Most midsize law firms are not deciding whether to adopt AI, they are deciding where to focus it. That question matters more than most firms realize because the wrong focus produces inconclusive results, skeptical partners, and a proof of concept that’s too thin to justify the next investment. The right focus makes the commercial logic visible quickly, builds the internal case using real data, and makes the path to the next work type obvious.

This is a framework that midsize law firms can use for making that choice well.

Volume is the wrong variable

The most common mistake in selecting an AI entry point is choosing on the basis of volume alone, picking the task that takes the most time and assuming that is where AI-driven efficiency will deliver the most value. Volume matters, of course, but it is not the primary variable.

Two variables that matter the most are verification burden and pricing model.

The verification burden variable

Verification burden is the time and expertise required to check, validate, and stand behind an AI output before it reaches a client or is relied upon in a legal matter. For some tasks, such as drafting the first version of a standard letter, summarizing a document, or researching a legal question in a trusted and authoritative system, the verification burden is relatively low. The lawyer reviews, adjusts, and approves. The time saving is genuine, and the risk is manageable.

For other tasks, like compliance checks, document extraction in which completeness is critical, or jurisdiction-sensitive advice, the verification burden may be so high that it absorbs most or all of the time saving. Worse, outputs that appear credible but contain errors can create hidden costs, such as rework, increased risk, and reputational exposure. The return on investment (ROI) calculation on these tasks must account for the full cost of verification, not just the speed of the initial output.

This is why not all AI tools are equal. A professional-grade system grounded in authoritative legal sources with clear citation trails can significantly reduce the verification burden on research and analysis tasks compared to a general-purpose tool that produces plausible-sounding but unverifiable output. When evaluating where to start, the question should not be Will AI do this task faster? but rather Will we be able to verify the output efficiently enough that the time saving is real?

Document extraction in which completeness is critical, or jurisdiction-sensitive advice, the verification burden may be so high that it absorbs most or all of the time saving.

Clearly, the right tool matters here. A professional-grade system reduces the verification burden structurally, and lawyers can check outputs efficiently because the sourcing is transparent. A publicly available AI tool, on the other hand, may produce fluent but opaque answers that place the full burden on the reviewer. For many midsize law firms, verification by design is not a nice-to-have; instead, it’s what determines whether the time savings is real or illusory. Firms that underestimate the verification burden and overstate the value AI is delivering are also the ones most likely to face a difficult client conversation when outputs fail or require a significant correction — one that is considerably harder to have than is the entry-point conversation.

Recommended action: Before selecting an AI tool for a specific work type, determine whether the tool produces outputs with clear sourcing and citation trails that make verification straightforward. Also, ask yourself whether your firm has a defined review process that ensures lawyers are genuinely checking outputs, or are being persuaded by plausible-sounding answers. If you find that either situation is not the case, then the verification burden will be higher than the time saving suggests.

The pricing model variable

The pricing model is the second variable, and for many midsize firms, it’s the more immediately clarifying one.

For work billed on an hourly basis, AI efficiency creates a conversation that most firms are not yet having but will need to. Currently, more than 90% of legal fees still run through the billable hour, which means most firms have not yet needed to confront what AI-enabled efficiency means for their revenue model. That will change.

Market standard realization runs at around 85%, which means roughly 15% of associate time does not make it to the client's bill. In fact, for midsize firms in which a significant proportion of associate work is already being written off, AI may make the business case through reducing write-offs alone, without any change to billing model, because fewer hours lost to non-billable or written-down work, the better the realization will be on the hours that remain.

Beyond write-off reduction, however, firms have three options for capturing AI's efficiency value: i) they can take on more matters with the freed capacity; ii) they can invest the freed time in higher-value work that justifies higher hourly rates; or iii) they can develop new service lines that include more fixed-fee or subscription models that capture the efficiency gain structurally rather than matter by matter. Most midsize firms will need some combination of all three options.

Market standard realization runs at around 85%, which means roughly 15% of associate time does not make it to the client's bill. In fact, for midsize firms in which a significant proportion of associate work is already being written off, AI may make the business case through reducing write-offs alone, without any change to billing model, because fewer hours lost to non-billable or written-down work, the better the realization will be on the hours that remain.

Beyond write-off reduction, however, firms have three options for capturing AI's efficiency value: i) they can take on more matters with the freed capacity; ii) they can invest the freed time in higher-value work that justifies higher hourly rates; or iii) they can develop new service lines that include more fixed-fee or subscription models that capture the efficiency gain structurally rather than matter by matter. Most midsize firms will need some combination of all three options.

This is why not all AI tools are equal. A professional-grade system grounded in authoritative legal sources with clear citation trails can significantly reduce the verification burden on research and analysis tasks.

For work that’s priced on a fixed-fee basis, the commercial logic is more immediate. If the task takes less time, the margin on that matter improves directly. The client pays the agreed fixed fee, and the firm delivers the work more efficiently. Then, that efficiency gain goes directly to the bottom line. There is no billing conversation to manage, and no client expectation to reset

This is why fixed-fee, repeatable work types — such as conveyancing, standard employment matters, straightforward contract reviews, or routine regulatory filings — tend to be firms’ clearest starting points for demonstrating AI’s commercial value. The proof of concept is measurable within weeks rather than months.

However, there is a harder question to answer: What if your highest-volume, most AI-suitable work is not currently priced on a fixed-fee basis? Some firms have answered this by creating a fixed-fee entry point by proactively repricing a well-defined, repeatable matter type that was previously billed by the hour. Of course, that’s a commercial decision not just a discovery exercise, but it requires a conversation with clients before adoption. For those firms willing to have that conversation, they are not just solving the immediate problem, they are getting ahead of the repricing transition that AI will ultimately require of most legal practice areas.

Recommended action: List your firm's five highest-volume fixed-fee work types — these are your best candidates for immediate proof of concept. If (a likely scenario for most midsize law firms), then identify where write-offs are highest. That is where AI-enabled efficiency is most likely to make the business case on existing hourly work, without any change to the pricing model.

Three criteria that reveal where AI will actually pay off

Identifying the right entry point for your firm’s AI initiative means asking three questions about each potential work type.

1. Is this work AI-suitable? The most AI-suitable work tends to be document-heavy, research-intensive, or work that is heavily templated. In short, tasks where the value that AI adds is speed, consistency, and coverage (the ability to search more comprehensively than time would otherwise allow) rather than judgment. Work that requires bespoke legal analysis, nuanced client relationships, or significant professional discretion is not necessarily unsuitable for AI support, but it is not the right entry point.
2. Is the verification straightforward? Can a lawyer review the AI output efficiently and stand behind it with confidence? If the answer requires checking every line against primary sources, or if an error in the output would have serious consequences that are difficult to detect, the verification burden may undermine the ROI.
3. What is the pricing model? Fixed-fee work provides the clearest and fastest demonstration of value. If the firm's most AI-suitable, most verifiable work is currently billed by the hour, identify whether the write-off rate makes the efficiency case independently of any billing model change.

There is a fourth consideration that the three criteria above do not capture, and that’s the cultural signal that the entry point choice sends internally. Starting with a core practice area matter type rather than an easier entry point — even if the proof of concept is slower to establish — signals that AI is a strategic capability. Neither is wrong, but the choice is a culture decision as much as a commercial one.

Recommended action: Score your prospective AI-enabled work types against these three criteria. The right entry point scores well on all three, being AI-suitable, straightforward to verify, and containing a fixed-fee or high write-off rate on hourly work. Then ask: What does this choice signal to partners about what AI means for this firm?

Moving forward down the path

Most midsize firms are not starting from zero. AI tools are already in use across most legal organizations, formally and informally. In some cases, they are through official adoption programs, in others through individual lawyers or practice groups working ahead of any firm-wide strategy. The question for most firms is therefore not how to begin, but how to turn experimentation into structured evidence and generate the kind of proof that builds the internal case for the next investment decision and makes adoption defensible rather than ad hoc.

Overall, this focus should strive to make commercial logic visible, build the internal case using real data, and make subsequent direction in the firm’s AI journey obvious.

By using the above framework, midsize law firms can establish a business case to support that journey. However, the framework only works if it is measured. Most firms have no baseline against which to assess whether AI adoption is delivering value, and that makes it impossible to build the internal case for the next investment, or to answer the client when they ask, What has your AI adoption actually changed for us?

By choosing the right entry point and defining what success looks like before rollout begins, midsize firms can create that baseline. You do not need to start again; you just need to start knowing what you are learning.

‍The third article in this series will address what comes next: What lawyers actually do differently with the freed time, and how to have the client conversation that unlocks the full value of that change.‍