Digital Catapult’s AI journey: from garage to national infrastructure

When the final cohort of the Machine Intelligence Garage completed the programme in 2023, Digital Catapult faced an unusual problem: the initiative had worked too well. 

Digital Catapult had spotted a gap when designing the Machine Intelligence Garage years earlier: the increasing cost of computation power for AI and machine learning models was a significant barrier despite rapidly increasing availability and decreasing costs of cloud computing. This apparent paradox was proving to be a big obstacle to the success of high-growth potential startups as costs quickly accumulated before technology proof-of-concept, funding and revenue can be achieved.  

Over the six years since it was conceived, Machine Intelligence Garage had become one of the UK’s most respected AI accelerators. It won multiple awards and supported entrepreneurial founders to turn cutting-edge artificial intelligence (AI) and machine learning (ML) based solutions into viable, investable products and services. By providing that much needed – free – access to the compute power required to develop machine learning models (£7.2 million worth of compute in total), Machine Intelligence Garage intervened to actively address, and fix, this problem giving startups on the programme a clear competitive advantage. It helped over 150 deep tech startups across 17 different cohorts to unlock £52 million in investment, resulting in an average of 2.5 jobs created per company.  

But over the course of the programme, the UK AI landscape had begun to change. Compute became less costly and more widely accessible to smaller businesses. AI capability started to spread beyond disruptive startups into more traditional sectors, and the biggest challenge was no longer just building AI, it was deploying it practically, usefully and at scale.  

As Machine Intelligence Garage concluded, Digital Catapult chose to take what had worked so well about that programme, re-engineer it and do something different.

Machine Intelligence Garage demonstrated three things very clearly.  

First, early-stage AI companies thrive when technical support and commercial reality are treated as inseparable. Founders needed cloud credits and hardware support, but they also needed access and connections, with peers, mentors, investors, and industry who understood the complexity of ML-driven products and services and were ready to be early adopters.  

Second, ethical AI could be practical, not theoretical. Digital Catapult embedded responsible AI practices directly into product development, supported by an independent AI Ethics Committee and a clear, usable Ethics framework. The focus on deployable, responsible AI was treated as a core part of building a better, more investable business and not simply a compliance exercise. 

Third, diversity strengthened innovation. The 17 cohorts of deep tech startups that joined Machine Intelligence Garage consistently showed higher-than-average diversity across ethnicity, background and experience, broadening both the perspectives and the applications of AI that were being developed during the programme. 

By its conclusion, alumni companies including Greyparrot, Carbon Re, ToffeeAM, Audiostack and LuffyAI, had moved far beyond the pilot stage into real-world deployment, tackling various critical industry challenges, and hiring staff as they grew and successfully scaled. Together, these five companies have collectively raised £19.5 million in private investments and almost £1 million in public investment within two years of engaging with Digital Catapult. Greyparrot’s analyser tool was even listed as one the Best Inventions of 2025 by Time Magazine. These impactful examples of how companies scale after working with Digital Catapult represent a handful of the hundreds of small businesses that have gone through our deep tech acceleration programmes.  

With graduate companies going on to achieve success and gain private investment, the question for Digital Catapult became: how to take this success beyond one single, albeit very successful, accelerator programme?

The answer was not another “garage”, but a shift from accelerator operator to ecosystem builder. Digital Catapult absorbed the lessons of Machine Intelligence Garage, and multiple other technology and sector specific accelerators from years past, into a broader innovation framework specifically designed to support deep tech companies at multiple stages in their growth journeys, working with different technologies (including AI and machine learning) across different sectors. This new approach enabled various pathways to success for deep tech startups, from early experimentation and development, right through to investment readiness and scaling.

This evolved approach paid off and since 2020 Digital Catapult has: 

• Supported almost 3,000 businesses 

• Helped these companies raise a combined £596 million in private investment 

• Generated an estimated £4 billion in wider economic impact for the UK 

AI and ML were no longer a standalone vertical. They were becoming foundational capabilities across Digital Catapult’s work. 

The next significant chapter in this story came with the launch of BridgeAI, a national programme funded by Innovate UK and delivered by Digital Catapult, the Alan Turing Institute, STFC Hartree Centre and BSI. BridgeAI was built on a simple insight drawn directly from Machine Intelligence Garage: “The barrier to AI adoption is rarely a lack of algorithms, it’s a lack of connection.”  

Many sectors critical to the UK economy have huge potential for the practical application of AI, in all sorts of scenarios. From construction to agriculture, transport and logistics, and the creative industries, all are rich in data but poor in AI capability. Meanwhile, AI companies struggle to access real industry players and their problems, representative data, and routes to market. As the name implies, BridgeAI set out to bridge those gaps. 

Drawing on Machine Intelligence Garage’s playbook, Digital Catapult designed the High Growth AI Accelerator, a short and intensive innovation programme focused on four key components: 

• Pairing deep tech AI SMEs with real sector challenges set by leading industry partners 

• Providing hands on technical and commercial support 

• Embedding responsible AI practices from day one 

• Preparing companies for practical adoption, not just demonstration.  

This marked a subtle but important change in emphasis. Success was no longer measured only in funding raised, but in deployments achieved, productivity unlocked, and confidence built within entire sectors. Industry partners jumped onboard – from Nestle to Transport for London (TfL) to Merlin Entertainments – and since its inception in the summer of 2023, Digital Catapult has welcomed 24 startups to the programme, supported the development of 22 proofs-of-concept and minimum-viable products, and enabled participating startups to secure £1.7 million in funding. The AI Adoption Toolkit, developed exclusively for this programme, gave organisations a unique opportunity to assess and benchmark their AI adoption readiness, and has been used by over 100 companies since it launched.

The tangible impact of this evolution in Digital Catapult’s interventions shows up in multiple places: 

• AI companies supported by Digital Catapult are now deploying solutions into live industrial environments, not just running pilots. 

• Traditional sectors are gaining the confidence, skills and frameworks needed to adopt AI responsibly, reducing risk both demand side and supply side. 

• Ethical AI practices developed during Digital Catapult’s early AI interventions are standardised and scaled, influencing dozens of companies over the long term rather than a single cohort. 

• The UK AI ecosystem has shifted from isolated innovation to collaborative problem-solving, with Digital Catapult widely regarded as a trusted and significant player. 

Companies emerging from these programmes are tackling challenges ranging from net-zero industrial processes to automation in waste management and data-driven decision-making in complex supply chains. Even where challenges are faced in getting enough data to develop a tangible solution the programme steps in to help both startup and industry challenge owner to overcome the problem.  

One startup on Digital Catapult’s High Growth Accelerator was developing an AI-driven multi-modal model to categorise a large media publisher’s content archive. It was given access to a partial dataset that was large enough to build a Proof of Concept (PoC) and once the PoC was developed the publisher shared more data for model refinement and operationalising the solution. This modification approach allows an adopter to maintain dataset control and privacy when piloting and collaborating with a startup before scaling up a system for launch. The startup was also able to identify new use cases for its model, such as helping filmmakers categorise hours of footage and streamline editing processes, a good example of how an AI model trained for one use case can open new opportunities in others. 

What’s more, AI is frequently harnessed in tandem with other deep technologies as part of the convergence of the deep tech “stack”.  

For example, as part of Digital Catapult’s work on the MyWorld programme in the West of England, small and medium-sized enterprises (SMEs) were supported to understand how to leverage AI responsibly within the creative industries for prototyping and design through a series of workshops that upskilled teams to analyse and evaluate solutions for potential unethical behaviour risks.  

AI-native networks and architectures where AI becomes a core component of network management and not simply an add-on for managing increasingly complex 5G-Advanced and 6G networks, is a particular area of interest for Digital Catapult. AI has enormous transformative potential in telecoms offering opportunities ranging from network optimisation and enhanced security to personalised services and intelligent customer care. REASON, a programme led by the University of Bristol in collaboration with Digital Catapult, is one example of work that is looking at how AI can be integrated into the development, management and automation of future 6G telecommunications networks. There are a multitude of challenges to integrating AI into telecoms networks and technical, operational and ethical concerns needs to be addressed for successful implementation, but the opportunities are almost limitless. 

The end of Machine Intelligence Garage was a transition in Digital Catapult’s approach to enabling the practical application of artificial intelligence and machine learning. It proved that the UK could nurture world-class, high-growth potential AI startups, and what followed proved that the same principles of deep tech support, ethical foundations, and strong connections to real end users and industry challenges, could be applied at a national scale. 

Today, Digital Catapult’s AI and machine learning work is less about a single programme and more about solidifying the UK’s digital infrastructure: the people, partnerships, and market pathways that allow AI to move from code to consequence. It’s also focused on supporting a wide range of businesses and organisations across critical sectors of the UK economy to understand how AI tools can deliver practical, tangible benefits.  

The garage doors may have closed, but what they opened continues to shape how AI is built, adopted, and trusted across the UK.