I know a lot of law firms are starting to worry that they might be falling behind on AI and thinking they need to rush now to embed it throughout their practice.
I think right now would be a good time to pause and reflect on where things are going before making any big, longer-term decisions.
Different approaches
Sure, there have been announcements of the world's largest law firm, Kirkland and Ellis, investing $US500 million in building its own proprietary AI system. That's a huge bet that they can build something distinctive enough to add value over what everybody else will get by subscribing to constantly improving commercial offerings from other parties, such as Legora and Harvey. It might work out; it might not. In any event, they're going to be on the hook for substantial ongoing costs to maintain and improve whatever they build to outpace the market. Still, for a firm of that size, it's manageable – probably only 1% of turnover.
At the other end of the market, there are regular news stories about AI-native law firms being set up, such as Superlegal in the U.S., which is offering construction contract reviews signed off by a lawyer for only $117. It's hard to comment on the quality of that service, the value they add, and the moat that protects them from other firms offering a similar service.
I would assume that there's not actually that much of that work around to support lots of firms offering that type of service at that price point on a profitable basis. As those capabilities become widespread, many clients can get their own AI review solution if it’s not doing much that's special.
Even if it is possible to pump very high volumes of that work through in that way, that's not the type of work that I would find satisfying and would want to be doing as a lawyer anyway. I’d feel like a hamster on a wheel and would burn out or become bored very quickly.
I would tend to look at that model as being a floor, which could be the base on which to start figuring out other things I could do that would be uniquely valuable and which clients would be happy to pay more for to build a new practice around, that was satisfying and allowed me to spend more time on things I enjoyed, like thinking and talking to people.
What technology to choose?
There's a further question of which technology stack do you use to build these new capabilities?
This year, the world has fallen in love with Claude, particularly after it released a bunch of new legal AI capabilities. I've started testing them and have found them interesting and can see the potential. However, at this stage, it also requires quite a lot of work to figure out and set up. On the Teams plan I subscribe to, it's quite common for me to run out of credits midway through even basic tasks and have to wait 5 hours to finish.
If I were going to use this as a core part of my business, I would need to increase my plan and costs significantly. In a larger business, I’d need to do that for everyone. We're already starting to hear stories of organisations that are realising AI is going to be very expensive once the true costs of tokens are passed through. This is because many of the workflows currently running are not targeted at areas where AI is actually required and adds the most value, and are not necessarily optimised to make the most efficient use of the allowed tokens either.
All organisations, including law firms, are going to need to take a more holistic look at their processes and the outcomes that they need to achieve, and then design an approach that uses AI where it genuinely adds a lot of value but also makes more use of other technology that they already have, which doesn't cost nearly as much. Good examples include tools in the Microsoft subscription, such as Power Automate. Another tool that we use a lot, which is very low-cost, is Cognito Forms. These tools can handle very high transaction volumes without additional costs.
The other point I would note is that we're still waiting to see which organisations are going to be the best core to work around. All the platforms are rushing to embed AI across their platforms, hoping to become the single source of truth that everybody else has to connect to.
Most recently, I've seen some interesting examples of what iManage and NetDocuments are doing in the document management space. In theory, it makes a great deal of sense that your document management system would be the place where all of your emails and documents sit, and AI can form a matter graph that is constantly updating every time a document is added or updated, so that you can immediately log in and see what is most important.
On the other hand, as AI becomes increasingly expensive, do we really need every single document across every single matter to be constantly reviewed and updated by AI? Just in case someone wants to log in and get a real-time update on what someone in a junior role has been editing as work in progress.
If it's not just the document management system trying to provide that level of AI and insight, are we actually paying multiple times for the same thing, whether we want it or not? Are the practice management systems going to be trying to do the same thing?
As a simpler example, I now have two AI systems that want to brief me in the morning on the content of my emails and calendar and tell me what is most important for me to do today. I also look at my own emails and calendar anyway because that's just what I like to do.
Things will be clearer soon.
I think in six months it will be much clearer where things are headed, and it will be possible to make more informed, better choices.
It should be more obvious by that time what the AI capabilities of document management systems and practice management systems will be, how they will fit together, and who is doing it best. Let those companies invest their own time and money in figuring out how to orchestrate all the various workflows and optimise the AI agents so they do what they should be doing efficiently.
You'll also be able to see then how much you can still do with the basic merge capabilities within your practice management system to generate simple documents, just as you currently do - consistently every time. If needed, you can supplement a specialist document automation system for more complex documents, which, through integration, can still draw on existing matter information. It provides the safety of deterministic logic without the risk of hallucinations or the high costs of AI.
You can then use AI for the distinctive parts where it will add the most value.
What are we going to do in the meantime?
You're probably asking: how are we going to prepare ourselves for this new AI future, and what should we be doing now?
I'm sure there is a great deal you can do over the next six months that doesn't require you to make three-year technology commitments.
The most important things I would be thinking about are:
- What types of work do you think that a law firm like yours should and will be doing in three years?
- Do you see yourselves doing $117 construction contract reviews, or would you like to be doing something else that's more valuable?
- If you are going to do something else that's more valuable, what skills will you need that you don't currently have? You could start preparing to learn those skills or find partners who already have them to collaborate with.
- You should also assume that hourly billing is not going to make much sense in three years, because one way or another, there will be a lot of automation and efficiency throughout the systems that you work with. Everybody will have access to it because they will use the capabilities of at least one of the systems they subscribe to. It will be well worth spending time now to think about what value looks like, how you can create it, how you can communicate it, and how you can fairly share in it when the value is not based on your inputs and the amount of time it took you, but it's on the outcomes that you can achieve for your clients. Working with a value pricing expert now would pay significant dividends down the line and will most likely help you understand what you want from your technology and how to choose the one that is best suited to you.
As you can tell, I'm not holding myself out as an expert on AI or any particular platform. I do think it's worth spending time exploring these tools, seeing what they can currently do, and who seems to be doing it best.
If you're planning on building your own AI workflows using Claude Co-work or something similar to try to embed your own distinctive AI throughout your firm in a way similar to Kirkland and Ellis, I think you'll likely be wasting time and money that could be better spent on more strategic questions about the nature of your business and business model. If those are things you're interested in, that might be the kind of thing I could help you with. I'm always up for a chat if you send me an email. I'm pretty sure one of my AI agents will let me know, and we can schedule a time.
