How to get press coverage on a startup budget
Learn from experts what makes content newsworthy, how to do media relations yourself and get PR for your product and com...
Do you need to embrace AI to get investment? How can you use AI to grow your business? Those were some of the questions that drew founders to Stripe’s office in London on 12 July 2023 for our panel discussion, Turbocharge Your Startup: Embracing AI, co-hosted with Hotbed and Stripe.
The visionary founders on our panel have paved the way in AI innovation. They’ve created AI initiatives within their own startups or developed cutting-edge AI solutions that can transform how you work. Read on for a summary of the insights shared, and a list of the AI tools that might just be the secret to supercharging your company.
We’re at an interesting moment in time with AI – in the middle of three eras. When work first started in the field, nothing really developed for about 40 years. You could spend your life building something that didn’t work and didn’t take off. In the third era, everything will be AI generated and the way we work now will seem completely unreasonable.
But right now, we’re in the second era: still experimenting with the early versions of AI. And it’s exciting because that’s where the opportunity is. We have the ability to productise things now that in the future won’t even be attached to the term AI, because the technology will be so widespread that it’ll seem natural.
Co-founder and CEO,
In this post
You can build an investible team without a dedicated AI expert. Our panel agreed that investors expect a different level of in-house AI expertise depending on whether your company is a ‘pure’ AI startup or AI-enabled. An AI startup that creates and builds new technology will need more domain expertise than an AI-enabled company that combines existing technologies in new ways.
The recent explosion of high-quality, affordable AI tools means building an innovative and impactful AI startup is no longer restricted to teams with deeply specific technical knowledge.
Sean: I think of AI as a general purpose technology, like electricity. We don’t employ highly specialised people with PHDs in machine learning, because we don’t need that level of expertise to put AI to use in our business.
John: To investors, your ability to build something is more important than the specific make-up of the team. Investors want to see that you can do something that others can’t. Show that you have talent in the team, and that you’re the team to meet the challenge.
It’s a misconception that investors are fighting over each other to fund AI companies without early revenue. To prove your startup has potential, the same principles for pitching to investors apply in the era of AI.
Our panel agreed that the most important metric you need is traction, whether that’s early revenue or another sign that you’re building momentum if you’re pre-revenue.
John: Investors are looking for potential – and traction is the proof of that potential. It’s important that investors see that there’s a problem that you can solve. The most important thing is that you prove your company can work at an early stage. That can buy you time to focus on revenue later.
Sean: For us, the first fundraise is about the tech, the second fundraise is about the traction. All the funding we raised in our first round was on proof of concept. And by the second round, we’d built up to seven-figure recurring annual revenue.
Deep tech is a bit different, but for AI-enabled businesses, you do have to meet the usual standards: get customers, make money, grow revenue.
Martin: You can almost always raise a pre-seed on the idea – the dream, the team. Ultimately, after that traction trumps everything.
Plus, traction isn’t just about reassuring investors. It’s also crucial to keeping your team motivated and onboard. If you’re not growing, eventually your team won’t want to stick around.
Beware jumping on a bandwagon. It’s tempting to think you’ll be able to raise at a higher valuation if you sprinkle your pitch with AI buzzwords, but our panel cautioned against this. Instead, use AI as a tool to help you improve what you’re already doing.
John: Don’t fake it. If you’ve got investors pressuring you to pivot to AI, it’s tempting to tell them the buzzwords they want to hear. But they won’t be the right investors to take you where you need to go, and you won’t get far branding yourself as an AI company if you’re not.
On the other hand, using AI as a tool will speed you up and reduce your costs. The industry AI is revolutionising most dramatically right now is coding. You don’t need as many engineers to build an MVP. That means you can get to market more quickly.
Martin: If an investor is encouraging you to use AI in your startup, try to look at it from their point of view; essentially they’re pointing you towards the benefits of AI. They want you to have the speed of execution and defensibility that AI can give you. If using ChatGPT can make your team 10x more productive, why wouldn’t you use it to be more capital efficient?
Using AI well can help you outrun your competition, but there’s also the risk that if you don’t work out how to integrate AI, you’ll get left behind.
Like any startup, companies specialising in or using AI will succeed only by developing products or services people actually want to use. Our panel discussed how important it is to listen to your users and continually iterate.
Sean: Build a product, ship it, get feedback – repeat until you have a product someone will buy.
John: Go and speak to as many people as possible. Explore what’s happening. AI is changing so fast that we can’t predict where the opportunities are.
How do AI-enabled companies convince investors that their use of AI is valuable? And that the value doesn’t just come from the underlying, non-proprietary technology? Our panel emphasised that to succeed as an AI company, your product or service needs to improve on the standard AI tools available to everyone.
Alex: The algorithm is only as powerful as the data you feed it with. And even then, hallucination is a real problem with AI-generated content. If the model doesn’t know something, it’ll make an answer up and be confidently wrong. Many models are also pulling from older data as well, which makes the problem worse.
Our tool solves the problem of hallucination. We’ve worked hard to train it through a series of prompts to admit when it doesn’t know something. That means customers can create a custom chatbot for their brand using My AskAI and know it won’t create misleading content.
James: With Hyperscale, our business uses AI to find sales leads and send personalised messages at scale. The value is not just in personalising with GPT4. It’s in the data we feed it, which ranges from targeting (identifying which customers are more likely to buy) to converting (assessing how your product is relevant to the customer based on their public profile).
Elaf: A lot of people can build on top of large language models, and the value comes from building that layer perfectly for your perfect target customer.
Thanks to scrutiny of high profile companies like Meta and Apple, consumers (including businesses) have an increasingly sophisticated understanding of how their data is stored, used and shared. Our panel agreed that if you’re an AI pioneer, you should take the lead by being responsible in how you use data from your users – whether that’s their personal details, their images or their prompts and chats.
James: That’s something we’re being very careful about. We only use public data points from LinkedIn and Crunchbase for our sales outreach work, for that reason. Eventually, we want to be able to feed in our own database, so we’re figuring out how to encrypt that data.
Alex: We see the risk as coming not from the AI model itself, but the people building on top of it. On our chats, people can choose whether or not their conversations are fed back into the model.
Elaf: As always, regulation is behind innovation. All the images we produce are public by default. We have to be very clear with customers that their images will be public. We’ve found that most people are happy with that, because if someone is using their branded image, it’s probably to market their products, so it’s a win-win.
If you’re running a tech startup, then of course your tech ops are vital to the company’s success. If you’re an early-stage AI company looking for investment and you don’t have a Chief Technology Officer, then it’s probably time you to move this up your to-do list.
Elaf: If it’s even a thought that not having a CTO might be a problem when raising investment, then it’s time. Remember that at an early stage, each hire will have a significant effect on the direction of the company, so it’s important to get the tech team right from the start.
Alex: It’s always risky to rely on a third-party sales provider over the long term. If you’re building a tech company, then you need the tech expertise in-house.
Our panel was keen to share their wins from trying out AI tools. If you have a tricky task ahead of you, try asking an AI tool to help – you might be pleasantly surprised by the results…
Elaf: We used Chat GPT to write a legal case against a company that copied our site. It was probably complete rubbish, but it looked official enough that the site was immediately taken down.
Perdie: At Hotbed, we used Midjourney to create visuals that represented how Hotbed is the perfect place for startups to thrive, even in less-than-ideal conditions. We wanted a garden of Eden flourishing on the bare landscape of the moon. Obviously, getting this as a real image is impossible. In partnership with a brand designer, we used Midjourney to create the graphics – and we were really happy with the results.
Looking for AI inspiration? Here’s a list of the tools built by our speakers or that came recommended by our panel and audience.
As well as helpful insights, new contacts and fresh inspiration, attendees went away with exclusive discount codes for several AI tools. To make sure you don’t miss out on any of our in-person events or webinars, sign up to our events newsletter.
Main image: Generated by DALL-E 2