As part of our ongoing AI in Software Engineering Insights Report, HI Technology & Innovation has been speaking with engineering leaders across industries to understand how teams are truly using AI.
At our recent breakfast event in London, founder Mike Daniel shared some of the early findings — and the patterns already emerging are both exciting and cautionary.
A Room Full of Ideas
Mike opened the morning by sharing his journey from software engineer to founder, and the evolution of HI Technology & Innovation into a growing team of more than 30 specialists. He set the stage for a frank and forward-looking discussion on how engineering teams are embracing (and sometimes resisting) AI.
At our recent breakfast event in London, founder Mike Daniel shared some of the early findings — and the patterns already emerging are both exciting and cautionary.
The State of AI Adoption in Engineering
Our data shows that 89% of organisations are now investing in AI in some capacity. Yet beneath that impressive headline figure lies a gap in maturity:
- Only 26% have a documented AI strategy.
- Fewer than 1 in 10 are tracking measurable outcomes from AI initiatives.
- 50% of engineers without official AI tools are using unapproved alternatives — the rise of “shadow AI”.
The message is clear: adoption is high, but governance and measurement lag behind.
Where AI Is Creating Value — and Where It Isn’t
Tooling investment is producing measurable efficiency gains, especially in greenfield projects using agentic AI for proofs of concept and rapid prototypes. These projects report 3–5x acceleration compared with traditional development.
Yet areas like training, R&D, and documentation remain underfunded. Only 21% of businesses are experimenting with internal tools or R&D teams dedicated to AI — the very areas likely to drive long-term value.
The Three Golden Rules for Responsible AI
Through our own research and conversations with industry peers — including experts from Deloitte — we’ve distilled three foundational rules for sustainable AI adoption in engineering.
- Human oversight is non-negotiable.
- The engineer who prompts the AI owns the output.
- AI should augment human ability — not replace understanding.
These guardrails are essential for maintaining quality, compliance, and trust as AI-generated code becomes commonplace.
Emerging Risks: Skill Erosion and the Tech Debt Paradox
Early data shows that while AI has dramatically increased code output — in some cases doubling it — productivity gains average around 20%. The difference lies in rework, refactoring, and review.
We call this the Tech Debt Paradox: as AI speeds up delivery, it also accelerates technical debt.
Add to that the risk of skill erosion, particularly among junior developers, and it’s clear that culture and process must evolve alongside technology.
What Comes Next
Through our own research and conversations with industry peers — including experts from Deloitte — we’ve distilled three foundational rules for sustainable AI adoption in engineering.
- Building internal AI strategy and governance
- Designing metrics to measure true impact
- Training teams to use AI responsibly and effectively
Our AI in Software Engineering Insights Report will be published in December 2025.
