Clause libraries and AI drafting: keeping your own standards
By the ChamberOne team 4 minute read
Ask a general AI tool for a limitation of liability clause and you will get something that looks right. It will be fluent, reasonably structured and drawn from an average of everything the tool has seen. It will not be your firm's clause. It will not reflect the positions you have negotiated over years, the wording you have learned to avoid, or the amendments a client insisted on last spring. For corporate and commercial lawyers, the real value of AI drafting depends on what you give it to work from. This article is about building that foundation.
Key points
- Generic AI wording reflects an average of unknown sources, not your firm's positions.
- Start with the clauses you negotiate most, each with a preferred version, a fallback and a note on what to resist.
- Tell the tool to use only your clauses and to flag anything it adds from outside.
- Review the library after each significant deal and again every year.
The problem with average text
Generic drafting produces generic risk. A clause pulled from an unknown source may favour the wrong party, assume a different legal system, use terms that do not fit the rest of the contract, or leave out a carve-out you always include.
The danger is not obvious error. It is smooth, plausible wording that nobody examines closely because it reads well. A lawyer who accepts it has adopted somebody else's position without noticing.
What a clause library is
A clause library is a set of approved clauses, each with notes. At a minimum it holds the preferred wording. A better one holds alternatives for different situations, the reasons for each position, and a note on when to use which.
Many firms already have one, in the form of a precedent bank, a shared folder or the memory of a senior associate. The task is to turn that into something structured that both people and tools can use.
Building the first version
- 1
Pick the ten to fifteen clauses you negotiate most often: liability, indemnity, termination, confidentiality, governing law, payment, assignment and the like.
- 2
For each, write the firm's preferred version, one fallback and a note on what you would resist.
- 3
Record the document type and the side you usually act for, since a supplier's clause differs from a customer's.
- 4
Mark the jurisdiction. A clause that works in one legal system may fail in another.
- 5
Name an owner for each clause, who reviews it on a set date.
Using a library with AI
There are two ways to link a library to a drafting tool. You can give the tool the relevant clauses in the prompt, and tell it to use them and to flag any departure. Or you can use software that searches your own documents and clause bank for the closest match and shows where it came from.
Either way, the instruction matters. Ask the tool to propose wording only from the supplied clauses, to say when none fits, and to mark anything it has added from outside. That last instruction keeps your standards visible. A lawyer can then see at once which parts are firm positions and which are the tool's invention.
- Reflects the internet, not your firm
- May favour the wrong party or assume a different legal system
- Reads well, so nobody examines it closely
- Proposes wording only from the supplied clauses
- Says when no clause fits
- Marks anything added from outside
Keep the lawyer's judgement in the loop
A clause bank does not know the deal. The commercial context, the relative bargaining power and the client's appetite for risk all change what the right clause is. Treat suggestions as a starting draft and read the whole document as a whole, because clauses interact. A neat indemnity can be undone by a definition three pages away.
Check also that defined terms line up. A clause lifted from a library often uses a term that the new contract defines differently or not at all.
Who may change a clause
Decide who can edit the library. If everyone can, it fills with half-finished variants. If only one partner can, it stalls. A workable middle is that any lawyer can propose a change with a reason, and a named reviewer for each clause approves or rejects it within a set time.
Show the status of each clause: approved, under review or retired. A drafter in a hurry should see at a glance whether a clause is safe to use. When a tool is fed from the library, feed it approved clauses only.
Keeping the library alive
A library goes stale quickly. Law changes, cases change positions, and the firm learns from negotiations. Set a rhythm: after each significant deal, ask whether a clause should be updated. Once a year, review every clause against current law and market practice, and remove anything nobody uses.
Keep the history. When a clause changes, note why. A year later, someone will ask, and the reason will be gone.
- 1Finish a significant deal
- 2Ask whether a clause should change
- 3Update it and note why
- 4Review every clause once a year
- 5Remove what nobody uses
- Then back to step 1.
Choosing what goes into the first batch
Resist the urge to catalogue everything. A library of forty clauses that nobody trusts is worse than twelve that everyone uses. Ask the lawyers who draft most often which clauses they rewrite each time. Those are the ones where the firm has a view that is not written down.
Start with one document type, such as a services agreement. Get the clauses, fallbacks and notes right for that, try them on three live matters, and ask the users what was missing. Then extend to the next document type. This keeps the work manageable and keeps the lawyers engaged, since they see the benefit early.
Include a short note on the counterparty's most common objections and how the firm usually responds. When a tool or a junior drafts a response, that knowledge prevents a repeat of a concession the firm decided long ago not to make.
Where this does not help
A library speeds up routine drafting. It does little for a novel deal, a first-of-its-kind structure or a clause that has to be built from principle. In those cases the lawyer writes, and AI may help with phrasing or with a check for gaps. It also does not replace the need to read the counterparty's draft with care.
Templates in ChamberOne
- Draft from your own templates, with merge fields filled from the matter, the client and the parties.
- Export the result to Word.
What it does not do
- The site does not describe a clause bank or AI clause suggestions, so this article's advice on clause libraries applies to whatever tools you choose.
- Click-to-sign is a simple electronic signature. Some documents need formalities it does not meet.
- A lawyer reviews every draft.
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
AI contract review: a first-pass checklist for lawyers
What to ask an AI to flag on a first read of a contract, what it tends to miss, and where a lawyer must read every word.
5 minute readWhat AI document analysis extracts well, and what it misses
What AI document analysis does well on parties, dates, clauses and citations, and what it misses, with ways to spot the gaps.
5 minute readAI-assisted engagement letters: what to automate and what stays human
Merge fields and templates handle the routine parts of an engagement letter, while scope, fees and limits of liability need a lawyer's decision.
5 minute read