How do you calculate ROI on legal AI?

Per-deal hour savings × associate billing rate, minus annual platform cost, minus deployment cost (training + integration). The math is straightforward; the inputs are what firms underestimate. Hour savings are typically 80-120 hours per mid-market deal, billing rates depend on the firm, platform cost is enterprise-tier.

Does the ROI work for small firms?

It depends on deal volume. A firm doing fewer than 5-8 mid-market deals a year has a harder ROI case because the per-deal savings have to cover the same platform cost spread across fewer deals. Above ~10 deals/year, the ROI is straightforward.

What about fixed-fee vs. hourly?

On fixed-fee deals, the savings are pure margin. The fee is fixed; lower hours means higher margin. On hourly deals, the savings change what the hours are spent on rather than reducing the total. Both models benefit, but the visibility of the benefit differs.

How do you measure quality, not just speed?

Recall against ground truth (what fraction of real issues did the tool find?), precision (what fraction of flagged items are real?), and rewrite percentage (how much does the partner edit before the memo ships?). Speed without quality is regression.

What's the deployment cost?

Variable by firm. Training (associate hours over the first 90 days), integration (DMS connection, data-room provider connections, firm-branded templates), change management (partner adoption, workflow updates). For a mid-sized M&A practice, total deployment cost typically runs in the range of one to two months of platform fees.

The ROI of Legal AI for M&A: An Attorney's Calculator

This is the ROI math attorneys actually use, with concrete numbers from real deployments. Built for the partner who has been asked "what's the business case?" and needs an answer in numbers, not in marketing claims.

The hours saved (per mid-market deal)

Typical mid-market deal, 1,000-3,000 documents, manual diligence vs. AI-augmented:

| Workstream | Manual hours | AI-augmented hours | Savings | |---|---|---|---| | Data room ingestion + classification | 8-12 | 1-2 | 7-10 | | First-pass risk review | 80-120 | 15-25 | 65-95 | | Amendment chain resolution | 15-25 | 2-4 | 13-21 | | Memo drafting | 25-40 | 8-12 | 17-28 | | Disclosure schedule (sell-side) | 80-120 | 20-30 | 60-90 | | Redline review | 20-30 | 8-12 | 12-18 |

For a buy-side mid-market deal: ~115-170 associate hours saved. For a sell-side deal with disclosure schedules: ~175-260 hours saved.

These ranges come from real customer deployments. Your mileage will vary based on deal complexity, document mix, and how the firm has staffed similar deals historically. The math holds across most of the mid-market.

The dollar value

At a mid-market law firm with associate billing rates around $700-1,000/hour, the per-deal savings translate to:

Buy-side mid-market: $80k-$170k in attorney time saved
Sell-side mid-market: $120k-$260k

For a firm doing 30-50 mid-market deals a year, that's $3M-$10M+ in annual saved attorney time. Platform cost runs in the low-six figures for an enterprise contract; deployment cost is modest. Net ROI is meaningful.

For a smaller practice doing 5-10 mid-market deals a year, the savings still cover platform cost, but the ROI is tighter. Deployment-cost amortization matters more.

What changes about the deal economics

Two scenarios, depending on how the firm bills:

Fixed-fee deals. The fee is fixed; hour reduction is pure margin. A $400k fixed-fee mid-market deal that took 600 associate hours of diligence pre-AI now takes 480 hours. The 120-hour reduction at $800/hour is $96k of margin captured. The firm wins; the client pays the same price; the partner runs a better practice.

Hourly deals. The hours go down, but billings track. The savings show up as redirected hours. The associate spends 50 hours fewer on first-pass review and 50 hours more on negotiation prep, structuring questions, and partner-mentorship moments. The associate's career trajectory improves; the firm's deal capacity expands; the bills are similar.

Both models benefit. Fixed-fee deals see the savings in margin numbers; hourly deals see the savings in deal capacity and associate-leverage numbers. Either way, the practice is materially better.

Quality, not just speed

Speed without quality is regression. The right ROI math includes quality metrics:

Recall against ground truth: what fraction of real issues did the AI-augmented review find compared to the manual review on the same deal?
Precision: what fraction of AI-flagged items are real issues, not false positives?
Rewrite percentage: how much does the partner edit the AI-drafted memo before it ships?

The bar should be: equal or better recall than manual review, ≥70% precision (so the partner doesn't read everything anyway), <30% memo rewrite (so the team adopts unprompted). A tool that misses these is a regression dressed up as an upgrade.

The deployment cost

The often-overlooked input. ROI is per-deal-savings × deal-count minus platform cost minus deployment cost. The first three are easy to estimate; the deployment cost trips firms up.

Realistic deployment cost, mid-sized M&A practice:

Training: 8-15 associate hours per associate, in the first 90 days, around two real deals. At billing rates this is real money but it's amortized across a long horizon.
Integration: DMS connection (iManage, NetDocuments), data-room provider connections (Datasite, Intralinks, ShareFile), firm-branded templates for memos and schedules. One-time engineering plus ongoing maintenance.
Change management: Partner adoption, workflow updates, internal communication, edge-case discovery. Less measurable but real.

For a mid-sized M&A practice, total deployment cost typically runs in the range of one to two months of platform fees. After that the ROI is per-deal savings × deal count.

How to compute ROI for your practice

A simple framework:

Estimate hours saved per deal type at your practice. Use the table above as a starting point and adjust based on your historical staffing.
Multiply by associate billing rates to get per-deal dollar savings.
Multiply by annual deal count for that deal type.
Sum across deal types for total annual gross savings.
Subtract: annual platform cost (roughly low-six-figures for an enterprise contract), and amortized deployment cost (one-time).
The result is annual net ROI.

For most mid-sized and large M&A practices, the math is positive by a wide margin from the first year.

The other ROI

The numerical ROI is the easy part. The structural ROI is harder to quantify and matters more long-term:

Junior associate retention. Associates who get freed from grunt diligence and pushed into high-leverage work earlier report higher career satisfaction. Lower turnover, lower training cost.
Partner capacity. Partners reviewing pre-prepared findings rather than directing from-scratch reviews can run more deals at higher quality.
Client experience. Faster turnaround. More thorough analysis. Fewer missed issues. Repeat client business.
Competitive position. A firm running AI-augmented diligence wins more fixed-fee bids than a firm not running it. Visible in win rates over a year or two.

These don't show up cleanly in spreadsheets but they show up in the firm's P&L over time.

Companion reading

How to Roll Out Legal AI at a Law Firm — 90-day rollout playbook
Evaluating Legal AI Tools — buyer's framework
AI Due Diligence: An Operational Playbook — workflow mechanics

If you want to see Mage on a real deal: request a demo. We will run end-to-end diligence and walk through the result, including the time-to-deliverable comparison against your manual workproduct.