We’ve identified and named a new problem in the world of digital design and development – The AI Governance Gap: the distance between how fast a product team can now build software with AI and how slowly the policies around them have caught up. Our research, run with Intelligent People and supported by CrowdPull PR, found four times as many product people have built working software with AI as feel confident putting it into production. This gap is the subject of our new report – The AI Governance Gap Study.
Product owners are no longer just defining software. They’re building it. We started seeing this first-hand in our own client work, so we went directly to product professionals to find out how widespread it is. The result is the AI Product Builder Study, run in partnership with Intelligent People, a specialist product, design and UX recruitment consultancy with access to a database of 70,000 professionals. We received 2,251 responses from UK product, design and leadership roles. Since then, we’ve turned that data into the AI Governance Gap study that you can read below.
That partnership came about thanks to our long-time partner and collaborator, Liam O’Neill at CrowdPull, who introduced Bulb Studios and Intelligent People and made this whole study possible.
What The AI Governance Gap Study found
- 62% have used AI to build a working application or product feature, with a further 9% currently reviewing or experimenting.
- 27% have shipped AI-generated software without engineering review.
- 31% do not know who legally owns AI-generated code.
- 15% feel confident deploying AI-generated software into production. A further 30% say they feel confident enough.
- 1 in 3 say they could build an MVP without developers.
- Around half believe coding will become a core product management skill within five years.
- 29% believe AI will replace prototyping within two years, while 61% believe it will assist but not replace it.
In short, the capability arrived first. Confidence, process and accountability haven’t followed yet. That’s what the AI Governance Gap measures.
Speed is not the problem. Ownership is.
This is where the AI Governance Gap hits hardest. Most operating models in use today were built for a world where code could only originate in one place. Access control, code review, security testing, licensing checks and release approval all assume an engineering team sits between an idea and production. However, when a product manager can build and deploy in an afternoon, that assumption disappears and nothing has replaced it yet.
The legal position is the least settled part of the picture. Ownership of AI-generated code, the licensing obligations inherited from a model’s training data and liability when generated code fails a security or accessibility audit are all areas where UK organisations are operating ahead of their own policies.
What The AI Governance Gap Study means for building your team
AI product-building is becoming a career advantage. It’s widening the AI Governance Gap at the same time. Employers increasingly expect product candidates to demonstrate AI prototyping skills. The ability to build is starting to shape who gets hired and promoted. However, what it doesn’t yet do is tell an employer whether that person knows when not to ship.
Every organisation will close its AI Governance Gap eventually. The only real choice is whether it’s closed deliberately, or by an incident. In fact, the full study covers the data, the methodology and the questions above in more depth than we can fit into one post. Download the 2026 report for free – no signup or data requested, below:
Frequently asked questions
It's the distance between how fast product teams can now build software with AI and how slowly the policies, ownership and review processes around them have kept up. Our research found four times as many product people build with AI as feel confident shipping it.
62% of respondents to the AI Product Builder Study have used AI to build a working application or product feature. In addition, a further 9% are currently reviewing or experimenting.
It depends on the contract and the tools used. 31% of product people surveyed didn't know who owned code they'd already written. Establishing ownership before code reaches production, not after, is one of the study's core recommendations.
27% of product people surveyed have already shipped AI-generated software without engineering review. Because of that, the study recommends giving AI-generated code the same security, accessibility and data-protection review as any other code before it reaches production.
The full report is available as a free PDF download, covering the complete findings, methodology and recommendations in more depth than this summary.
Bulb Studios ran the study with Intelligent People, a specialist product, design and UX recruitment consultancy. Since Intelligent People holds a database of 70,000 contacts, 2,251 UK product, design and leadership professionals were able to respond.
