Aimed at managing partners, innovation leads and IT directors at UK law firms running the build-vs-buy conversation internally. Both options have a place — the trick is knowing which problem belongs in which bucket.
The Off-the-Shelf Legal AI Market
The market for AI tools sold to law firms is now substantial. Harvey, Lexis+ AI, ContractPodAi, Luminance, Robin AI, Spellbook — every legal tech vendor has an AI proposition. The horizontal AI products (Microsoft Copilot, ChatGPT Enterprise) are also being adopted at firm level for general drafting, research and summarisation.
These tools have real advantages: they are available immediately, they have been tested on large volumes of legal documents, they handle a lot of the data security and compliance plumbing, and they come with vendor support. For drafting assistance, legal research, conversational document Q&A and general productivity, a well-chosen off-the-shelf tool is often the right answer.
They have meaningful limits when your use case is specific, your document types are unusual, your output format is fixed, or your workflow needs to integrate with systems the vendor was not built for.
What Off-the-Shelf Legal AI Does Well
Off-the-shelf tools excel at tasks that are general in nature and where the output is flexible. If you want to ask questions of a lengthy bundle, get a draft of a standard clause, or run free-text search across a document set for relevant precedents, a well-designed legal AI tool handles those tasks well.
They are also the right choice when your needs are common enough that the vendor has trained the system on similar materials to yours. The major legal AI vendors have processed enormous volumes of UK commercial contracts, court documents and regulatory filings — that domain coverage is a genuine asset for general work.
Where Off-the-Shelf Tools Fall Short
The limits become visible in several situations:
Your output format is fixed
If the output needs to look a specific way — a DD report in the firm's house style, an entry in a specific field in iManage or your matter management system, a spreadsheet column that feeds a downstream process — most off-the-shelf tools cannot deliver that without significant rework. A custom system produces output in exactly the format required, every time.
Your document types are non-standard
General legal AI tools are trained on common commercial documents. If your practice involves unusual document types — heavily negotiated bespoke agreements, sector-specific instruments, project finance documents, regulatory filings in a niche area — performance on those will typically be lower than on standard materials. A custom system trained and tested on your actual documents will outperform a general-purpose tool on accuracy for those specific types.
Your workflow involves non-document data sources
Some of the highest-value automation is not pure document review. Monitoring Companies House filings for borrower clients. Watching regulatory feeds for FCA enforcement updates. Pulling structured data from court listings or insolvency notices. Off-the-shelf legal AI is built around documents and text — custom systems can be built to handle arbitrary data sources and combine them however the workflow needs.
Your data residency requirements are strict
Some firms — particularly those handling highly sensitive matters, government work, or under sector-specific obligations — require document data stays inside specific geographic boundaries or is not transmitted to third-party AI infrastructure. Custom systems can be deployed on-premises or inside the firm's own UK cloud environment, giving complete control over where data is processed and stored.
The Cost Comparison
Off-the-shelf legal AI is typically priced per user per month — often £100 to £500 per user per month for professional-grade legal AI. For a 50-fee-earner team that is £60,000 to £300,000 per year, every year, regardless of usage. The cost is a per-seat tax on having lawyers, not a function of how much value you actually extract.
Custom automation has a higher upfront cost — typically £8,000 to £30,000 to build a focused system — but low ongoing costs (LLM API usage usually £100 to £500 per month). Over a three to five year horizon, a custom system built for a specific high-volume workflow is typically cheaper than a per-seat SaaS tool, even before you account for the higher accuracy that comes from being purpose-built for your documents.
The crossover depends on volume. For occasional or exploratory use, off-the-shelf is usually cheaper. For a defined, high-volume workflow that runs continuously — DD on a steady stream of mid-market deals, lease portfolio reviews, NDA processing — custom is usually cheaper by year two.
The Practical Decision Framework
The question to ask is: do I have a specific, defined, high-volume workflow where the output format is fixed and the input types are consistent? If yes, custom is likely the right call. If the answer is no — if you need general AI assistance across a range of tasks, or the use case is exploratory — start with an off-the-shelf tool and revisit custom automation once the workflows are clearer.
Most firms end up with both: off-the-shelf tools for general AI assistance across the practice, and custom automation for two or three specific high-volume workflows where the precision and output control of a bespoke system pays back the build cost quickly.
A Note on DIY
There is a third option: building it yourself, using AI platforms that make development accessible without deep technical expertise. For firms with capable in-house IT, this is worth considering for simple, low-stakes workflows. For anything involving privileged client data, complex document types, or production-grade reliability, the time investment in internal development typically exceeds the cost of commissioning a professional build — and the result is a system that needs ongoing internal maintenance rather than a delivered, tested solution.
If you want a steer on which workflows at your firm belong in which bucket, get a quote.