Trevor Gowing, Co-Founder and Managing Director, Decode

Decode is a boutique software product studio. We work with small businesses to build bespoke software products using open source web technology, and these days we increasingly use the term applied AI: integrating AI agents into the products we build for clients while still maintaining a focus on quality. That’s a big conversation in tech at the moment. You can go fast, but can you go fast and maintain the quality? The jury’s still out.

How is AI changing the way your team builds software for clients?

It’s changed things a lot. We started the year quite anxious, because AI really took a step change when it came to software engineering. But we’ve developed a new strategy based on a number of hypotheses about how AI will affect our industry.

One is that if it becomes cheaper and easier for people to build, more people will be building, not just traditional software engineers, but people in other roles across a business, which we think means more opportunity for businesses like ours. We also believe larger businesses will keep smaller in-house engineering teams focused on their core IP and look to outsource everything else. And we believe AI will kill time-and-materials billing, which we’re quite happy about. We never believed that was a good commercial model for either the client or the vendor.

So we’re moving toward value-based billing, packaging our services around delivering business value rather than billing time. If AI is doing a lot of the work, whether that’s automated agents or AI-assisted engineering supporting human software engineers, the value isn’t the amount of time spent on it. It’s the output.

In our experience, the biggest mistake people make when building bespoke software is either under- or over-engineering. If you under-engineer and your go-to-market team is successful at bringing customers on board, but your technology can’t keep up, you risk breaking the trust you worked so hard to build. If you over-engineer and build a system that’s too complicated for the stage your business is at, you risk eating up revenue or capital maintaining something that’s more than you need.

Building a system that can grow with the business, one that’s right for that particular point in the business’s life, can be tricky, and it’s something we feel we get right.

One of our best clients has a product used by music talent management agencies to run their businesses and handle billing in the music industry, which can be complicated because there are so many different players involved: the agency, the artist, the manager, the promoter, the venue. Reconciling billing across all of that is genuinely difficult.

They’ve been in the market for more than ten years and are one of the trusted leaders, and their product works well, but it was originally built more than a decade ago, so the interface is a bit dated. We’ve been building an integrated AI agent into their platform. We started by offering their clients natural language to SQL translation, letting them ask questions of their own data and the market data in the system in plain English. Then we moved on to accepting unstructured documents, converting them into structured data, and submitting that into the system.

Most recently, we pulled the AI agent out of the existing interface entirely and gave it its own full chat interface, similar to ChatGPT or Claude. Clients no longer have to log into the existing system at all to do their work. That’s had a real impact: it’s helped improve their sales and revenue and increase retention among existing clients as they adopt more AI into their own businesses.

I think people get too focused on the technology instead of the actual business problem and how to solve it for the people at the end of the day. Technology is there to serve people, not the other way around.

Traditionally, partly because of the cost of development, we sometimes had to settle for the lowest common denominator. That’s part of why we’re excited about AI: it makes it so much easier and more cost-effective to build tailored, personalized software. I believe we’re entering an era of personal software, where people build their own software for themselves, personally or for their businesses, and work with companies like Decode to help build and maintain it.

One of our biggest supporters and advisors since we started the business in 2019 was an ex-colleague of ours, Carl. From the start, he recommended we take on support work for existing systems to build a recurring revenue base. We didn’t do that. We focused on winning project work up front, with the idea of converting it into support and evolution retainers afterward.

We got lucky with a few good opportunities early on, which I think made us a bit overconfident and misaligned our expectations of what was practical. After a rough period following COVID, we realized our expectations had been too high. So now we’re genuinely focused on finding a small problem to solve for a customer, converting them into a retainer client even at a small value, adding consistent value, and upselling from there. That’s been a big pivot for our business over the last couple of years, and it’s really starting to bear fruit.

Following on from our new strategy, we’re launching what we call our zero-to-one service: a one-month product build. The first week is a design sprint with a UI/UX designer, a software engineer, and a product manager to plan the scope. Then three one-week engineering sprints deliver a production-ready system, not a prototype or a demo, something genuinely ready to be used by real people. We include two months of support and evolution services alongside it.

It’s a clear, understandable package of value that’s affordable enough for people to buy, and something we can build on from there, converting those clients into ongoing, retained customers and eventually annual projects. We kicked this off in the second quarter of this year, and so far it’s going well.