispute. It is also the practical form of the fee communication the engagement letter contemplates.

When does a fixed fee for a legal research task make sense?

A fixed fee for a legal research task makes sense when the task has a defined work product. A legal research memorandum on identified issues, a claims analysis of a fact pattern, and a research analysis of an opposing brief each qualify. A defined work product lets the fee state exactly what the client is buying.

The definition is what makes the fee work. A fixed fee attaches to a deliverable, so a task whose deliverable can be named can be priced before it runs. The write-off problem largely dissolves with it. A client who balks at open-ended research hours will pay a stated price for a stated work product, because the value can be judged against the price before the work is done.

Stating the fee in virtual hours keeps the price legible. The client reads what is being bought in the hours a lawyer would have spent, the measure the profession already uses to judge the worth of legal work.

Hourly treatment still fits the genuinely open-ended inquiry, the investigation whose scope emerges only as it runs. Even there, staging keeps the budget honest. Run a bounded first task at a fixed price, read the memorandum, and scope the follow-up from what it found.

How do you measure the ROI of AI legal research?

You measure the ROI of AI legal research against your own baseline, in a pilot on matters you know. Tally what a research task cost the old way, in hours billed, hours written off, and days elapsed. Compare the same task run by the agent, priced as quoted plus the lawyer's review time.

The cost comparison is the narrowest measure. Set beside it the elapsed time from question to memorandum, since days matter to a motion calendar in a way invoices never show. Then compare the work product itself. Read the two memoranda side by side for completeness, for authorities the earlier research never reached, and for whether the quality holds across researchers of different experience.

The record also outlives the pilot. The same file that supports the adoption decision documents the firm's supervision of the tool, the reasonable-efforts record Chapter 9 describes.