How AI Can Supercharge Your Scale-up

When Prime Minister Rishi Sunak addressed the capital’s innovators at London Tech Week in June, he was quick to raise the topic of the moment. “AI is surely one of the greatest opportunities before us,” he said. “Already we’ve seen AI help the paralysed to walk, and discover superbug killing antibiotics. And that’s just the beginning.” This general-purpose technology, he said, would fundamentally transform society. “Right now there is an opportunity for human progress that could surpass the industrial revolution in both speed and breadth.”

Amid such grand rhetoric, it’s unsurprising that some small and medium-sized enterprises (SMEs) feel at sea. Is AI relevant to their operation? Is it affordable? What are the risks? And, well, where should they even begin?

“I’m hearing from a lot of SME owners that it's all so fast moving and so huge, it’s causing a deer in the headlights kind of response,” says Melissa Snover, founder of tailor-made nutrition company Nourished. “But really the biggest danger is not doing anything.”

Nourished is part of the WIRED Trailblazers programme, an ongoing partnership with HSBC UK supporting innovative UK businesses as they scale. Many of those taking part are already using the latest wave of AI tools – known as ‘generative AI’ thanks to their remarkable abilities in generating new content – to boost productivity, reduce costs and find a competitive edge. The use cases span everything from conducting research to creating web imagery and writing emails.

Clearly, AI is no longer just for large organisations. In fact, says Snover, one reason there is so much excitement around generative AI is precisely because it’s so accessible. These tools are inexpensive, work out-of-the-box and have natural language interfaces. Ask ChatGPT in plain English to draft a marketing blog or write some code, and it does your bidding. No specialist knowledge or implementation required.

"And that makes it more relevant to more everyday tasks than traditional AI, which is usually narrow in its focus and more technical in its applications."

So what advice do some of the UK’s trailblazing businesses have for SME leaders looking to adopt this technology? Here are five key tips…

Generative AI is evolving so rapidly it can be hard to keep pace. Reading specialist news sites, following clued-up Twitter feeds and learning from your peers is essential. But it’s not enough simply to learn about AI in theory – you need to try it out yourself. “One of the most bizarre things is I come across a lot of sophisticated people with great opinions about AI, but they’ve never even used it,” says Dan Vahdat, founder of remote patient monitoring company Huma. “I'm like: if you haven't even used it, don't pretend you understand it.”

As a starting point, Vahdat advises simply searching the web for the popular tools and trying to incorporate them into your daily routine. “If you're reading a document, get AI to summarise it for you; if you're writing an email, or preparing a speech, work with generative AI,” he says. “Then you’ll understand not only how it works but also its limitations – and get more ideas for how it can be used.” "

Typical IT implementation is top down: management identifies a problem, settles on a software solution and then asks employees to use it. Generative AI doesn’t lend itself to this approach, argues Peter Lipka, co-founder of virtual worlds company Improbable. “It's just overwhelming in some ways to try and target a particular area,” he says. “I think it's about empowering people to bring the right tools into their workflows – and my advice is go far and wide because this needs to be about wholesale movement within your organisation.”

Employees will need support throughout that process. At Nourished, which has also asked its employees to proactively seek out and start using generative AI, fortnightly all-hands meetings are used to bring everyone up to date on industry developments and also offer learning sessions. “I have taught them about ChatGPT; I have worked with them on content creation with [image generation tool] Midjourney,” says Snover. “One of the things that has been recently making a lot of difference is prompt training, to make the AI do what you want it to do.”

As with all organisational change, inertia can be a problem. Teams may pay lip service to the idea of using these tools, but in practice may stick to what they’re comfortable with. Improbable believes the best way to overcome this is to correctly align incentives. “A principle that we've put in place is that we’re going to be assessing individuals as though they have access to AI,” says Lipka. “That's quite bold, but it's that acceptance that we all want to be augmenting ourselves.”

AI is powerful, but it comes with risks. It can “hallucinate”, for example, producing content that is convincing but false. It can be biased, reflecting prejudices in its training data. There are also intellectual property concerns: AI tools might incorporate your inputs into its underlying model; equally, if you use its output verbatim, it could expose you to a copyright claim.

That’s why it’s essential to establish clear rules for how your organisation will use AI. “We have some basic guidelines,” says Snover. “Never, ever, put customer or confidential data into an AI unless you’re certain it’s a private environment. When you are using AI, you must check the answer. And you can never copy and paste the output – it should be a baseline starting point for your own work.”

To some extent, the rules you set are about your risk appetite. “We decided that our ability to augment ourselves for productivity is more important, in some ways, to our competitive advantage than IP concerns,” says Lipka. “So we obviously do our best for IP protection, but as some of these issues are genuine unknowns, we don’t let that make these tools off-limits.”

When assessing how much value AI is delivering, standard metrics can be helpful. If your team is more productive, or customer retention increases, that’s obviously useful data. But the technology and its adoption is so nascent that it can be hard to track return-on-investment directly. For one, there might be a multitude of tools being used by different people in different ways, so it’s challenging to pinpoint which is helping best. For another, generative AI is not just about improving productivity, but also the less tangible aspects of work such as quality of ideas. Moreover, people are still figuring out how best to use them.

“The jury's out for how we measure this the right way,” says Lipka. “I think ultimately, for now, the most important metric for augmentation would be actual utilisation of the tools – how many tools are they using and how often?”

For Vahdat, gut feeling can also play a role. “If an app delights you, that usually means it’s a great thing,” he says. And if it's like, ‘Wait, what is this?’ then it usually means it's a bad thing. It’s like your phone – you spend hundreds of pounds on it, not because of some ROI calculation, but because you just know that it really improves your life.”

Staff will naturally be worried about unemployment, so communicate how the company envisages AI impacting their teams, and address concerns openly and honestly. “I've had to stand up in front of groups of people, and they say, ‘Will AI take my job away?’ That’s not them trolling me, they’re genuinely asking that. So you have to have good answers and a clear understanding of what that really means for them. I've always said: don't look at it as a replacement. Always see this as augmentation. And if the organisation can be more productive as a result of using it, we can actually do more things and that will create more work.”

In fact, notes Vahdat, AI can help people adapt as the workforce recalibrates. “AI is a very powerful tool to help people reinvent themselves,” he says. “To give you one example, we didn’t used to be open to hiring customer support people from anywhere in the world because of language barriers. But now with translation AI tools, someone who has basic but limited English language knowledge can probably communicate through email chat as well as anybody. And these tools are only going to get better and better in years to come.”