Why coding faster isn’t making deployment faster

James Jarvis has spent 25 years in revenue roles, including stints at GlaxoSmithKline and Forrester Research and a spell running a Symantec sales team. Along the way, he founded two companies of his own: a lead-generation agency and, with his wife, a shaver brand.

He later helped grow digital transformation consultancy ECS, which was acquired by GlobalLogic in 2020, before joining ClearRoute — first as a non-executive director, then as chief revenue officer, then stepping in as CEO in March.

With TechInformed, he talks about building ClearRoute’s own AI platform, the gap between coding faster and shipping faster, the cultural fallout of AI adoption and how he uses AI himself day to day.

How has it been in the CEO seat these past few months?

Wild, in a good way. What an amazing time to be leading a software engineering company. We started investing heavily in AI early last year, on the basis that if we’re going to advise customers on AI strategy, we should build something ourselves first. Led by our global head of engineering & AI, Justin Wilkin, we built an MVP [minimum viable product] platform on primitives we felt were foundational to getting enterprise value from AI.

And in January we decided to become an AI-native company. We gave everyone at ClearRoute, technical and non-technical, a Claude license to build internal skills and agents, and now have agents running parts of HR, talent acquisition, finance, ops and sales, all connected into Orbit, our internal agentic platform, with governance, access policies and a token-cost model built in so we know exactly what we’re spending and where. We’ve also got over 50 Claude architects, effectively “forward-deployed engineers,” embedded with clients to identify workflows and turn them into governed, scalable capabilities rather than one-off builds sitting on someone’s laptop.

Your State of the Route to Live 2026 report found that AI is speeding up development, but not getting software live faster. Can you unpack that?

AI is making developers code faster than ever, but that was never really the problem. The real problem is: can we get well-understood, well-built features actually live? Because it’s only when something’s live that anyone gets value from it, otherwise it’s just a sunk cost.

A newer firm can release an idea within hours or days; larger, more traditional organizations take months. So writing more software faster just compounds the existing bottleneck (governance, security, release controls, quality gates) rather than solving it.

AI amplifies whatever state you’re already in: if your engineering practices are strong, you go faster and get real advantage; if you’re cumbersome, it just amplifies the problem.

What are the big problems you’re seeing enterprise leaders wrestle with right now?

A few things, from the five or six C-level conversations I have every week with large, regulated customers.

First, identifying and prioritizing the right use cases, and knowing how to measure the value you’ll get.

Second, working out when you can be confident the value will justify the investment — ideally the return should pay for itself many times over.

Third, scaling AI work from pilots into genuinely operational capabilities across the wider business.

And finally, the human and cultural impact: if you free up someone’s time, like a stock manager no longer needing to manually forecast fruit and [vegetable] orders, what do they do with that time, and how do you help them adapt? Multiply that across an organization trying to save hundreds of millions a year, and it’s a huge conversation about culture and ways of working.

How do you use AI day to day, personally?

I’ve got an agent that goes into Salesforce for me, pulls reports and flags key pipeline changes so I barely need to log in myself. On calls, an agent transcribes, logs the activity in Salesforce and drafts a follow-up email in my tone that I can send or tweak. All our board reporting is AI-enabled now too, pulling from MCP [Model Context Protocol]–connected systems into a branded pack with analysis and recommendations we then validate and adapt.

And in talent acquisition, agents now screen the 100-plus CVs we get per role against our ideal profile and produce a weekly report on interviews, applications and conversion by channel — freeing up a tiny team to spend their time where it matters.

How do you take your coffee?

I have a flat white, extra hot.