Start-ups move fast with AI-generated code

{"text":[[{"start":null,"text":""}],[{"start":6.8,"text":"Even by the standards of start-up life, Aleksi Tukiainen drives his new hires hard. Not content with the controversial 9-9-6 working week (9am to 9pm, six days a week), he expects them to work 24 hours a day."}],[{"start":21.45,"text":"Luckily Lisa and Dan, as AI agents, are not complaining — yet. Lisa, a sort of hybrid executive assistant and software engineer, has so far been the most useful to Tukiainen, founder of financial intelligence start-up Arctal. But it is Dan, the data guy, who shows the most promise."}],[{"start":40.849999999999994,"text":"Arctal’s bread and butter is data, and the start-up has tens of thousands of documents in multiple formats that previously would have taken a human analyst days to process. “Dan just grinds away in his data dungeon. No one needs to talk to him,” he says. “It is scaling quite nicely.”"}],[{"start":58.99999999999999,"text":"Advances in AI that have taken place over the past 12 months have changed what it takes to build, run and scale a business. Start-ups looking to grow quickly, yet with few resources to deploy, are leaning heavily on large language model-driven tools to gain an advantage over their better-established rivals."}],[{"start":null,"text":""}],[{"start":77.35,"text":"Since implementing AI tools across the business two years ago, ClickMechanic, a UK-based automotive marketplace, has increased revenue by 76 per cent while only increasing its headcount from 26 to 30. “We haven’t hired in customer support over the past two years, despite scaling from £3.4mn to £6mn in revenue,” says chief executive Andrew Jervis."}],[{"start":101.19999999999999,"text":"Another factor has been that ‘no-code’ AI tools — which allow users to create new systems using simple interfaces such as drag-and-drop — have allowed non-technical employees to develop products without relying on software engineers. This has eased a bottleneck that traditionally existed within start-ups with limited — and expensive — engineering resources. Instead of relying on engineers, non-technical staff are simply building their own resources."}],[{"start":128.7,"text":"“The people who use the tools build them and own them,” says Andrius Slimas, co-founder of Saltz, a Vilnius-based food marketplace start-up. “It is nuts, the efficiency that you have. You have someone who has never seen code in their life say: ‘I want this structure, or this interface, or this feature.’ We could not do this four months ago.”"}],[{"start":null,"text":""}],[{"start":151.85,"text":"But the advent of no-code solutions does not necessarily mean that start-ups no longer need engineers, says Julio Martinez, founder of Abacum, a New York-based financial information start-up. He cites the Jevons paradox, a theory attributed to 19th-century economist William Stanley Jevons, which states that even when technology makes a resource more efficient, the use of that resource actually increases rather than falls as a result."}],[{"start":178.95,"text":"“We see that [ . . .] using cloud or GPT is way easier. You get answers a lot faster. But people have 10 times as many questions. We haven’t laid anyone off. We haven’t replaced anyone with AI. In fact, we are hiring a lot of engineers, product managers and designers. Every new hire is like [hiring] two people in the past,” says Martinez."}],[{"start":198.79999999999998,"text":"Investors, too, speak of AI tools having tangible results. Unrest, a London-based VC fund, has a programme that matches investors with start-ups. Since integrating AI into the scheme, it has reduced the size of the team working on it from five to one. Not only that, but co-founder Pan Demetriou says that the number of investors participating is up 52 per cent and the number of meetings generated by the programme has more than doubled, partly because of the connections suggested by AI."}],[{"start":230.84999999999997,"text":"Orr Vinegold, Unrest’s other co-founder, says the advent of AI has also opened up the field of potential start-ups beyond software engineers or those who can afford to pay them."}],[{"start":242.69999999999996,"text":"“What we’re seeing now is sole founders or teams of two founders that are not technical but who are able to buy in the code and say: ‘I’ve got an initial [product], and I have got X many people already signed up to it’,” he says. In turn, that means a much greater number of potential investments for venture capital firms."}],[{"start":260.15,"text":"While AI has eased bottlenecks and saved money for a lot of start-ups, the extent to which it is used can vary a lot. Nvidia founder Jensen Huang has publicly said he wants to see his engineering staff burn through $250,000 of tokens each — the AI unit of currency and a proxy for how much AI people are using — per year."}],[{"start":281.84999999999997,"text":"Tukiainen is sceptical, both about the scale of token outlay and the ability of AI in its current form to perform the tasks that would justify that kind of spend. He expects the total spending across his team to be about £5,000 per person per year, or £25,000 in total."}],[{"start":301.54999999999995,"text":"Across the board, the big fear is that AI is moving too fast. Abacum’s Martinez admits that the risk of AI hallucinating — or inventing false content — already causes him to lose sleep. “We work with listed companies and their CFOs. Being accurate and rigorous matters a lot,” he says."}],[{"start":320.29999999999995,"text":"Saltz’s Slimas says his biggest fear is that the rapid improvement in AI accuracy leads to complacency. “They say that the highest danger in aviation is after a pilot gets 200 hours of flight time,” he says. “That’s when all the mistakes happen because you think: ‘I got this’.”"}],[{"start":345.69999999999993,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1777878082_1729.mp3"}