How AI is Re-Shaping Start-Up Engineering Teams In this new era of vibe coding and software builders, it has never been easier to start a tech company and put software in the hands of users.

There are stories of founders being accepted by YC with prototypes that they built in a matter of weeks. This would have been impossible just a couple of years ago. When I started my career in Silicon Valley 20 years ago this would have seemed like the stuff of science fiction. Anything that allows founders and developers to build faster is good news for the start-up ecosystem. It encourages more innovation, faster testing and unlocks the potential of founders who don't have technical backgrounds.

However, AI is not yet capable of producing truly production-grade software. And some founders are realising that too late. We know this because the big AI companies themselves are telling us. If AI was building production-grade software, then why are OpenAI and Meta offering $100m salaries to the best software developers in the world? So where exactly does AI help start-ups, and when does it start to trip them up?

Inception
AI software builders are excellent at building prototypes and software that can stand up to early testing. For the first time, solo founders can build software that can attract users in days. When start-ups are still in the days of going from 0 to 1, they should embrace AI to support with ideation, prototyping and QA testing. These are all vital elements of the early days of a startup. Most importantly, AI has taken down the barriers for non-technical founders to get to proof of concept and market fit with software companies. I hope this will lead to an explosion in BIZ Experiencesial innovation with new ideas and companies coming online in the next few months and years.

Early-stage
However, from my perspective, any start-up that is relying on software built purely by AI beyond pre-seed stage is going to run into difficulties. Even when startups are still early-stage companies, investors and customers alike will be looking for metrics that AI generated products can't yet deliver.

The truth is that AI is great for basic tasks, but complex infrastructure and projects still require expert developers to implement. So founders who have raced to a new prototype and have loudly talked about how they have built a new software company by themselves will start to face questions they can't answer about security, about how their software integrates with larger systems and about how their product or platform scales. So as start-ups grow, they still need to invest in strong teams of developers and engineers. The latest research suggests that AI tools are saving developers an average of just under 4 hours a week. That isn't nothing - that's a 10% increase in productivity. But it isn't quite as game-changing as the AI companies would have us believe. The biggest thing is to fight for talent. Developer talent will become more expensive.

Scale-up growth stage
For later stage tech companies, AI will be improving the efficiency of individuals and teams, but it hasn't re-written the rule book for how they operate. Klarna was a high profile example of what can go wrong when later stage tech companies swap developers for AI. Less than a year later, they were backtracking and trying to rehire everyone they had let go because the quality of AI agents wasn't good enough. Beyond the actual technology itself, the single biggest change for growth stage companies will be the race for engineering talent. AI has made simple coding tasks very simple which has created a very challenging environment for junior developers and coders.

But it has also increased the demand for experienced, highly skilled engineers. Developers who know how to build complex infrastructure, integrate agentic flows where required and leverage new AI technologies whilst maintaining the rigour and discipline of classic engineering will become gold-dust. More and more scale up companies will be competing for world-class engineering talent and will need to invest in individuals and consultancies who can deliver that work for them.

What comes next
The caveat to this advice is that everything can, and will change. The new generation of LLMs will bring new innovations and breakthroughs, and AI software builders are improving all the time. Production-grade software built by AI will become a reality in the next few years. So the impact on productivity will continue to improve with time. However, many of the fundamentals for software companies remain the same, especially if you are operating in a highly regulated or complex industry. Founders forget that at their peril.