Responsible AI adoption gets easier when an enterprise brings a real problem and a startup builds the answer with it. That is the model Jimmy Jarvis, Lead for Policy and Engagement, AI, at Digital Catapult, set out at aiblLIVE London 2026.
Digital Catapult runs challenge-led accelerators that pair AI startups with industry partners. Three of them show how a pilot becomes something people use, from Merlin Entertainments to Transport for London.
All figures as stated on stage by Jimmy Jarvis.
Most firms use AI somewhere, but few have embedded it in live operations. ONS data puts business use at 46% and extensive use at around 12%. Staff use runs ahead of sanctioned use, which creates ungoverned shadow AI. Productivity gains often fail to reach the bottom line, and most barriers are not technical.
When we talk about AI adoption, we’re referring to deep, extensive adoption. This is AI in live operations, at scale and governed with the skills, data, and governance to sustain it.
Jimmy Jarvis, Lead for Policy and Engagement, AI, Digital Catapult
An industry partner sets a real problem, and startups compete to solve it. Digital Catapult scopes the challenge for AI suitability, selects a cohort and provides technical and commercial support, from compute to coaching. Responsible AI practice is built in from day one, and the adopter is prepared for deployment rather than a demo.
Merlin got a bespoke product shaped around its own data. Placemaking AI adapted its visitor insight platform for leisure, then built virtual guests trained on Merlin’s reviews and social content. Staff can test how a new attraction might perform. A separate startup, Tell Me AI, trialled a museum guide grounded in the attraction’s own documents.
Merlin didn’t buy a finished product from Placemaking. They collaborated with the startup and shaped an entirely new product and service around its own operations and data.
Jimmy Jarvis, Lead for Policy and Engagement, AI, Digital Catapult
Digital Catapult’s accelerators point to four things: a real problem, data shared in stages, end users involved and effort planned. A small data sample comes first, with wider data later. Both sides need the buy-in to push through to a minimum viable product and a commercial deal.
Jimmy Jarvis is Lead for Policy and Engagement, AI, at Digital Catapult. The UK deep tech innovation organisation has supported more than 3,000 companies since 2018. He spoke on the Growth track at aiblLIVE London on 20 October 2026.
If your AI pilots stall before they reach live use, this model is worth a look. The full PDF covers all three partnerships in detail, in the speaker’s own words.
For innovation leads, COOs, digital and AI leads and startup founders. Six pages, about an eight-minute read.
