Veiw all resources

AI for Legal Contracts: How Your Startup Can Use It, Pros and Cons, and Best Practices

October 25, 2024
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Large Language Models (LLMs) in Artificial Intelligence (AI) have represented a huge breakthrough with countless applications, and that includes legal use cases like contract review.  While this tech could be inserted throughout your Contract Lifecycle Management (CLM) process, the typical division is for either post-execution (i.e. after a Contract has been signed) vs. pre-execution (i.e. while being drafted or negotiated).

Most earlier stage companies don’t have a large enough volume of contracts that post-execution AI applications like organizing documents, extracting renewal or notice obligations, understanding and analyzing legal data, is useful enough for the time and cost it takes to implement.

However, budget-sensitive startups and time-starved executives might be very interested in pre-execution AI solutions around contracts.  We’ll discuss good and bad use cases, pros and cons, and finally offer our verdict on the whole.

Keep in mind that you should only rely on AI for contract review if you have no other choice, or if you can do it in conjunction with a lawyer; humans and AI both make mistakes, but you can much better vet a human’s work product through various signals than you can vet the work product of AI.

What Are the Pros and Cons of Using AI for Legal Contracts?

Beyond the generalized pros and cons of using AI, there are a few ways that the pros and cons of AI applied in the context of law, especially for Silicon Valley startups, express more uniquely:

Pros

  1. Speed:  While Legal and Ops is proud to offer next business day turnaround time, we know not all legal vendors do, and sometimes you may need something even faster.  We got big brains, but not nearly the computing power of OpenAI, Anthropic, or whatever other LLM you’re using.
  2. Cost:  While lots of dedicated legal AI solutions can cost a few hundred dollars a month, other more generalized tools are much cheaper; either way, if you can safely outsource work to AI rather than a lawyer, you’ll definitely be saving money!

Cons

  1. Venture Capital Sensitivity:  We’re not aware of AI tools specifically trained on Silicon Valley standards around acceptable contract provisions.  What’s good for a telecom or services business may not be good for a software product with a different business model and expectations around diligence; in a word, being in a niche can make a difference.
  2. Accuracy:  Beyond general accommodation for a set of cultural expectations from VC, we all know that AI accuracy is not yet 100%, and it can often confidently make statements that would be untrue for any industry.  These probably hurt more.
  3. Accountability:  Often, software vendors can limit their liability to the amount of money you’ve paid them for their services.  Lawyers aren’t generally allowed to enter into limitations of liability with their clients–at least in California.  That means lawyers will be more accountable if something goes wrong.
  4. Ethics:  We don’t just say this because we’re biased–AI solutions have work to do in getting consent in the training data they use.  This is a thumb on the scale, in our humble opinion.

What Are Better and Worse Use Cases for AI in Startup Contracts?

The main idea is to use AI on provisions that (1) are not overly technical, (2) are not overly sensitive, and (3) are not context-dependent–unless you’ve given AI a lot of background on you, your counterparty, the deal, risks on your mind, key objectives, etc.

Better Use Cases

  1. Commercial terms; it’s relatively non-technical, not sensitive, and not super context-dependent to have AI make revisions to term, termination rights (especially if you are the vendor), auto-renew, payment terms, interest rates, etc.
  2. Acceptable Use, Warranties, Restrictions:  While these are important terms, if there is strong limitation of liability and indemnification language, then the risk is managed significantly.  They’re also a bit less technical and not scrutinized as closely by investors in diligence as some of the other terms presented here (of course, always strive for being as close to your template as possible).
  3. Confidentiality: Especially if changes are made mutual, then you have less reason to be concerned about a major risk allocation towards you; further, confidentiality breaches are something you will try very hard to avoid anyway, and in any case, tend to be harder claims to recover under.  Combined with the other reasons described here, it’s a term we’d be more comfortable delegating to AI.
  4. Miscellaneous terms:  Terms like governing law, arbitration, severability, entire agreement, notice provisions, and more, might be important in terms of the impact they have, but they are often less technical, context-dependent and more often dependent on the specific dispute that comes up, which is hard to predict.  

Worse Use Cases

  1. Limitation of Liability:  Both because this is one of the most important levers on risk allocation, and because it’s technical insofar as carveouts may be defined in certain ways that can have a big difference; and because AI will likely lack context on which party is more incentivized to cap this, you should delegate this with extreme caution.
  2. Indemnification: This is a technical, sensitive, and context-dependent provision insofar as different words will have different legal implications (e.g. negligence vs. gross negligence; arising from vs. relating to); it is sensitive because it is a major lever of risk allocation; and it is context-dependent since your own business’ practices, industry, the particular customer relationship, negotiation capital, and other factors, will each have a big impact on what outcome you want to shoot for as a starting point and as a minimum requirement for closing the deal.
  3. IP Rights:  AI doesn’t yet necessarily know very well the kinds of IP rights a future investor would expect to see in your material contracts during your future diligence; it may not differentiate between the kind of work product you may in fact intend to assign to a customer, like unique reports, vs. the kinds of rights you simply want to give as access, e.g. the right to access our cloud service.  Because an inadvertent assignment of IP can be fatal, this is a technical & sensitive area you should hesitate to outsource.

Our Verdict on AI for Legal Support

You should use this if you are very budget sensitive, or have worked closely with a lawyer to apply it only to those provisions that are not (1) technical (i.e. where a small word change makes a huge difference, or carries a different legal meaning), (2) sensitive (i.e. make a big difference in liability), or (3) context-sensitive (i.e. requires deep knowledge of the parties, transactions, or even industry–at least unless you’ve thoroughly caught your AI tool up).

Beyond Contract Negotiations

Drafting Contract Templates

Two things here:

First, this isn’t a use case where being “generative” is necessarily an advantage; instead, you’d rather stray as close to the beaten path as possible, so you don’t accidentally miss or add something you didn’t mean to.  

Second, this is an asset that you are going to use many, many times–so any shortcomings are are multiplied by the number of times you use it.  

For these reasons, we’d first, of course, recommend working with a lawyer.  If that’s not an option, the next best bet might be to look at very comparable peers and getting inspiration from their terms.  You could then lawyer an AI review into *that* new work product, first giving it context on our business, then asking it about what other risks it would consider, what other changes it would suggest making.

Corporate Workstreams

This is an area where you especially do not want to be “generative.”  You want to be super static, and in fact, ideally just like everyone else.  That’s true for a lot of reasons:  You’ll avoid extra scrutiny when your investors do diligence; you’ll avoid mistakes that have a more significant legal impact on the entity itself; you’ll avoid violating more specific tax and securities laws rules that can be very technical and sensitive.

We would not recommend using AI, therefore, to do things like issue option or stock grant agreements, prepare board and stockholder consents, make changes to your Charter, or other corporate workstreams.

General Advising 

While much of the same analysis as above applies, we would say a great use case is to save lots of time that you’d otherwise spend with a lawyer getting the basics on facts.  For example, if you’re doing a SAFE round, consider uploading the YCombinator (YC) primer on SAFEs into your AI agent and asking it to give you feedback based on your questions; same with other heavier legal documentation.

Once you have a deep lay of the land, you can set up time (1) to confirm your findings, especially in the context of Silicon Valley Venture Capital, (2) ask any outstanding questions, and (3) make a game plan about what kind of documents you need, and the best way to do it.

Tl;dr:  Great for educating yourself more efficiently on the basics or help digesting resources put out by the subject matter experts that matter most to you; less good for relying on entirely.

What AI Tools Can You Use for Legal Contracts?

We won’t speak for these products, but they’ve established some considerable presence, so may be worth checking out, especially for pre-execution workstreams (e.g. especially  negotiating contracts).

  • LegalOn is a market leader in AI-enabled contract review and redlines with built in AI-answers and summaries of terms and questions you may have.
  • DraftPilot takes in your preferences from playbooks you plug in, and calls out suggested changes that can be auto-inserted into third party contracts.
  • Pincites is a YC-backed AI contract review tool that embeds into Microsoft Word to apply AI-generated redlines to first- and third- party contracts.
  • Paxton is handy for not just contract review, but all the more so for tasks around legal research, including statutes, case law, and more.
  • Juro was first and foremost a Contract Lifecycle Management tool for managing version control, execution, and post-execution organization, and now also includes the ability to insert AI-generated language into contracts.

Best Practices for Using AI for Legal Contracts

Beyond the legal-specific best practices described in using AI, consider the more general one:  Giving your AI agent lots of context including guidance from Silicon Valley-specific resources; spend time prompting your LLM to test out its responses for future references, continually review and refine based on its answers!

Wrapping Up 

AI tools can definitely be leveraged for reducing legal spend and the time it takes you to work with a lawyer.  Like any powerful tool, though, it should be used carefully; in this case, by knowing when to use it (especially if you are budget sensitive or have scoped out a clear application together with a lawyer for a highly recurring tasks), and especially for workstreams that are less technical, sensitive, or context-dependent.

Again, keep in mind that you should only rely on AI for contract review if you have no other choice, or if you can do it in conjunction with a lawyer; humans and AI both make mistakes, but you can much better vet a human’s work product through various signals than you can vet the work product of AI.

Ready to learn more?

Enter your work email to get started

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.