Law Firm AI ROI: Measure Client Value First
Thomson Reuters reports a growing AI value gap. Here is how smaller firms can measure client impact instead of tool activity.
Quick answer
Measure law firm AI ROI by client and firm outcomes. Licenses purchased, logins, and AI-generated draft counts are activity, not results. A useful scorecard ties one workflow to a service result: faster intake, fewer missing documents, quicker preliminary assessments, lower rework, more predictable billing, or clearer client updates.
Thomson Reuters has been reporting on an AI value gap. Its September 2026 analysis says 91% of surveyed professionals believe their organizations are falling short of AI's potential value, while 78% of clients see AI-enabled quality improvements as essential and only 6% believe providers are delivering them. Treat that as a market signal from one research program, not a universal census. The practical lesson holds anyway. Buying a tool is not the same as changing how a firm delivers work.
For a small firm, the strongest response is a 30-day pilot with numbers attached. Choose one recurring workflow. Name an owner, define the baseline, record corrections and exceptions, then compare the client experience before and after. Keep the responsible lawyer visible at the point where an AI-assisted process becomes advice, advocacy, or a client commitment.
This article is informational and not legal advice. Professional-responsibility, confidentiality, billing, data-protection, employment, and court-filing requirements vary by jurisdiction, client, and matter. Consult qualified counsel and the applicable bar or court authority before changing your firm's practices.
The September signal: an AI value gap, not just an adoption gap
The September 2026 Thomson Reuters Institute article describes a pattern most managing partners will recognize. Leadership approves AI tools, communicates an innovation strategy, and schedules training. Months later, usage is limited, lawyers remain uncertain, and clients notice no meaningful change in service.
The article works from a composite scenario, not one firm's case. Its value is the operating question it raises: did the firm change how it delivers value, or did it only add another application to the technology budget?
The same analysis reports that 35% of professionals say their AI ambitions are not reflected in daily work, and that 32% of clients have reconsidered, or plan to reconsider, their relationship with a firm they view as falling behind. For a smaller practice those are consequential signals. A firm does not need the largest budget to compete. It does need to know whether the investment improves a client interaction, a matter handoff, or a result the client can understand.
The American Bar Association's recent operations guidance makes a related point. AI can streamline repetitive work. Effective adoption still requires governance, workflow design, training, verification, and leadership. The guidance recommends starting with a contained, lower-stakes task and measuring time saved alongside corrections required. That beats reporting that the firm has implemented AI.
What client value means in law firm operations
A lower invoice is one kind of client value. So is the experience of working with the firm, and the quality of the decisions its process supports.
For a small firm, useful outcome categories include:
- Responsiveness: how quickly a new inquiry receives an accurate next-step message.
- Clarity: whether the client can understand status, responsibilities, and timing.
- Predictability: whether the firm can provide a more reliable scope, schedule, or billing update.
- Quality: whether review finds fewer unsupported statements, missing facts, or preventable errors.
- Capacity: whether lawyers reclaim time for strategy, judgment, and relationship work.
- Continuity: whether another team member can understand the matter without reconstructing its history.
Each category can be made concrete. Improve responsiveness becomes: reduce the median time from approved intake to first human follow-up. Improve quality becomes: cut missing-document rework in the selected workflow. A metric earns its place when a team can name the event that starts it, the event that ends it, and the person who can change the process.
Start with a value hypothesis
Write one sentence before you choose a tool. Connect a workflow to a client result:
If we use AI to organize approved intake information and propose the next task, the team will reduce the time to a complete matter-opening package without increasing conflict-review errors.
Three things sit in that sentence: a workflow, a benefit, a guardrail. It also tells the firm what not to claim. The pilot tests whether a narrow coordination step improves service while professional review stays intact. It does not try to prove that AI can practice law.
Set the baseline before you enable anything. Record the current median turnaround, handoff count, corrections, escalations, and how many client follow-ups the work generates. Without a baseline, a good result may reflect a busy month or a quiet one rather than an improvement.
Name a client-visible result wherever you can. A complete response package by the next business day. A status update within two business days of a material event. Fewer repeat requests for the same missing document. A clear target gives staff a reason to use the workflow and gives leadership a fair test of its value.
Three small-firm pilot patterns
1. Intake to matter opening
The system organizes what a prospective client supplied, flags missing administrative fields, suggests a practice-area queue, and prepares a proposed next-step checklist. A person still performs the conflict review, decides whether the firm can proceed, and determines what information must be requested.
Measure the time from submission to a complete review package. Count the fields discovered missing later, and the inquiries that get no timely response. Counting automatic classifications does not measure success. A classification is only useful if the client reaches the right human action sooner.
2. Matter status to client update
The system collects approved matter-status fields, spots stale tasks, and drafts a plain-language update. The responsible lawyer or team member reviews the draft, checks the promised dates, and sends the communication through the firm's normal channel.
Measure the time from a status-changing event to a reviewed update, the corrections needed before sending, and how many avoidable status questions arrive. A useful pilot may find that more automated messages are the wrong fix. The better fix is often a status record good enough for a human to answer accurately in one conversation.
3. Document collection to review readiness
The system compares an approved request list against documents received, flags apparent gaps, and prepares a review queue. It should not declare a matter complete without a human-defined rule and a responsible reviewer.
Measure incomplete handoffs, time spent reconciling folders, and the lag from receipt of the final expected document to review readiness. Keep an exception log for unreadable files, duplicate documents, unusual client circumstances, and any case where the automation proposed an incorrect status.
Use a balanced scorecard
A practical scorecard balances speed against quality, experience, and control. Four dimensions are enough for a first pilot.
Speed asks whether the selected workflow takes less time. Record median and worst-case turnaround. A median can improve while a small group of clients sits through a serious delay, so review the outliers.
Quality asks whether the output is more complete and reliable. Count corrections, missing facts, duplicate requests, escalations, rework. Have one reviewer classify errors the same way each time. A raw correction count is less useful if the definition changes halfway through the pilot.
Experience asks whether clients and staff can feel the improvement. Where it fits, ask one short client pulse question, such as whether the next step was clear. Ask staff whether the workflow reduced searching and handoff friction. Do not make a client answer a survey every time a system changes.
Control asks whether the process can be explained and governed. Track who approved the input, who reviewed the output, which version of the workflow ran, and what happened when the system was uncertain. Review access, retention, confidentiality, and vendor terms. The right controls depend on the matter and the tool.
Those four dimensions catch a common mistake: declaring success because a team produced more output. More output can raise risk when it creates more review, more client confusion, or more untracked data movement.
Do not confuse usage with adoption
Usage counts activity. Adoption is a sustained change in a process that people trust and clients can experience. A dashboard of prompt counts, active seats, or generated drafts may help with capacity planning. It is not proof of return on investment.
Ask five questions instead:
- Did the selected workflow change in a documented way?
- Did the responsible staff use the new steps consistently?
- Did a client-visible result improve without a new material risk?
- Did the firm capture exceptions and correct the process?
- Would the team keep the workflow if the pilot ended tomorrow?
Vendor announcements help a firm see what capabilities exist. Thomson Reuters describes HighQ and CoCounsel Legal as a governed collaboration environment that can connect matter management, permissions, audit controls, and AI-assisted work. That is a useful example of an integrated design, not proof that every firm should buy that product or that integration alone produces value. The buyer's test does not change: what result will improve, how will it be measured, and who remains accountable?
Governance and audit evidence
Governance is part of ROI. A process that cannot be trusted will not survive. Start with a written description of the pilot's purpose, approved inputs, prohibited inputs, review point, escalation path, and retention expectations.
Keep a small evidence set: baseline timing records, the correction log, exception notes, approval records, and a short decision memo. Collect no more client information than the pilot needs. If a vendor processes data, document the relevant contractual and security questions and involve the firm's privacy or security reviewer.
Make the review point visible in the workflow. A human should know what the system proposed, what source material supports it, and what still requires professional judgment. The firm should also be able to stop the automation without losing the underlying matter record or confusing the client.
A 30-day pilot plan
Days 1 to 5: define the test. Select one workflow, owner, baseline, client outcome, and guardrail. List the matters that are in scope and the matters that are excluded. Write a stop condition before the first test.
Days 6 to 10: prepare the process. Configure the smallest useful workflow and train the people who will use it. Create a review checklist. Test ordinary cases and at least a few exceptions. Confirm that permissions and data handling match the firm's requirements.
Days 11 to 24: run with human review. Collect the same measures for the old and new processes. Record corrections, delays, escalations, client questions, and staff workarounds. Do not change the success criteria because the first result is inconvenient.
Days 25 to 30: decide. Compare the scorecard, review the exception log, and interview the people closest to the work. Continue, revise, pause, or retire the workflow. If the pilot saved time but did not improve service, ask whether the baseline was wrong, whether the workflow was too narrow, or whether the saved time should go to a client-facing activity.
Management takeaway
Owning AI proves nothing on its own. The business case is a defined service process that works better for clients and professionals, and a firm that can explain and control what changed.
Start with one value hypothesis. Measure speed, quality, experience, and control. Keep a human accountable for the judgment that matters. Then decide whether the result is strong enough to expand.
Firms that want a broader operating model can read the small-firm AI playbook and the verification-first legal research workflow. To discuss a practical workflow assessment, contact The Hammer Lex.
Frequently asked questions
How should a law firm measure AI ROI?
Connect one workflow to a measurable result and compare it with a baseline. Include time, correction and rework rates, client or staff experience, and governance evidence. Financial savings belong in the picture too, as long as they do not replace quality and control measures.
Are AI logins a useful ROI metric?
Logins show activity, not value. They can tell a firm whether a tool is available and in use. A login never shows that a workflow improved or that a client received better service. Pair activity data with outcome and exception measures.
What is a good first AI pilot?
Choose a recurring, contained, lower-stakes coordination task with a clear owner and review point. Intake completeness, matter status preparation, and document-collection reconciliation are reasonable candidates. Avoid starting with an unbounded process where the firm cannot define success or monitor exceptions.
How can a firm protect client trust?
Limit the pilot to approved data and approved people. Document the review step, protect access, record exceptions, and describe accurately what the firm does. Check professional-responsibility, confidentiality, privacy, contractual, and court requirements before changing a client-facing process.
What if a pilot saves time but does not improve service?
Treat that as a useful result, not a failure to hide. Recheck the client outcome, see where the saved time goes, and inspect whether quality or control changed. The firm may need to redirect capacity to communication or strategy, narrow the workflow, or stop it if the net result is not worthwhile.
Sources
- The AI impact gap: Why law firms are failing at AI implementation, Thomson Reuters Institute, September 2026.
- Delivering the future of legal services with HighQ and CoCounsel, Thomson Reuters, September 2026.
- Transform Workflows and Operations for Good with AI, American Bar Association, 2026.
Sources
- https://www.thomsonreuters.com/en/institute/spotlight/ai-impact-gap-failing-at-implementation
- https://legal.thomsonreuters.com/blog/delivering-the-future-of-legal-services-with-highq-and-cocounsel/
- https://www.americanbar.org/groups/law_practice/resources/law-technology-today/2026/transform-workflow-and-operations-for-good-with-ai/
Turn AI activity into client value
Talk with The Hammer Lex about connecting matters, workflows, billing, and client communication so your technology investment produces visible results.
Book a demo