The problem with building businesses with AI

Gareth Jones - Founder & CEO, TownSq

Subscribe to the Businessin Wales daily newsletter for FREE here.

AI has taken the time out of building apps, but building a successful business is harder to shortcut.

At the excellent NxtGen AI conference in June I was asked a question during my session about the best platform to use to build apps on – but it’s clear my answer split the room.

My session was on using AI to write business plans, and it was mostly a vehicle for me to rant about all of the terrible business plans I’m seeing come through at the minute that completely misuse AI as a tool.

It’s so much fun using AI to vibe code small apps for fun quiz games at pub dinners with the family, or to map out a DIY project, but I also know of Welsh entrepreneurs who’ve vibe coded their way to six figure businesses.

The worry, and risk, is when non-technical founders take a product that superficially looks ready for the big world, but is really setting you up for a lot of headaches.

Where the barrier until recently has been the cost of designing and developing an app, from the human cost, management, and prioritisation, AI has made it as easy as an afternoon sat at your desk fielding emails to create powerful and sophisticated tools.

Paying to play

But that cost is normally just an entry fee. The cost to successfully deliver an app is not just down to building the app – the significantly more expensive part is to drive engagement and keep it online.

To the question I was asked about which platform to use to build apps, my answer would probably annoy the vibe coders: focus on finding a tech-minded co-founder who can utilise the AI tools to build what you need.

When you launch an app, especially one that handles any personal data, you’re creating a whole range of responsibilities from keeping it safely online, to regular patches and bug fixes.

More importantly, you need to drive people to the app. Growing your user base is normally a very stressful and expensive exercise, which is confused by people believing that your app will go viral and you won’t need to worry about it.

Better than slop

The American VC General Catalyst – an early-stage investor in companies including Airbnb, Stripe and Canva – recently introduced a test for prospective investments: before committing capital, its team attempts to recreate a company’s software or app using AI in-house within a day, as a way of judging whether the business is defensible.

The big idea will become less important, and I’m sure I’m not the only one who is constantly bombarded with AI slop in my inbox and even on social media platforms like LinkedIn. My level of attention online has already massively dropped off, and my likelihood of opening an email which feels even slightly like a cold email has dropped to zero.

The hard work to get people to care about what you’ve built is only going to get harder, no matter how much your marketing strategy is, in the words of Claude, “genuinely impressive”.

Subsidised AI won’t last forever

My bigger worry behind all of this though is the sustainability of this model. We all know (I assume) the immense pressure on energy and water systems, but AI is another industry that has been propped up by VC and hyper-inflated valuations. There was a report released in June by SemiAnalysis that suggested that Pro subscribers to OpenAI (the company behind ChatGPT) are potentially getting up to $14,000 of usage per month from a $200 a month subscription. 

This is a bit like saying that someone at a £20 all-you-can-eat buffet could eat £1,400 of food, while knowing it isn’t humanly possible – but with chatbots you’re constantly being encouraged to use it more and more – which makes this scenario much more likely that you could be using way more than your subscription.

This discount won’t last forever, and there’s growing suspicion that the prices are going to rapidly increase as we become more and more hooked and reliant. This is mostly speculation – though API prices have already increased three times in the last year – but for someone looking to grow a tech company the idea that your developer, designer, and marketer are going to request a 70x pay increase is a big red block on the risk register. That’s what we’re relying on AI being for our companies if we’re not careful. 

You could end up with an app that you’ve spent ages working on, and can no longer afford to sustain or maintain, and no idea where your blind spots are.

In that scenario, the founders who aren’t afraid to ask for help and be honest about their shortcomings will be at an advantage – and the ones who relied on AI for everything will be left scratching their heads and wondering where it all went wrong.