Many legaltech pitches to a law firm focus on a feature matrix and compare their platform against a few competitors. But relevance doesn’t really reside in what looks like a semantic contract search here, an automated diligence there, and so on.
What building it yourselves actually costs
Two moves in the market have made this question unavoidable. In March 2025, Cleary Gottlieb acquired the UK AI consultancy Springbok, and its founder along with roughly ten engineers became Cleary’s internal AI Acceleration team. In July 2026, Legora acquired the litigation startup Wexler to absorb its fact-intelligence framework. Both sent the same message to the law firms watching: if off-the-shelf software doesn’t fit, the alternative is buying the engineering capability outright, not building it from a standing start.
Which path is cheaper? Early estimates put a 30-person in-house legal AI team at around £4.1 million a year. Real-world execution in 2026 has shown that figure to be a floor, not a realistic budget. The fuller picture looks like this:
| Role | Headcount | Fully loaded cost per head | Annual total |
| Principal AI/ML engineers | 4 | £250,000 / $330,000 | £1,000,000 |
| Senior full-stack developers | 10 | £160,000 / $210,000 | £1,600,000 |
| Legal process/prompt engineers | 8 | £130,000 / $170,000 | £1,040,000 |
| Product managers & QA | 4 | £140,000 / $185,000 | £560,000 |
| Security & DevOps | 4 | £170,000 / $225,000 | £680,000 |
| Base salary & benefits subtotal | 30 | £4,880,000 | |
| Infrastructure, API tokens & cloud | — | — | £1,500,000 |
| Model fine-tuning & evaluation tooling | — | — | £600,000 |
| Total annual fully loaded cost | £6,980,000 (~$9.1M) |
For a 3,000-lawyer firm, that puts the honest run-rate at £6.5-8.5 million a year, before counting the 18 to 24 months it takes to build, test, and actually deploy anything production-grade. When we look at that timeline alongside the cost of the build, a law firm comes face to face with how long it is paying before it has something its lawyers can use.
Why this looks different depending on where a law firm sits
The build-versus-buy calculus isn’t the same everywhere. We have built a table below where law firms can identify which column describes which firm before a vendor tries to sell the firm on the wrong logic.
| Market | Dominant strategy | Typical spend threshold | The objection worth having ready |
| US BigLaw | Strategic acquisition and internal labs – AmLaw 50 firms running $10–20M innovation funds to acquire niche teams or build custom layers on Azure OpenAI | $8-15M/year | SaaS tools risk long-term vendor lock-in and per-seat fee inflation; internal builds generate firm-owned IP |
| UK Magic/Silver Circle | Hybrid M&A – firms like Cleary or A&O Shearman acquiring boutique consultancies to accelerate capability rather than starting from zero | £5–10M/year | Generic vendors often don’t meet SRA regulatory compliance and client data isolation requirements |
| India & APAC Tier-1 | Consolidated SaaS purchasing – local AI talent costs make full internal builds cost-prohibitive against licensing an existing platform | ₹ 8–20 Cr/year | Global vendor pricing in USD/GBP quietly erodes local margins when converted |
Telling the real pitch from the one borrowing your fear
The difference usually shows up in how vendors frame the number itself.
| What a feature- first pitch sounds like | What a pitch that’s actually done the math sounds like |
| “Our contract review engine is $500/seat/month with a 94% accuracy score on indemnification clauses.” | “Building this in-house requires 12 senior ML engineers, $1.5M in annual compute, and 18 months of dev time – a £5M + commitment. We deliver that same infrastructure tomorrow, at roughly 15% of that cost, with zero maintenance risk.” |
The left column is trying to win a features argument. The right column is answering the actual question your finance committee is going to ask, which is never “is this good software”. The question is: “What does the alternative cost, and how fast do we get value either way.”
Three things worth demanding before you sign, or before you approve a build
Ask for the build calculator, not just the demo. A vendor who understands your economics should be able to hand you an interactive model of what building this internally would actually cost including hiring, compute overhead, security certification, ongoing fine-tuning. If they can’t produce that comparison, they haven’t thought about your decision from your side of the table.
Ask who eats the maintenance cost. Foundation models turn over every three to six months. An internal team can spend up to 40% of its budget just keeping pace reworking pipelines every time a new model generation lands. A vendor is meant to absorb that cost across their entire client base instead of your firm absorbing it alone. If a vendor’s pricing doesn’t obviously reflect that difference, ask why.
Ask for the number in your terms, not theirs. Per-seat pricing and ARR are the vendor’s language, not yours. The number that actually matters to your partnership is what this protects in partner margin, or what it’s worth in billable-hour-equivalent efficiency. A vendor who can translate their pricing into that language has clearly sat with your P&L. One who can’t is still selling software, not understanding a law firm.
The build-versus-buy question was never going to be settled by a feature comparison. It’s a capital allocation decision, and it deserves the same scrutiny your firm would give any other seven-figure commitment regardless of which side of the table brings the spreadsheet first.
Lawfinity Solutions advises international law firms on cross-border legal market positioning. If the India corridor is a live question for your firm, we would be interested in a conversation. Lawfinity works with one firm per jurisdiction. Engagements begin with a single conversation about your firm’s current position and where the corridor question is live for you. Write to Prachi Shrivastava
