Andrew Powell is CIO of UK top 50 law firm Macfarlanes, which is one of the early adopters of generative AI tool Harvey. We asked how law firms are approaching the adoption of generative AI tools and whether IT budgets are likely to increase.
How would you summarise 2023?
2023 has been relentless at one level and going back to a point that Rob Archer, a psychologist from Cognacity, made at ILTA Europe, whilst we plan the workload to have peaks and troughs, this year we have seen a few unexpected projects arrive and occupy the troughs. That’s fine short term as long as it doesn’t become the norm. For the sector though, the big thing has been the arrival of generative AI and the continued move to SaaS, and the number of projects covering one or both of these.
What have been the key surprises and challenges?
Useable generative AI has come round far faster than many people expected. Looking back a few years at the cycle times for Kira or Luminance, they were a fair bit longer and I think the speed to market of gen AI tools may have caught people by surprise. If you look at Copilot, there are some challenges for law firms: documents are usually in iManage or NetDocuments, not SharePoint; information barriers are controlled by a separate system, not the file security on the document store; the code version and patching cadence for Outlook don’t suit most firms who are using Outlook add-ins (though that is changing next year); and the 300 minimum user limit for Copilot (on top of an E3 or E5 subscription) is unpopular – for many firms that’s a firmwide rollout, not a pilot.
Do you think law firms are still reticent to adopt gen AI products?
No! We’ve gone ahead with Harvey but we’re looking at other stuff as well. What Harvey offers is inherently different to Copilot, for example. Whether it is a license cost or the resource costs of building it yourself, these tools are not cheap and you have to look carefully at what the use case might be and what that will give to the firm. If you’re going to spend six figures on Copilot, what are you going to do differently to justify that? Some of those benefits you can articulate well and some are harder to put a value against. Is it helpful to summarise a Teams meeting? It depends on who you are, who was there, and what you’re doing with the output. If it was free, you’d be all over it, but there’s an opportunity cost and we have to think about what else that investment might have enabled. I’m surprised extractive AI has not had the same attention this year (“go and find me the last three times we did something like this”) but maybe that will happen next year.
Will IT budgets go up?
Probably, yes. Gen AI tools might drive some of that but in most cases, headcount increases and subscription costs going up by RPI (or more) will push up the cost of providing a base service to a firm. The other thing with SaaS subscriptions is that you can’t delay costs for a year: in the old world, you could have delayed an upgrade for a year and pushed the project and license costs out a bit; whereas now if you’re in the cloud you’ll have the same subscription costs next year if you’re lucky, or a big increase if you’re not. Those of us renewing a Microsoft E3 or E5 will see a big uplift on the pricing from three years ago.
So, the other theme was the continued move to SaaS?
Yes, that came out loud in ILTA’s annual technology survey and one I do myself every year – both surveys show the same direction of travel. Many firms have a SaaS strategy – and why would you not with the complexity of DMS and PMS. Firms have tried to simplify the estate they’ve got, so not having to support complex on-premises stuff is attractive, though the migration often is not. You also have a layer of cybersecurity that’s increasingly complex for on premises systems, with increasing pressure to patch regularly and respond very quickly to vulnerabilities. It’s also pretty clear that most vendors are focusing their functionality improvements in their SaaS offerings first.