John Craske on CMS’s Agile GenAI Strategy After the Harvey Deal

As of last week, international legal group CMS has now started to use the ‘£1,000 per person, per year’ legal genAI system Harvey on paid client matters. But what is the wider AI strategy at CMS? Artificial Lawyer caught up with John Craske, Director of Innovation, to find out.

First, Harvey, the legal AI company that seemed to come from nowhere 18 months ago to now figure in the tech stack of many of the world’s largest law firms, has OpenAI as an investor and is built upon its GPT LLM series.

Craske explained that they started to work on a wide-ranging pilot back in November 2023 with around 100 people in the UK getting to use Harvey for tasks such as: reviewing documents, drafting clauses, and summarising case reports. They have identified around 90 use cases in total. He added that ‘Harvey is not going directly to the DMS’ of CMS, which naturally narrows some of the tasks it can do for them.

Now begins the real thing, when Harvey will be used as part of billable work for clients. For this they’ll be using the legal AI system ‘as is’. As Craske noted, they’re using a ‘vanilla’ version without any special fine-tuning for CMS itself.

They will have licences in the ‘low hundreds’ and have ‘found that lawyers saved more than one hour of time for each task’ undertaken. He added that, as with nearly every other law firm and inhouse team, they will not be putting all their AI eggs in one basket and will keep working with other tools they have already.

Or as Craske put it: ‘It’s too soon to make one big bet.’ He added that innovation teams ‘will have to be quite agile because the pace of change is ever increasing’.

‘[GenAI] is now more experimental. We play with products, understand them and then move on. We couldn’t write a genAI playbook now, as it would be wrong in a couple of days.’

He added that overall, ‘I think Harvey is good for the market because it inspires competition.’

In short, don’t believe that Harvey will swallow the entire legal tech market anytime soon, even if it has made rapid inroads very quickly, at least among some major law firms and inhouse teams. And it has to be said there is a lot of competition, both from established companies and new ones that are rapidly emerging.


Craske would not comment on Harvey’s cost, but from sources elsewhere in the legal market, this site has heard that at least some firms have been quoted pricing that amounts to about ‘$1,200 per seat, per year for Harvey’, or about £1,000 per seat.

It’s also understood that Harvey’s sales team have at least on one occasion demanded that 100 seats is the minimum number of sign-ups, with one year as the minimum term. So that would cost at least $120,000 for a 100 lawyers, and you can extrapolate easily from there.

Several global law firms have over 2,000 lawyers and then more than that again in support staff. So, a roll-out for just the fee earners would be around $2.4 million for some of the largest law firms. Although one could argue that for a law firm with $1bn-plus revenue even a couple of million is not a massive amount, especially when you look at spending on essentials such as Microsoft 365 for everyone in the business.

In comparison, Microsoft’s Co-Pilot is about $30 per user per month, hence $360 per year per lawyer, but according to Craske ‘it is not fine-tuned to lawyers….and [produces results] that are shorter than a lawyer would want’, i.e. not that useful on its own for the legal sector.

Also, when considering Harvey’s price, this rules it out for most small law firms, which might refuse to add such extra costs to their tech stack. And it’s worth remembering that at the bottom of the UK 100 law firm table, revenue levels are around £30m per firm, in fact it’s not until you approach the Top 50 firms by revenue that you get to £100m and firms with considerable resources for tech projects.

In which case, smaller firms are going to have to find other ways to engage with genAI tools.

Artificial Lawyer put these numbers to Harvey and asked for an official comment, but they didn’t respond.

The Bigger Picture

OK, so that’s Harvey. Now what? What is the bigger picture? For example, has the firm’s management considered what an expansion of the use of legal AI tools, and other legal tech products that drive efficiency via automation, will mean in terms of the economics of the business?

‘My view is that law firms are at the beginning of the digital transformation journey,’ Craske explained. ‘Before, computers were just glorified typewriters [for lawyers].’

I.e. we’ve only really just started to see what we can do yet.

So, if we transform how lawyers work, what then happens with the time-based legal economics upon which the entire structure of law firms is based?

‘We have debated the economics. There have been discussions at board level and among practice groups. The junior end has been the engine room for generating revenue. If things are more efficient, then by the nature of it then there are less hours [for an element of a matter]. And if you follow that model then there will be less revenue,’ Craske observed.

He added that clients will also ask about AI’s impact with regard to their legal spend. And clearly, no firm wants less total revenue. So, how to navigate this?

‘I think there will be a more nuanced conversation about value,’ he added, but noted that they do not yet have a playbook for pricing with AI as part of the equation.

There will be experimentation also now with pricing. There will be an openness with clients, so there is trust and they can see the value,’ he suggests, then underlines, ‘the real story here is the people and the business model. This is about human-plus-machine. At the moment the focus is on the machine, but the ‘human-plus’ is the most interesting part.’

And this is a key point, which Artificial Lawyer totally agrees with. The real-world outcomes that stem from the use of AI, or any other tech, are way more important to the clients and the law firms than the tech itself. The tech is always a means to an end – albeit a fascinating area of tech that as a sector we’ve only just started to leverage on client work.  

But, the question then is: what is the goal? (See earlier article: ‘GenAI Has Split the Legal Information Atom – Now What?’)

The reality is that CMS – like most other firms – is still feeling its way forwards. It’s identified use cases for genAI, it’s onboarded a chunky system like Harvey, and it’s exploring other avenues as well. It’s talking through pricing in an AI world, even at board level, but there is not yet a systematic way to price with AI. In short, they are in the same position as many other leading law firms around the world.

The main thing is that they are doing something and learning. They are getting into the details and the outcomes of legal AI through real world testing and that’s got to be a good thing. Plus, as Craske stated, they are staying agile on genAI, this new phase of legal tech is only just getting started.