A law firm does not earn a return because an AI tool produces a summary quickly. Return appears when a complete matter stage improves and the firm converts that change into additional profitable work, lower write-offs, avoided external cost, changed staffing need, better fixed-fee margin or another financially approved outcome.
This guide provides a measurement method for commercial law firms. All figures are illustrative, not Binarify results or industry benchmarks. The firm’s finance, practice, technology, risk and professional-responsibility owners should validate the measures and treatment.
Choose one matter stage and unit
Define a repeated unit such as:
- one prospective-client intake packet;
- one conflict-search preparation;
- one agreement reviewed against a question list;
- one document in a disclosure or due-diligence population;
- one chronology event set;
- one research memorandum; or
- one monthly matter-status report.
Split the data when practice area, matter complexity, document type, language, jurisdiction or billing model changes the work materially.
Record a representative baseline:
- eligible units started and completed;
- lawyer and staff time for setup, preparation, review and correction;
- elapsed time and work in progress by stage;
- rework, write-downs and write-offs linked to the stage;
- fixed-fee budget performance where applicable;
- external review, discovery or temporary-staff cost;
- adoption and unsupported-case volume; and
- material quality, confidentiality and client-service events.
Use matter-system events, document records, billing data, a short time study and a reviewed sample. Time entries alone may not show non-billable preparation, correction or waiting.
State the route from workflow change to value
Write a testable chain before the pilot. For example:
A source-linked contract review matrix reduces associate preparation and partner correction on a defined fixed-fee service. The team completes the same reviewed work within budget while material omission and correction rates remain within the agreed limits.
Another valid hypothesis might convert capacity into additional matters without a planned contractor. A third might reduce unrecoverable time on intake for enquiries the firm cannot accept.
If there is no credible conversion action, report saved time as operational capacity. Do not multiply every saved hour by a headline billing rate and call the result profit.
Calculate realised hours
Annual gross hours released
eligible annual units × measured minutes saved per unit ÷ 60
Annual realised hours released
gross hours × adoption rate × supported-case rate × workflow availability
Suppose a defined review workflow covers 1,200 eligible documents a year and reduces complete preparation and lawyer-review time by 14 minutes per document. Expected adoption is 75%, supported-case coverage is 85% and workflow availability is 97%.
1,200 × 14 ÷ 60 × 0.75 × 0.85 × 0.97 = 173.1 realised hours
The 173.1 hours remain capacity until the firm applies them to a documented commercial or cost action.
Convert capacity without double counting
| Benefit route | Evidence needed | Financial treatment |
|---|---|---|
| Additional matters | Demand, capacity and contribution after delivery cost | Count contribution, not fees billed |
| Fixed-fee margin | Same scope and quality, lower complete delivery cost | Use observed cost change by matter type |
| Write-off reduction | Billing reason and changed workflow | Count only the observed attributable change |
| External cost avoided | Supplier invoices and replacement scope | Deduct new internal and technology cost |
| Overtime or temporary staff | Approved baseline and changed requirement | Count actual reduced spend |
| Hire avoided or deferred | Approved workforce plan and timing | Recognise from the date the cost would start |
| Faster turnaround | Retention, pricing, capacity or cash mechanism | Keep operational until finance approves a value |
Do not add the value of released hours to the profit from matters completed with those same hours. Assign each realised hour to one benefit route.
Treat hourly billing honestly
In an hourly matter, less production time may reduce billable time rather than increase immediate revenue. The commercial case may depend on additional demand, a revised pricing model, lower write-offs, stronger client retention or the ability to redirect senior attention.
ABA Formal Opinion 512 discusses reasonable fees and actual time in the US model-rules context. Billing obligations vary by jurisdiction and engagement. The responsible firm owners must decide how technology use affects fee arrangements and client communication.
Keep three measures separate:
- hours no longer required for the task;
- hours redeployed to other work; and
- financial benefit actually recognised.
Model work in progress and constraints
Use arrivals and completions by week:
closing work in progress = opening work in progress + units received − units completed
Track stage and age. Faster extraction will not improve matter turnaround when client instructions, partner review, counterparty response, disclosure decisions or court deadlines are the actual constraint.
Measure whether work moved to another role. An associate may spend less time preparing a summary while a partner spends more time checking sources and correcting context. The complete unit must improve.
Include full ownership cost
Count:
- workflow mapping and review-protocol design;
- product licences, model usage, storage and hosting;
- integration with intake, practice, document, research and finance systems;
- data clean-up, metadata and permission remediation;
- supplier, security, privacy, professional-risk and client review;
- evaluation sets, testing and double review;
- lawyer and staff training;
- human review and ongoing quality sampling;
- monitoring, incident handling and change approval;
- internal product ownership and support; and
- exit, export and rollback work.
Include the native-product option in the comparison. A capability already available in Clio, iManage, NetDocuments, CoCounsel, LexisNexis, Microsoft 365 or another approved platform may have a stronger return than custom development.
Calculate ROI and payback
First-year net benefit
realised financial benefits − first-year incremental costs
First-year ROI
first-year net benefit ÷ first-year incremental costs × 100
Payback point
The first month in which cumulative realised benefits exceed cumulative incremental costs.
Use a monthly model when rollout, adoption, matter demand and billing cycles change over the year. Present conservative, expected and upside scenarios. Do not combine every favourable assumption into one forecast.
Binarify legal AI value ledger
This proposed artifact makes each benefit claim reviewable.
| Measure | Baseline source | Pilot comparison | Conversion action | Financial owner | Quality guardrail |
|---|---|---|---|---|---|
| Intake preparation time | Intake and practice-system events | Same practice area and channel | Capacity or staff-cost plan | COO or practice manager | Missing party and conflict-search rate |
| Review time per document | DMS events and time study | Same question list and document type | Fixed-fee margin or capacity | Practice head and finance | Material omission and correction rate |
| Write-offs | Billing reason codes and review | Same matter type and fee basis | Observed reduction | Finance partner | Client outcome and complaint rate |
| External review cost | Supplier invoices | Equivalent approved scope | Avoided supplier spend | Matter partner | Recall and precision sample |
| Matter turnaround | Matter-stage timestamps | Same complexity and period | Approved capacity or client value | Matter owner | Rushed review and missed deadline rate |
| Operating cost | Vendor bills and support time | Observed production volume | Deduction from benefit | Technology and finance | Availability and failed writes |
Reject a benefit row without a baseline, conversion action, owner and quality guardrail.
Keep professional quality as a release condition
Pair financial measures with:
- incorrect or omitted parties in conflict preparation;
- material facts, clauses or documents missed;
- unsupported or fabricated sources and authorities;
- lawyer corrections, rejections and overrides by reason;
- wrong-client, wrong-matter or wrong-version results;
- confidentiality, privilege, access and information-barrier events;
- advice, engagement or filing steps taken without authority;
- client complaints and required disclosure failures; and
- downstream corrections after work is used.
The Solicitors Regulation Authority’s 2026 warning notice says professional responsibility remains with regulated people and firms. A pilot that saves preparation time while weakening source verification, confidentiality or lawyer supervision has failed its release condition.
Use industry research as context, not a forecast
The Thomson Reuters 2026 AI in Professional Services Report reports wider use of generative AI across professional services while only a minority of respondents said their organisations track ROI. This supports measuring workflow and business outcomes rather than counting licences or prompts.
The report covers several professions and markets and comes from a legal-technology and information provider. It does not establish the return available to a particular firm or workflow.
Run a decision-grade pilot
Agree the eligible population, baseline, financial treatment and stop conditions before implementation. Run shadow mode on a representative set, then a controlled assisted phase with named lawyers and quality sampling.
At the decision point, answer:
- Did the complete matter stage improve after setup and lawyer verification?
- Which released capacity has a funded or revenue-producing use?
- Did material quality, confidentiality and client obligations remain within agreed limits?
- What will licences, operation, review and future changes cost?
- Which matters or documents remain unsupported?
- Who owns benefit realisation, professional approval and rollback?
Apply the calculation to client intake and conflict-check preparation or source-linked legal document review. Explore our AI consultancy for law firms, or book a 30-minute conversation to define one baseline.