JD is a software fanatic. At university he ran a PC repair business. After leading various software ventures, he founded Raygun, where they provide diagnostics services for fellow developers. JD was nine years old when he wrote his first piece of software on his parent’s PC. Fast forward to 2024 - and it’s AI that’s capturing his imagination.
AI compels founders like JD by having the potential to:
With that context laid out, here’s what he had to say:
JD possesses an unapologetic urgency when it comes to using AI. Having a plan is half the battle. He implores all team members to experiment with AI, not just developers.
“I'm in my 40s, I've seen two or three of these hyper cycles now. And every single time they get faster. And I don't know that the human mind is really capable of understanding exponential innovation. We do have some time, but the longer you leave it to actually try and upskill and understand how to leverage this, the higher the chance you become roadkill. It's as simple as that.”
At Raygun, they ran an AI week. The entire company participated, including non-coding folks like finance, customer success and design.
“Primarily, I just want to be able to have a high fidelity conversation about AI with my team. And at the moment, I can't, that's not because I know it all, absolutely not. It's actually that all of us will have different interpretations and different understandings. And if we can then have a dialogue around that, we're going to be able to make better decisions on how to leverage AI within our organisation.”
JD says that “AI is not just for nerds. That's one of the huge mistakes that I've seen some organisations make. It's going to affect everybody. It's like pretending in the 90s that the internet would only affect software developers. This is going to affect everybody. So we need to take everybody on the journey.”
AI Week started with everyone attempting the onboarding challenge their new engineers are tested on. Everyone passed.
After that, seven scrappy cross functional teams were formed. They worked without the constraints of JIRA or their usual development processes. Drawing on models like ChatGPT4.0, Claude 3 Opus, and Mistral, they trained models using their own data, safely hosted on an internal server. Having an internal server gave them peace of mind around data security and privacy.
“At the start of this week. We hadn't really used this technology at all. By the end of the week, we had a computer that we could talk to that could answer our business questions, and talk back to us, just like J.A.R.V.I.S. from the Iron Man films.”
How many times a day does a question in Slack disrupt your workflow? People must learn to fight muscle memory and use AI tools for generic queries that busy peers need not bother with. JD says his team has committed to new habits around using AI, and the company has backed that up with budgets for premium tools.
“GPT4.0 is a lot better than the free version. GPT4.0 is where you can upload like a CSV and ask questions. It's twenty dollars a month. Seriously, just pay for it.”
Having established data security guide rails within the team, JD says he’s trying to create a culture where it’s okay to play around as a team.
“Set up some channels in Slack. Have an AI one where you're just sharing some news about what people are seeing. Encourage people to be a bit goofy. As we get older, we get more and more afraid that we're going to break stuff and that we're going to do it wrong.”
AI can redefine how internal teams service other parts of the business. JD cites a big win from AI week at Raygun where they created a self-service business intelligence (BI) model. Previously, BIs were responsible for generating reports. Now, team members can upload CSVs into a model themselves and ask for insights directly. Now, the BI team is freed up for innovating on the business as a whole, rather than handling the technical side of queries they receive from others.
Ultimately, it's about how much toil they can subtract from their business as they scale up.
“Rather than going, oh my goodness, people are going to kind of lose jobs. What you want to do is be able to say hey, we're a 20 person company, a 30 person company, or a 50 person company. But we can operate as [equivalent to] a 10,000 person company now. That is the great democratizer.”
In the world of startup and scaleup, speed, in JD’s words “is, frankly, sexy.” Success means moving quickly and being faster than your competition.
“If you can try more things in the same amount of time, you're more likely to win.”
JD says they are already looking beyond LLMs towards agent-based-models (ABMs). ABMs are the next level above LLMs, where processes come together to provide a cohesive service.
“Agents are where you kind of string together complete capabilities with these things. Maybe it's a customer support bot, and it analyses the request, maybe it auto sets a priority, maybe it suggests a response, maybe you trust it enough to send the response, but maybe it puts the human in the loop.”
What could all this lead to? Faster, better quality customer service, and faster, less risky product launches.
One of the hard things with AI is that it’s not always easy to understand or score how you're operating and the impact it's having on productivity and profitability. JD encourages us to look at it through the lens of scaling. For product businesses, it’s typically seen as a mistake to scale on headcount, so tracking revenue per employee is the kicker.
“I would fully anticipate that in a business that is able to dramatically leverage AI that the revenue per employee skyrockets. You know, when we look at ‘mega-techs’, they're typically in the sort of $1 to $2 million of revenue per employee. We think we're probably looking at a future where you're going to see the first billionaire, that was a one man band, right, because you can leverage these systems and scale up a person substantially. They are getting better every single day.”
JD looks back on the proliferation of cloud services ten years ago as a blueprint for services business to harness AI.
“The opportunity if I was a service operator would be in education. We saw multibillion dollar businesses get built in sort of the 2014-2020 around helping the tech industry understand how to leverage the cloud, right? The cloud had already been around for a decade in the tech industry. Relatively speaking, it's a very tiny industry compared to the whole world. AI is the whole bloody world and everybody needs to learn it. So if I was building some sort of services organisation, I would try to be ahead of the ball on AI, because the people who aren't tech leaders need people in tech to help them understand how to leverage it themselves.”
Read, watch and grow in under 5 minutes a fortnight.
A newsletter designed by Atlas Digital.
