AI is beginning to change how software is built and who builds it. Discuss.

This virtual HotTopics C-Suite Exchange, in partnership with Nearform, brought together senior technology leaders to explore how AI is redrawing the boundaries of software ownership—and what new operating models may be required to balance speed, autonomy, and accountability in the AI era.

Across organisations, the boundaries between engineering, product, and data teams are becoming less defined. AI-assisted tools are enabling a wider group of practitioners to design, prototype, and even deliver software. At the same time, engineering teams are evolving—shifting focus from writing code to shaping systems, governing quality, and enabling faster, safer delivery at scale.

For CIOs and CTOs, this introduces a new set of organisational questions. If more people can build, who is accountable for what gets built? How should ownership of platforms, products, and data evolve? And what becomes the role of engineering in an environment where capability is more widely distributed, but risk remains concentrated?

AI is beginning to change how software is built and who builds it.

Here is what we learned

The strongest C-suite lesson from the session was that AI is accelerating software production faster than most organisations are improving the systems around it. In fact, the discussion repeatedly moved away from coding itself and towards strategy, governance, data, talent and organisational design. Specifically:

The constraint is shifting from “Can we build it?” to “Should we build it?

AI is making software materially faster to produce, but participants saw much less improvement in identifying the right problems to solve. Organisations with strong links between P&L owners, product teams and engineering appeared better equipped to turn that additional development capacity into business outcomes. 

AI is exposing weak product operating models rather than fixing them.

Several participants argued that businesses now discovering problems with AI-native engineering often never completed the earlier transition to genuine product-led working. Agile ceremonies are insufficient if customer need, commercial outcomes, ownership and experimentation are still poorly defined. AI makes that organisational debt more visible because poorly conceived ideas can now be turned into software much faster.

Data readiness may determine AI readiness more than access to models does.

Across conversations, unreliable, fragmented or poorly governed data repeatedly emerged as the obstacle between prototypes and dependable enterprise systems. One organisation had almost quadrupled its data-engineering team over five years and was only now seeing the payoff in faster product development.  Another described having petabytes of data but being unable to join it sufficiently well to generate the required insight. 

Governance has to move into the production process.

A sharp debate concerned whether governance should still be described as a “foundation”. One argument was that in an AI environment it needs to become an active part of delivery rather than applied afterwards as a “compliance gate”. In one organisation, governance and quality checks have effectively become invisible parts of analysts’ everyday workflows, we heard.

Democratised software creation creates a new enterprise integration problem.

Executives should expect more employees outside engineering to arrive with functioning AI-generated applications and prototypes. The problem is now how to decide how engineering teams safely absorb, industrialise and integrate that work into complex brownfield environments. The session described an emerging divide between rapid “vibe-coded” creation and the technical rigour required to reach production.  

Specification may become a much more important management discipline.

Spec-driven development was discussed as one mechanism for translating business requirements into AI-assisted engineering workflows. Early experience suggests significant acceleration is possible, but success depends heavily on context engineering, governance and the tacit knowledge that employees often struggle to articulate in a specification. It remains experimental for many organisations rather than a settled operating model.

Human judgement becomes more valuable as production becomes cheaper.

A recurring concern was that AI can generate more code and more potential solutions without necessarily improving judgement about quality, relevance or risk. Participants questioned how organisations will preserve senior expertise as junior work becomes increasingly automated.

AI readiness is ultimately a question of trust.

Organisations whose commercial proposition already depends on reliable data appeared to have an advantage: governance, provenance and quality are already linked directly to reputation and revenue. Businesses where data has historically remained invisible may face a harder executive challenge—making the economic case for disciplines that suddenly become essential once AI begins operating on that data at scale.  

Engineering the AI Era

In partnership with Nearform

Created in partnership with Nearform, Engineering the AI Era brings CIOs, CTOs and engineering leaders together to modernise software delivery and turn AI into measurable business impact.

Find out more