A coffee with…Jason Hill, UK CEO, Reply

Starting out as a tech support worker, Jason Hill  has accumulated 30 years’ experience in the IT and consulting industry, with the last 15 spent as a tech leader at Reply, a global network of companies specialising in enterprise-based business solutions.

TechInformed met Hill in June, at the London leg of Reply Xchange, an annual multi-territory set of events that bring together IT professionals, creative thinkers, and tech enthusiasts to explore the role of technology in reshaping industries.

Among the Reply clients present at this year’s event were HSBC, Schroders, easyJet, UK Ministry of Defence and car maker Aston Martin.

Gen AI dominated this year’s ReplyXchange. In terms of your clients, how many are at the PoC stage with this technology, and how many are using it on live projects?

We’re at the end of the beginning with Gen AI – it’s no longer considered something scary.

While not every customer is doing something, if you saw the numbers on my chart so far this year, just for the UK alone we’ve had 500 requests for pure AI. We will do around 5,000 projects this year, so even if 10% of these were pure AI and we didn’t take any more requests… it’s still a significant amount. I think by end of the year it will be around 25% pure AI projects, while it will influence others.

Where do you see Gen AI use cases dominating?

One of our customers is applying AI across their contracts which has led to savings in tens of millions.

When you look at the characteristics of an AI project the first involves consolidating data into one place; the second is then putting a model on top of that to understands what this data means. So, the customers that have done the first piece and have applied data on top of a large data set, they are the ones that a really going hard on this.

With sectors such as banking and finance, there seems to be a friction between governance and regulation when it comes to AI….

What we must look at is how to better understand what some of these things do using AI, and if we can understand them, then we can regulate them. If we regulate them, we have compliance and if we have compliance then we can have resilience.

While there is some trepidation in financial services, we are certainly seeing some customers in this area look at AI ‘reg tech’ and how they can understand really what’s happening.

One of the regulations, the BCBS 329, for instance, focusses on data lineage – understanding where the data has come from – that’s a great use case for AI because its quite a manual process – things hop from different systems or go through black box algorithms that we don’t understand. So, we have some pilots involved with how we can apply AI to understand data lineage.

In terms of use cases, which sectors are embracing Gen AI?

Certainly, Customer Service and Sales. They are the two largest because they are the most obvious use cases.

When we look at CX, one of the challenges is that human capital is expensive. But that’s not the real problem – it’s that we offshored it , which had some success in reducing cost but then involved producing scripts and so what we did was take the intelligence out of CX because its expensive in the form of people.

What AI and LLMs allow us to do is put that intelligence back in so we can take the drudgery out of CX, but more importantly, because we can hook up some of the data that the agent wouldn’t have acquired before they received a blind call, they are able to help customers more.

And then with sales. One of our customers in automotive – I did hear a stat on our dashboard – that the drop out rate of car purchases has fallen by 35%. That doesn’t mean they are selling 35% more cars, but 35% more journeys are finishing end-to-end with Gen AI whereas, before some customers would drop out along the way, so it is reducing friction.

The Aston Martin case study which we saw there today, is focussed on the AI propensity model and how Gen AI is helping it target which deals in its pipeline are likely to convert.

Shield Reply is also doing some interesting work in defence, generating war scenarios…

Yes, that project is focussed on the digitisation of military scenarios for operational use/ practice. Some are done in the field; some are done in the classroom. Classic scenarios include rescuing hostages, defending airspaces and setting up exclusion zones.

As a training leader you typically need to change the scenarios that people play, and the planning cycle takes more time than the playing time  – so we’re trying to flip this on its head.

Each model will learn from itself but also the planners are still able to build their own models because there’s  a lot of data and a lot of variables.

Do you see areas where AI is going to create new job roles?

We’re moving from governance roles to stewardship roles with AI. There’s a subtle difference. With governance you are trying to control a process or regulation but for stewardship roles you are thinking more about taking care of something and shepherding projects through various processes. How will these things work? And when they do work, who is taking care of it? And who is responsible for driving the right motivations?

Stewardship roles are not classic IT management either. It’s taking a slightly different approach of how you want to manage something. It’s an interesting set of skills that people will need to develop. But they are very human skills.

Would these ‘AI stewards’ necessarily have to come from a tech background?

It depends on the context. In the call centre someone’s got to manage the way someone responds to text, emails and IBRs. That level of curation doesn’t need a tech background. But under the hood if you are looking at cloud and data and LLMs you’ve got to be techy. The idea that AI is simple isn’t true. All we really do with technology is abstract away, so it becomes more usable for more people. But we’ve still got all the complexity underneath. But things are getting more technical rather than less technical.

If you have a conversational interface –  fine –  but how do we check what it’s doing and whether it integrates and shares and is compliant in the right way?

What do you do to switch off from work?

I’ve two sons and right now I’ve been helping them with exam revision. I’ve been doing GCSE computer science and A Level Business Studies and Chemistry.

Don’t all the kids use ChatGPT to help with homework now?

We’re not encouraging it. The school where they go to has an AI test so that they can test whether it was written by an AI or not.

What else do you do to relax?

Run, walk the dog, do the housework and cook. I’ve also taken up yoga recently – that’s probably the only thing I do where I  have to switch off for half an hour without thinking about something popping into my head – mainly because I’m not very good at it – I have to concentrate, or I worry I might break something!