The Translator: Solving the ‘Build vs Buy’ Dilemma with Ben Lee, Bidwells

Richard Breeden: Welcome, Ben. You’re Head of Data and AI at Bidwells, a leading property consultancy founded in 1893. You’ve been instrumental in bringing interesting new technologies into what is quite a well-established and traditional business. Your background is actually in journalism — how does that lead you to a job in AI?

Ben Lee: To me it seems very logical. I was a journalist 25 years ago, then spent my career in communications within real estate. I joined Bidwells as a communications manager nine years ago. Technology has always been my passion. I was writing about Facebook when it first found its way into Cambridge University. I could see how much it changed the newspaper industry, and I could see how much it was changing real estate.

Ben Lee: About three or four years ago, I first started to see how AI was impacting the marketing role I was in. We were using what was essentially an early version of GPT-2 in a product called Jasper. I could see immediately how generative AI could solve a lot of the problems our team was having. So I effectively appointed myself Head of AI from that moment on. I was talking to the business about it nine months before ChatGPT launched. When it did, everyone thought I could see the future and gave me the role and the runway to see where I could take it.

Richard Breeden: Did you have a technical background prior to that or were you educating yourself as you went?

Ben Lee: I know enough about how the technology works to be dangerous. But I think I was appointed because I understood the sector, the business and the people in our business so well. They could have gone out and got a machine learning expert who could train a data centre in their sleep, but there would have been nothing to apply it to at the time. What they really needed was someone who understood the operations and the business and could translate that. To me, it’s all about translation. I operate as a translator between business users and the technology team.

Richard Breeden: Bidwells has publicly stated some ambitious goals around doubling revenue by 2030 without proportional headcount increases. Where do you see AI actually contributing to that?

Ben Lee: First thing to say: AI is not going to get us there. We’re certainly not under any illusions that by adopting a few AI tools we’re suddenly going to be able to double the size of the business. That’s a growth strategy and there are going to be many different parts that will hopefully come together to make it happen. That’s going to be led by our people. The reason we framed it this way is that it gives people in the business a clear idea of what we’re using technology for: how can we enable our best people to do more so they can go out and grow the business, and how can we do things in a way we haven’t done before? The alternative is to maintain the size of the business but cut costs. By stating that, and having everything we do aimed at growth, we’ve given people a clear North Star. You have to be ambitious. If you get halfway there, you’ve done well.

Richard Breeden: Tell me about your technology and innovation committee — the idea of workflow first rather than tech first.

Ben Lee: Lots of people have formed AI councils in the last couple of years and I think there’s nothing wrong with that. Having a place where you can organise innovation and give people a clear channel to get support or funding is important. The reason we went for a technology and innovation committee is that AI is quite often not the answer to the team’s problem. We try to shift people’s thinking: let’s put the technology to one side for a second. What problem are you trying to solve here? What is your goal? What problem, if you removed it, would enable you to meet your stated goal? How well-aligned is the removal of that problem with the strategy?

Ben Lee: That helps to triage some ideas quite quickly, but it also encourages people to put their ideas forward and know how to get support. The committee is representative across all ten departments. What you’d have previously seen was lots of departments doing their own due diligence on technology in isolation, ending up with duplication or systems they’re not getting value from. The committee helps people in teams see what’s going on in other parts of the business and realise they can perhaps use something that’s already being developed.

Richard Breeden: How do you manage the fact that different parts of the business have different levels of influence and the best ideas might not always come from the loudest voices?

Ben Lee: That’s the biggest challenge. In mid-market companies, and particularly in professional services, you’ve got teams who are better at making their voice heard and people who are more influential. But some of the best ideas often come from parts of the business that don’t often get heard. We’re also the number one adviser to the science and technology sector, and what we hear from our clients is that when you bring together two teams from completely different disciplines, that’s where you get the best ideas. So we try to live by that.

Ben Lee: In the end, what my role fulfils is that time-consuming part of technology work — talking to people about what they’re trying to do, helping them understand what the right solution might be, helping them reframe problems. That is not a technology problem. That is a culture problem, a stakeholder management problem. Technology can’t replace that. We spend a lot of time understanding people’s problems and defining requirements really clearly. By doing that upfront work, the build and deployment of things becomes much smoother.

Richard Breeden: How do you think about which AI projects are applicable across the whole business versus solutions for specific pockets?

Ben Lee: We have four buckets for projects. Some are self-serve: people with the right AI tools can solve their own problem with a bit of curiosity and the right licence. Some require a bit more support — they might need technical expertise from our team, or have a data problem that needs solving. Then you’ve got the point where you need to procure a solution, go through a detailed requirements-gathering exercise and work out whether you build or buy. And then you’ve got enterprise-wide challenges that are something very different again.

Ben Lee: In terms of use cases, real estate is a very traditional, repeatable business. There’s a whole layer of things which are effectively document generation exercises with some orchestration of data — how meetings are run, how leases are reviewed, how scoping reports are written, how site appraisals are carried out. These things translate across all ten departments. The things that translate across the whole business are the quick wins — the things you can build a tool for, make available, and everyone can find a really good use for.

Richard Breeden: Do you have an inbuilt preference to build or buy?

Ben Lee: Our default position is to build, because our technology team is as big as our IT team and we will always feel we’ve got the capability in-house to build a lot of the tools I’ve talked about. But it comes down to resources. There comes a point where you have to prioritise and think: yes, we could build that, but actually there’s a solution that fits our requirements very well and we’re going to get value from it — so let’s just buy it and focus our in-house resources on something we can’t find elsewhere. It’s portfolio management: making sure the portfolio of projects we’ve got is distributed in the right way.

Richard Breeden: You recently ran a hackathon with around 40 university students. What did you learn from it?

Ben Lee: Really smart students from some of London’s best real estate institutions. We gave them a real problem — something our planning department had told us would be really valuable if they could solve it. We gave them a detailed brief beforehand so they had a good understanding of the planning sector and what we were trying to achieve.

Ben Lee: The teams that won weren’t necessarily the ones who came up with the most sophisticated use of technology. The most successful ones spent a lot of time understanding what we were trying to achieve, spent a lot of time asking questions of the planners who were at the event, and really took as much time as possible to define the problem before they chose the right technology. We were impressed by that more than anything else. It is quite easy to jump straight to a really spectacular machine learning exercise — when actually the winners didn’t need to do that.

Richard Breeden: So it’s critical thinking rather than the tech being the thing that comes first?

Ben Lee: The people I really listen to in all of this are the ones who aren’t talking about technology. They’re talking about their disciplines and thinking about what’s going to be left, what are we going to do? And that has to be understanding your domain, your discipline, your sector, your business — super, super well. And being able to explain that to the technology itself.

Richard Breeden: What’s the one thing you think people are missing or underestimating?

Ben Lee: Defining the problem and being clear about why you’re doing something. That’s vital. But also: if you don’t understand yourself the steps involved in doing a task, and you don’t understand what good looks like, and you don’t understand the thought process and judgements made whilst doing that task, then of course AI is not going to be able to do the job the way you want. What is a differentiator is understanding what it is that you do in your business that is different to your competitors. Build that into whatever tool or task you’re working on, and then use AI to help you leverage that expertise.

Richard Breeden: How do you balance governance with the pace of change?

Ben Lee: Having an IT and technology team working together very closely is so important. We use the Microsoft stack that we’ve had for 20 years and we will continue to use that to deploy Copilots, agents and workflows. We can build things quickly with a third-party tool, prove it works, and sometimes you can move more quickly that way. But if we’re going to deploy anything across the whole business, it has to be in the places where people are already working — not just so that people adopt it, but so we and our clients have the comfort that their data is being housed in the UK in the places we’ve always housed it.

Richard Breeden: If you could go back to day one and give yourself one piece of advice, what would it be?

Ben Lee: I’m surprised at how slow adoption has been. I thought there’d be lots of people like me spending time learning, teaching themselves how to use the new platforms, getting excited by the potential. I would say to myself: actually, it’s going to be a bit slow. But it’s been good in a lot of ways because it’s meant you can be more strategic and more thoughtful. What I would also do — and what we’ve done more recently — is have the leaders in the business be really clear that AI is not really an optional activity any more. Everybody in our business now has a stated performance objective to identify an area of their work which could be supported by AI. It solidifies that this is going to be part of the way everybody works.

How is a 130-year-old property consultancy using AI?

Bidwells, founded in 1893, is using AI to target a doubling of revenue by 2030 without a proportional increase in headcount. The strategy is growth-oriented rather than cost-cutting: the aim is to enable Bidwells’ best people to do things they haven’t been able to do before. AI is being applied to document generation, data orchestration, lease review, site appraisal and project management — tasks that are formulaic and repeatable across all ten departments — as well as to more specific, high-value work tailored to individual client needs.

What is the build vs. buy decision framework for AI tools in the mid-market?

Bidwells defaults to building in-house, given the size of its technology team. But the decision ultimately comes down to portfolio management: if an off-the-shelf solution fits 90 per cent of requirements and the cost is comparable to building, the right answer is to buy it and focus in-house resources on problems that cannot be solved by the market. The risk of defaulting to build is that engineering talent ends up solving problems that have already been solved elsewhere — creating technical debt and slowing down the work that genuinely requires proprietary expertise.

Why is a journalism background useful for leading an AI function?

Ben Lee’s argument is that the most valuable skill in an AI function is translation — the ability to bridge the gap between business problems and technical solutions. A journalist learns to understand a domain deeply, ask the right questions, identify what matters and explain it clearly to others. These skills transfer directly to the work of running a technology and innovation committee, defining requirements for AI projects, and helping teams across a business articulate what problem they are actually trying to solve before reaching for a tool.

What is a technology and innovation committee and why is it better than an AI council?

Bidwells chose to form a technology and innovation committee rather than a standalone AI council because AI is frequently not the right answer to a team’s problem. The committee’s first question is always: what problem are you trying to solve? Only after the problem is clearly defined does the committee consider whether technology — and specifically AI — is the right response. This approach reduces duplication, prevents departments from buying solutions they will not use, and allows good ideas from less-heard parts of the business to surface.

What did Bidwells learn from running a hackathon with university students?

The teams that produced the best results were not the ones with the most sophisticated technical approach. They were the ones who spent the most time understanding the problem. They asked more questions of the planners present at the event, spent longer defining what success would look like, and only then chose their technology. This reinforced Bidwells’ core principle: the quality of problem definition is the most important variable in any AI project, and critical thinking matters more than technical fluency.

How should mid-market firms handle data governance when adopting AI?

Bidwells uses the Microsoft stack it has maintained for 20 years to deploy Copilots, agents and workflows, specifically to ensure that client data remains housed in UK data centres. Ben Lee’s advice is to prove new tools quickly using third-party platforms, then deploy at scale only within the governance framework that already exists.

Why has AI adoption been slower than expected in professional services?

Ben Lee attributes slower-than-expected adoption to the fact that most professionals love what they do and have developed strong working habits over years or decades. AI adoption requires them to change those habits, and that takes time. He also notes that his own perception of the pace of change was skewed by the fact that he was immersed in it — most people are not. He expects another step change in adoption in 2026 as the impact of AI on software engineering becomes visible to the wider workforce.