Skip to content

AI, Contract Drafting, and Gold-Standard Precedents: Some Thoughts Prompted by Allen & Overy’s ContractMatrix

Luna the Albino Alligator

I learned from this article on that "Allen & Overy is set to roll out an AI tool to assist with drafting contracts—marking a new phase of AI use among law firms." This is from A&O's press release:

ContractMatrix is a contract negotiation tool that uses AI and other techniques to free lawyers from processing and allows them to focus on strategic questions. In some cases it saves around seven hours when negotiating a contract. Expert human decision-making, and judgement, is at its heart.

And A&O's web page for ContractMatrix advertises these features:

  • Generative AI-assisted interrogation and drafting
  • Real-time access to your gold-standard precedents and policies
  • Inbuilt risk management and governance designed by A&O lawyers

I've been told that as used by A&O, the AI model itself isn’t trained on A&O contracts but does leverage them using retrieval augment generation (RAG), with each practice area leveraging their own templates. The same goes for clients—so they will use a system that leverages only their own templates or precedent contracts (to ensure confidentiality of client data). (RAG is an AI framework for improving the quality of responses generated by a large language model (LLM) by grounding the model on external sources of knowledge to supplement the LLM’s internal representation of information.)

So ContractMatrix uses RAG to avoid the nightmare scenario that underlies this 2022 blog post—having AI hoover up a stash of contracts, or all of EDGAR, and then regurgitate it, warts and all. I’m in no position to assess whether ContractMatrix works as advertised. Instead, I have a more basic concern about quality. Let's consider A&O's reference to "gold-standard precedents."

I'll assume that "gold-standard precedents" refers to templates—contracts structured for use in future transactions—or contracts created using a template. "Gold-standard" is puffery suggesting something of such good quality that it can safely be used as a benchmark. But I've found that in templates and in contracts that were presumably generated from templates, you can expect dysfunction in how contracts say what they say:

  • I've written about shortcomings in company contract templates launched with some fanfare. For example, see what I've said about templates created by IBM and GE Aviation.
  • I've also selected some company contracts at random and found them wanting. For example, see what I've said about Google and Salesforce contracts.
  • The same goes for templates created by trade groups. I had a quick look at FIDIC and ISDA templates.
  • I've critiqued contracts drafted by big law firms (here, here, and here), although generally I don’t bother, because I assume no one cares enough: BigLaw drafting faces challenges I describe in this 2011 article, and the incentives seem skewed to favor getting the deal done over quality and efficiency. 
  • I've examined and found wanting contracts offered, for free or for money, to anyone who wants to use them, the most recent example being my critique of Practical Law's mutual confidentiality agreement (here).
  • In 17 years of giving Drafting Clearer Contracts presentations, I’ve reviewed hundreds of company templates. With only a couple of exceptions, they’ve offered plenty for me to discuss with participants.

As regards contract substance, my writings on contract boilerplate (listed here) show that conventional boilerplate contains much that's unhelpful. And see the first three blog posts listed here for my critique of how some issues are addressed in standard M&A drafting. It’s safe to assume that subject-matter experts under strong editorial control would uncover many ways in which template substance could be improved.

So I suggest that gold-standard precedents are as rare as albino alligators. Just as most of us haven't seen an albino alligator, I suspect that most of us haven't seen anything that could objectively called a gold-standard precedent. So despite the precautions A&O has taken, ContractMatrix will likely replicate the systemic dysfunction of mainstream contract drafting.

To some extent, that dysfunction is a matter of expediency. Overhauling contracts is a nuisance—it takes time and money and requires expertise that isn't available at most organizations. And more generally, changing your templates is challenging, because it inflicts dreaded change on your contract process. But conceding to expediency doesn’t turn dysfunction into something that's gold-standard.

(Postscript 1: Some of you might be wondering, Who died and made Adams an arbiter of contract quality! Go here to find out more about me. For someone else's take on my urge to scrutinize contracts, see this 2018 blog post by Casey Flaherty.)

(Postscript 2: Why are templates bad? See this 2020 blog post.)

(Postscript 3: My new business Adams Contracts (you’re on Adams Contracts’ website) offers an alternative to tolerating dysfunction or overhauling templates—it offers a fresh start. See this blog post for more about that.)

(The photo of Luna, the albino alligator, is available here. I'm using it under a Creative Commons license.)