Attorney Practice Guide
How to Calculate ROI on AI for a New Law Firm
A practical method for measuring AI through business won, revenue protected, and administrative capacity returned using the new firm's own workflow baseline.
Reviewed
Resource record
A practical method for measuring AI through business won, revenue protected, and administrative capacity returned using the new firm's own workflow baseline.
- Reviewed
- Aug 13, 2026
- Evidence
- Firm-level measurement model
- Useful artifact
- AI return worksheet
Review note: Sources, workflow, and professional limits reviewed for the current edition.
“AI saved time” is not enough to justify a workflow. A new law firm should be able to say what changed, how it was measured, and whether the result improved the business or merely moved work somewhere else.
Use three outcome categories:
- Business won
- Revenue protected
- Administrative capacity returned
Then subtract the complete cost of the technology and implementation. This creates an honest ROI model without relying on an industry-average savings claim that may not resemble the firm's practice.
The method is intentionally workflow-specific. The NIST Generative AI Profile organizes AI risk work around governing, mapping, measuring, and managing a defined use. A law firm can apply the same discipline commercially: identify the job, measure the current state, test the changed state, and keep the result connected to its source records.
Begin with a workflow baseline
Choose one recurring job. Examples include reviewing a client document packet, preparing a consultation from intake, following up on incomplete records, or moving a retained client from engagement to matter opening.
For several representative examples, record:
- Attorney minutes
- Staff minutes
- Waiting time
- Number of client correction cycles
- Errors or omissions requiring rework
- Systems opened
- Messages sent
- Whether the work delayed a consultation, bill, filing decision, or matter stage
Do not estimate the improved workflow yet. Measure the current path first.
1. Business won
AI and automation may support business won when they improve a measurable client path.
Examples include:
- A useful public answer produces a qualified inquiry.
- Guided intake lets the firm make a faster consultation decision.
- Appointment preparation improves the appearance rate.
- Prompt, accurate follow-up helps a suitable prospect complete engagement.
- Capacity returned lets the firm accept an additional matter it otherwise would have declined.
Use actual records:
Business won = collected revenue from attributable retained matters
Do not count every lead at the firm's average case value. A lead is not revenue. If the firm wants to show pipeline separately, label it as qualified pipeline and keep assumptions visible.
For capacity-based growth, be conservative. The fact that a workflow returned five hours does not prove the firm sold five additional billable hours. Count additional work only when it was actually accepted and collected.
2. Revenue protected
Revenue protected is value already available to the firm that a process helped prevent from being lost or delayed.
Examples include:
- A qualified inquiry recovered after an approved follow-up
- A stalled engagement completed
- Unbilled completed work surfaced and invoiced
- An invoice or payment plan returned to a current status
- A retainer replenishment issue surfaced before work stopped
- A matter returned to a billable stage after the exact client item arrived
Use the agreement, invoice, payment, matter, or campaign record.
Revenue protected = collected value attributable to the recovered event
If a tool surfaces a possible $8,000 matter but the client does not retain the firm, do not report $8,000 protected. Report one qualified opportunity surfaced and zero collected protected revenue.
3. Administrative capacity returned
Capacity is the easiest outcome to measure and the easiest to exaggerate.
Use:
Hours returned = old workflow time − new workflow time − correction time
Include the time spent reviewing AI output, correcting mistakes, maintaining templates, and handling exceptions. A workflow that cuts first-pass work from 30 minutes to 8 minutes but creates 10 minutes of cleanup returned 12 minutes, not 22.
You can assign a cost value:
Capacity value = hours returned × loaded internal hourly cost
Use internal cost for administrative savings, not the attorney's highest billing rate, unless the time was actually converted into collected work. Keep recovered billable revenue in the business-won category so it is not counted twice.
Complete ROI formula
For a defined period:
Total benefit = attributable business won + attributable revenue protected + capacity value
Total cost = software + setup + training + review + maintenance + switching cost
ROI = (total benefit − total cost) ÷ total cost × 100
Also report the raw numbers. A percentage without the underlying period, workflow, and assumptions can be misleading.
Attorney review is a real implementation cost, not overhead to omit from the calculation. ABA Formal Opinion 512 addresses competence, confidentiality, supervision, output review, and reasonable fees when lawyers use generative AI. Those professional obligations are one reason a credible ROI model records review and correction time instead of assuming every machine-produced minute is fully recovered.
An illustrative example
Assume a new firm pilots document follow-up for one month. These are fictional numbers for method demonstration, not a DocketBuddy performance claim.
- Old review and follow-up time: 18 hours
- New review, correction, and exception time: 7 hours
- Capacity returned: 11 hours
- Loaded internal value: $75 per hour
- Capacity value: $825
- One stalled signed engagement completed after the missing records arrived
- Collected value attributable to that recovery: $1,200
- Software and allocated setup cost for the month: $350
Total benefit is $2,025. Net benefit is $1,675. The illustrative ROI is approximately 479 percent.
The most important facts are not the percentage. They are the 11 measured hours, the one collected recovery, the $350 complete cost, and the specific workflow that produced them.
Keep an ROI receipt beside the work
For each AI-enabled workflow, preserve:
- Workflow name
- Baseline period and sample
- Source event
- Machine job
- Attorney review time
- Correction or failure time
- Business event affected
- Collected value, if any
- Hours returned
- Assumptions
- Owner and review date
This prevents a quarterly dashboard from becoming detached from the matter, campaign, or administrative job that created the number.
Include quality and risk measures
Financial ROI does not excuse a weak workflow. Track:
- Source accuracy
- False positives and false negatives
- Material omissions
- Attorney acceptance, correction, and rejection rate
- Client confusion or repeat requests
- Deadline or filing exceptions
- Data incidents or policy deviations
- Situations where the workflow correctly stopped
A workflow may be worth keeping even if the immediate financial return is modest because it prevents a serious recurring error. Label that separately as risk or quality improvement rather than inventing a dollar amount.
Measure one pilot before measuring the platform
Do not attempt to prove “the ROI of AI” across the entire firm on day one. Prove one source-to-action chain:
- A known event occurred.
- The system performed a defined job.
- The attorney reviewed the result.
- A business or capacity outcome changed.
- The complete cost was recorded.
DocketBuddy's interactive ROI workflow calculator helps establish the initial cost model. The New Solo program sets a baseline around business won, revenue protected, and administrative capacity returned as the opening workflows are configured and rehearsed.
Article feedback
Was this helpful?
What should we improve?
Thanks for telling us
We’ll use this to improve the guide.