Leadership lessons from corporate life to founding AI-powered startups
Moving from corporate leadership into startups taught me to rely on experts, question assumptions and give AI agents clear responsibilities.
AI-imagined illustration of our team using character-inspired AI agents.
I recently appeared on the Sustainable Tech Leader podcast with Mita Patel where we talked about leadership lessons and how we are now embedding AI into our our company operations. Below are some of the key takeaways from that conversation. I recounted one experience, where I took the role of CFO and COO of a company without any background in accounting or finance. I went from being the expert in the room in GE Commercial to having to rely entirely on my team.
Knowing the work can make it harder to let go
At GE, I worked my way from an intern to an executive leadership role in the commercial organisation. I understood the work because I had done it myself.
That experience was useful, but it also created a temptation to step in and do the job myself. When you know how you would do something, it is easy to take over before somebody else has had the chance to work it out. You can also dismiss an idea too quickly because it does not fit your experience.
The CFO role changed that. I could not fall back on being the person who knew every technical detail. I had to ask the team to explain things, sometimes from the beginning.
Questions about why we were doing something a particular way often turned out to be useful. I could understand the purpose, challenge the logic and help set the direction without pretending to know more about accounting than the accountants.
That did not remove the need to coach people. Much of the coaching was about prioritisation, working with colleagues and deciding what mattered most, rather than teaching the detailed skills of their job.
Changing direction when the commercial case changes
Another lesson I brought into founding businesses was the need to keep questioning an idea after I had committed to it.
My commercial background means I tend to start with the underlying business case. During my time working on green hydrogen, I kept modelling projects and trying to understand how the economics would work. I could not find a commercial case I believed in.
That led me to look for a different direction, and eventually towards the hybrid renewable-energy businesses I co-founded.
It is difficult to ask people to believe in a direction while also accepting that you might be wrong. But I do not think the answer is to stop questioning it once you have persuaded the team. I can champion an idea and still look for gaps in the assumptions behind it.
For a founder, that matters because there is nobody else who is necessarily going to tell you to stop. The more effort you have invested, the harder it can become to acknowledge that something no longer makes sense.
Building operations from the beginning
Starting a business gave me an opportunity to apply another set of lessons from being a CFO and COO.
I had seen how much work goes into reconciling information across different accounting systems, banks and spreadsheets. Different departments could each have their own version of the data, and getting a consistent view took effort.
With a new business, I could think about those foundations from the beginning. I started with the information we needed to run it, such as budgets and suppliers, and built from there. AI agents could then work with that information and the context of the business.
That is a different starting point from opening a chatbot and repeatedly explaining who you are and what you need. The agent has a defined role and knows where the relevant information is. The team can ask for something in ordinary language without rebuilding the brief each time.
The opportunity for me was to bring what I had learnt about operations into a small company, without simply recreating the collection of disconnected systems I had worked with before.
A clear role helps an AI agent too
I have found that agents work better when their responsibilities are specific. A finance agent needs different context from a project management agent or one helping with communications.
If I give one agent too many responsibilities and instructions, it can become confused. If its role is narrower, I can give it the information that actually matters for the task.
There is a familiar management lesson in that. When I moved outside my own expertise, I had to become clearer about what I needed from the people around me. Working with agents also forces me to think about the brief, the relevant information and the task I am asking them to do.
I still need to decide whether the work makes sense. Producing something more easily does not settle whether it is the right thing for the business to be doing.
Curiosity matters alongside experience
One of the things I enjoy most is working with people who want to experiment. Experienced people can understand an industry problem and recognise what is worth solving. Students bring questions and a willingness to try things that we might not have considered.
In the interview, I mentioned a student who produced a video animation of the business without being asked. He brought it to us, and we started discussing how to improve it. That kind of initiative is valuable because it gives the team something real to respond to.
AI gives curious people more ways to test an idea. It also makes the experience of the people around them useful in a different way: helping them understand which problems matter and whether what they have produced answers the question.
More capacity has not meant less work
I was less convincing when Mita asked about sustainable working. I admitted that I am probably working more than ever.
AI makes it easier to take on another task. Something I might previously have put aside now feels possible, so I say yes. I cannot honestly claim that this has automatically given me more free time.
We make sure that the students working with us put their studies first and have space away from the business. I need to take the same question seriously for myself.
The tools have changed a great deal since my corporate career. What I keep coming back to is deciding where I am useful, relying on the people who know more than I do, and being willing to question what we are doing. As AI makes more work possible, deciding what deserves our time is becoming an even bigger part of my job.
Listen to my conversation with Mita Patel on Sustainable Tech Leader.