Careers, in their own words

Clients tell me I found them a great candidate. Candidates tell me I kept them updated. Both are true, but neither is really it. The difference is that they felt understood — I could tell them how I was going to get to the result, not just promise it.

It's the symptom. The placement, the shortlist, the speed — those are all outcomes. What sits underneath is asking enough questions to understand the business, the hiring manager and what the candidate actually wants, before anyone else has clocked the risks. Recruitment isn't really about matching CVs. It's about reducing uncertainty for everyone involved.

It usually doesn't. A CV tells you what someone has done. It rarely tells you how they think, how they influence a decision or how they change the outcome. That only really shows up in conversation — which is why I still back interviews and references over keyword matching.

More often than not, honestly. A client wanted a senior infrastructure engineer with a long list of technologies and certifications. A couple of meetings in, it was clear the real problem wasn't infrastructure at all — they needed someone who could communicate, influence people and bring structure to a growing team. We rewrote the brief and hired someone with a different technical background but far stronger leadership. Recruit against the written brief and you find exactly what was asked for. Recruit against the actual problem and you find what they needed.

AI's been brilliant at taking the admin out of recruitment — summarising interviews, drafting adverts, tidying up job descriptions. What it still can't do is judgement. And interview prep is the one thing I'd never hand over. The insights and tips that get a candidate genuinely ready are the recruiter's job, and it's in my interest to do it well, because how they perform reflects on me.

Early on I was focused on the craft — the detail, the execution, making something people would look at and think that’s impressive. Now I can step into almost any problem space, however messy, and trust that I’ll understand it and make good calls. That didn’t come from getting things right. It came from years of experimenting, failing and shipping real products. The shift was going from needing to prove I could design something, to knowing I create value through the decisions I make.

Customer research has changed dramatically. It used to be slow and expensive, a lot of investment before you got to anything worth acting on. AI tools can now gather and synthesise insight at a scale and speed we didn’t have. What they don’t replace is the interpretation — the judgement and empathy to work out what any of it actually means. The point isn’t removing people from the process. It’s removing the repetitive part so we can spend our time on the harder problems.

The feedback I value most is when someone says I saw something they hadn’t considered. Most of my work has involved tangled customer journeys, multiple systems and competing priorities, and the rewarding part is connecting those pieces and finding the opportunity that was hidden in the complexity. Taking something that feels unclear and making it simple and actionable — that’s where design does the most.

I start with facts. When people disagree it’s usually because they’re coming at it from different experience and assumptions, so we establish what we actually know first, then what we believe, then what we still need to test. And a lot of disagreement isn’t really disagreement — it’s people wanting to be heard. Once they feel part of it, alignment gets much easier. Healthy debate is a sign of a strong team, not a broken one.

The invisible parts of leadership. A CV shows outcomes and deliverables, but not the judgement and influence it took to get there — how you brought clarity to something ambiguous, or kept people motivated through uncertainty. It doesn’t capture the bit outside your own role either: unblocking colleagues, mentoring, building trust across teams. The biggest value a leader brings is often not the things they personally create, but the conditions they create for others to succeed. That’s incredibly difficult to put on a CV.

Usually that I made something complicated feel clearer and more manageable. And that they felt heard, especially when things were difficult or emotionally charged. I try to understand not just what someone's saying, but what's driving it and what they need to move forward.

That's the outcome, not the underlying thing. The real difference is probably that I connect things across disciplines — I've worked across commercial, operations, technology, people and strategy, so I don't look at a problem through only one lens. What looks like clarity is usually just having asked what the goal is, what it'll take to deliver, what could go wrong and who it affects — and caring that everyone can see a path forward, even if they don't all agree.

Listening for what isn't being said. Watching who's speaking and who isn't, and whether someone's point is being dismissed for how they said it rather than what it was. I try to make sure the person closest to the work is heard — seniority and insight aren't the same thing. Mostly I'm helping a group get from discussion to a decision they'll actually act on, rather than one that's been imposed on them.

That's exactly the problem. A CV can say I led operations, technology or transformation. It's much harder to show that a lot of my value is in connecting those areas, bringing clarity to ambiguity and helping people make better decisions. CVs describe what you were responsible for. They're not good at how you actually operate — and in senior roles that's most of the job.

Judgement. You can delegate activity, analysis, even big areas of decision-making, and you can use technology to make all of it faster. What comes back is the call you have to make when there's no perfect answer, or when the technically correct answer isn't the right one for the people involved. That, and accountability — you can give people ownership, but you can't outsource creating the conditions for them to succeed.

That's changed in the last couple of years. Five years ago I think they'd have told you I was diligent about understanding the problems and the drivers of a client's business. Twice in the last twelve months, two different clients have told me that what they admire is fearlessness. Both times it was a surprise. Both times they said it was something they wished they saw more of.

I think they're actually the same thing. In both cases I was pitching ideas that were unlikely to succeed, but that were hard to deny the value of. My thinking was that even though they'd probably say no, the way they said no would be really valuable. With one of them it led to a subsequent deal, because the place the no came from was their own internal way of solving the problem.

Putting decks together has got so much easier with AI. Account planning, territory breakdowns — there's still a big advantage to experience in those, but the tools available now let anyone do them.

Execution. Balancing time against activity, having the human conversation with a client or a prospect. Although — filtering all of that into next steps is something I now lean heavily on AI for. So I can imagine that all these things are only a matter of time until the human is not necessary.

Great question. Don't know how to answer it.

I was handed a project that was bleeding red. We had a serious budget overrun, and my job was to bring it back on track financially. Part of the issue was poor time recording on work already delivered; another part was broad assumptions about the work still to come. Between these issues, it was almost impossible to get executive leadership to approve budget for anything outstanding. So I meticulously analysed the root cause of the overruns. I fixed the time recording process, built a tighter estimation model so effort and resource allocation matched what was actually needed, then put in a weekly review of the burn rate and the forecast. There was still about a third of the project to deliver. If it hadn’t been brought under control, there was a real risk the whole project would have been canned.

I mention this as an accomplishment for effective budget management within a specific project.

In a matrixed organisation that’s most of the job. The hardest was a programme with multiple streams running across in-house teams and several vendors, where every team’s deliverable depended on someone else’s. What it came down to was being explicit: articulating what I needed from each of them and what a delay from one team did to the critical path for everyone else. And constantly reminding people what was in it for them to get it done, and what the repercussions were if they didn’t. A structured, regular cadence with the right stakeholders is what actually held it together.

What I choose to photograph hasn’t changed. I use the generative tools to assist the process rather than replace it. I have been able to fix images with them that I’d otherwise have rejected, or wouldn’t have photographed at all. Generative tools have helped me overcome a few limitations I had with my equipment or environment.

Generally the beneficiary of my work is a stakeholder, a customer or a CEO, who has had some burden eliminated. That might come through refining a product’s user interface, or through managing a critical business operation. They’ll inevitably say that improving the process is what made the difference.

They’re describing the benefit, the outcome. What really makes the difference is organisational focus, ownership and enablement. Without a clear mandate that understanding and improving customer experience matters, you don’t get the resourcing or the processes that enable it. Without ownership, a team won’t take the initiative to work through complex multi-party problems, and the stakeholder’s pain persists unnoticed. They report the resolution of their symptom, but the seed of the solution was laid far earlier by the organisation’s culture.

We engaged an external firm to redesign the mobile purchase path for our flight booking product. After the first few iterations it was clear they didn’t understand the nuance of the product or the constraints we were working within. Rather than let it run on, I pushed to bring the project in-house and applied a prototype and user-feedback loop until we had something robust, de-risked and ready to go to development. Without that intervention I believe the project would have incurred significant costs and delivered a high-risk, unvalidated solution, leading to a significant loss in sales. Identifying the risk early and applying sound process is what made the outcome work.

This is essentially the binding thread of my multi-faceted career, and it sits at the top of my CV: the ability to identify risk early and mitigate it with best practice. It spans web development framework selection, design process methodology and the establishment of startup financial operations.

I documented one case on my blog where I tried handing off the build of a complex feature of Web3Connect to AI. It ultimately broke the site, and I had to roll back days’ worth of work, break it down into smaller chunks and try again. It was a good lesson in working within the constraints of the team you have, whether that’s an AI agent or a human.

Ownership. Historically I’d have focused primarily on validating skills and simply assumed ownership, but after leading multiple startup teams it’s clear how damaging a lack of it is. Beyond that, being AI-native is now a must-have, although it wouldn’t have appeared on resumes ten years ago!

This is a multi-faceted question to answer, and it depends upon the perspective of that someone. If the work I did lands with someone, then from their perspective it (the difference) would be the impact it made to their work i.e. removal of manual work, whether it’s gathering data to compile a report that shows the performance of a newly introduced product, or the simplification of a canvas. I rarely get feedback on the approach taken to deliver the work i.e. it’s the collaboration and asking the right questions to figure out what the problem is. And then the collaboration to design the best solution.

Of course they are right. Perception is the truth. I think of it like keeping the cooking and eating of the sausage separate from the making of them.

When approaching a problem, I can fossick around for a collection of undisputed facts, and then use an AI agent e.g. Co-pilot (within a set of guardrails) to quickly coalesce these into a set of available options for considerations. I am still needed to collect the facts, and consider the options (and discarding any hallucinations!).

I can’t help but think of the quote in Top Gun: Maverick “… it’s not what I am, it’s who I am.” Who I am is someone from the travel industry who came to appreciate the use of technology is a way to deliver it. Experience and knowledge. BUT it’s those who join recently that keeps you open to change, as they challenge the sacred cows. Adapt or die (another quote).

This is not an easy question to answer (in the conventional way). For vague problems, using Five Whys has always been a good starting point to really understand the problem and the root cause. Then you can start designing a (potential) solution. And if the solution is known to need iterations, then the approach espoused by Dr Alistair Cockburn is the one I would follow: probe the world. Put something out there and get feedback. You know it (the solution) is not perfect, but it will do - even if it causes some users to encounter a problem, or the system does not handle their expectations. The feedback you get will guide you (and your team) on the next (best) direction. Rinse and repeat. Conventionally (I guess) most people don’t see this (as a good approach).

Sometimes (or some in) the team can go “down rabbit holes” concerning features. Having an instinct that something is awry or not quite right comes down to the person, their experience (not necessarily in the solution at hand, but in the wider world), and their ability to communicate this within the team or the business. Guiding the conversation to return to focusing on value-to-the-customer rather than the minutiae of the feature is the most harmonious. Calling ELMO (Enough Let’s Move On) is a fallback I have used on occasions. Getting the team’s buy-in that getting imperfect, but valuable software solutions (that do not cause harm) into the hands of users earlier is a preferable approach, given the feedback that data will provide. This will then give you and the team the confidence on what to do next.

The hardest thing to show on my CV is probably the problem-solving behind the implementation. A bullet point might say that I built a platform with Next.js, Supabase and PostgreSQL, but in reality a lot of my work involved figuring out unclear requirements, understanding what users actually needed, and then connecting different parts of the system to make the workflow work reliably. That combination of business understanding and technical execution is difficult to fully capture in a CV.

AI helped me to build up the structure, check the input and output parameters, documentation drafts, code explanation, and explore possible fixes when I hit an error. What still needs me is to understand user’s need, considering better solution, review the output of AI, and deployed to the public.

I think currently the most valuable part is to understand why we do it and how AI does it. Like how the API works, how to authenticate, why we add human in the loop. But for some repetitive work, AI would take ownership on it. So I don’t think the goal is to compete with AI at writing every line of code. I think the goal is to build enough understanding to make good decisions, catch mistakes, and take responsibility for the final system. All the actions are to get a nice product done.

I usually look at two things: do I have the strong evidence to support my idea; what the cost it will be if we adopt a wrong decision. But in my early stage, I prefer to listen more and speak less, to understand why others would make this decision, what supports them, and frequently review, finally I would get their mind of thinking.

I wish someone had told me that I didn’t need to know everything before making the jump. I used to compare myself with people who had studied computing from the beginning, but I’ve realised that the gap closes much faster once you start building real things and solving real problems. And it’s AI era, you could bridge the gap quickly, the point is you have the passion and you are eager to learn new technology.

When I get a strong response from someone, it is usually after walking them through how a particular technology or process actually works and de-mystifying the subject. I enjoy cutting through the buzzwords and marketing hype and seeing the point where someone finally gets to the “why” behind an answer. In most business situations, people are given lists of requirements that often seem arbitrary, and explaining the ‘why’ helps prevent frustration.

I see it as a symptom of how technical skills are often taught — memorize this list of things, learn the expected answer, and move on to the next exam, quiz, or design review without necessarily learning how to reason through the subject. Giving people the tools to reason through their approach makes people happier and much more self reliant.

Walking people through the issue and explaining the train of thought behind a conclusion, not just handing off a directive or blunt answer to a question.

I would list it as mentorship and coaching. Most businesses claim to offer mentorship but I usually experience it as an initial few attempts to engage, then the pressure of schedules, deadlines and so on offer a convenient excuse to stop. What happens in the first few sessions, aside from socialization, is usually more statements of ‘what’ without the ‘why’.

AI is extremely useful for deep nuanced content search, and tuning written material for a given audience / reading level. Some agents are brilliant at learning repeatable presentation styles (themes, formatting, etc.) and allowing me to consistently format content for a particular medium. I have escaped the drudgery of tuning PowerPoint, and only have to author the content. I do not trust current LLMs to do critical analysis, make reasoned conclusions or pretty much do anything beyond operating as a Retrieval Augmented Generation (RAG) solution.

Almost every incident I have researched suggests a failure in basic Cyber Security controls, not some exotic LLM failure. When things go really wrong, someone usually failed on containment, permissions, trust delegation, or other basic concern.

In this age of near constant confabulation, I use a strict agentic reasoning framework with requirements for citations and reachable links, clear labeling in statements to classify statements as citations, synthesis, data source type (baseline knowledge, document, internet search etc.) and then walk through all the external citations. Sometimes I set up a task to review each citation, test the link, and get a second opinion on synthesis statements.

Sometimes the difference is something quantifiable, like a project getting back on track or a metric that changes in the right direction. But more often, it’s more about seeing a clear path to a decision and the ability to move forward from a place where they felt stuck. What I think I really do is help people think clearly in an inherently complex and unpredictable world when there is a lot of noise, and the signals are not clear. It’s the Product instinct in me to listen, take things apart, and then put them back together in a way that shows a way forward. These days, as a coach and advisor, that is usually what is happening when something ‘lands’, and it is a very satisfying and collaborative effort. Of course, the quantifiable wins are important, but the real impact is that change in thinking, and hopefully that keeps evolving even after my part in it is done.

They are probably right about what it feels like. But yes, it’s very likely the symptom rather than the cause. Underneath it is usually a mix of things. Sometimes people are stuck because they lack experience and not necessarily a lack of information. They may not have figured out how to find or filter the information, or they don’t quite have the judgement to assess risks. Sometimes it’s the politics or bias in their environment, with unspoken rules about who actually gets to decide things or what is safe to say. And sometimes it can simply be a personality or a working style of an individual who struggles with ambiguity or with taking action in the face of seemingly conflicting signals.

Ideally, most of that work of ‘change’ happens before we get in the room. A lot of it happens in one-on-one conversations and relationships beforehand, where I get to know their actual positions and concerns. I make sure I am very upfront about my own biases, which often gives others permission to be honest about theirs. It’s really about meeting people where they are in their thinking and then drawing out their points of view. People often say things to me they wouldn’t say to each other, and I think that’s less about my role and more about there being no hidden agenda in how I show up. Even in my career in large and small organisations, this came from being relationship-driven rather than from any authority I had, and that’s more true now.

Yikes! Most of it is really hard to get on a CV, and that’s a tough thing to admit on a page about improving CVs. For most of my career, the CV was not really the gateway to new roles. The roles came from building a reputation, the work that leaders and peers saw up close, and relationships built over years, which is really just about trust and integrity. That said, this is where quantifiable metrics become useful in a CV, and attributing a real number to your work is an important signal of the value you create. But an astute reader might still notice the unsaid things — for example, a career that revolves around the same people, or a progression of roles with fairly broad remits. Perhaps some AI tools could also discern those signals if trained! Speaking of AI, if I had one thing to say to graduates today, it would be to start building that reputation early, through whatever study and work you can do and start demonstrating the creativity and integrity that beats any CV. Since AI seems to be taking over most of the first CV filtering, that human connection will matter more than ever.

Yes, a career of skirmishing with technology debt made me realise that judgement and creativity debts were far more insidious in organisations. I saw it in numerous hubris-driven decisions made despite contrary evidence, and in the systemic compromises to good design and architecture to chase artificial deadlines and bad revenue. And now, the AI craze is exposing these debts quite vividly. A familiar example is a self-proclaimed technology company that attempted to cut engineering costs on the simple assertion that AI can write code. That decision, supported by the board down, was not really about AI, or even managing costs. It exposed judgement debt in a leadership team that buried itself in accounting spreadsheets and had never fully understood how engineering actually works, the architecture, the deep domain and systems knowledge, or the tradeoffs behind successful products. Within a couple of quarters, delivery slowed, the backlog doubled, customer complaints went up, and any chance of any ‘innovation’ was a pipe-dream. As expected, over the next several months the team was quietly rebuilt back to close to where it started, after eroding a lot of customer confidence. This pattern is unfortunately continuing to play out in numerous companies around the world and notably in Silicon Valley. And I see the same evidence of debt in things like the decisions to stop hiring interns and juniors, because ‘AI can do those jobs’. We forget that it’s how organisations renew and build the judgement and creativity of the next generation of leaders. Unfortunately, the bill for decisions like these usually doesn’t come due in the short term. It often takes many months or even a couple of years, in delivery slowdowns, employees quiet-quitting and customer attrition. And often this happens well after the people who enabled the debts have moved on to their next roles, and someone else is left to sort out the mess.

Surprisingly, it’s not whether an organisation has good product processes or how productive they look. I typically find some far more fundamental indicators — for example, I almost always start by asking how they are deciding what to prioritise and build. Quite often I get pointed to some months-old strategy deck, or because of a competitor, or because (insert eyeroll) ‘the CEO wants it’, or even worse, a customer offered to pay for it. That is really what I mean by theatre. A good product org will always talk about what customer problem they are trying to solve, why it matters in the market, and go from there. Other things I look for are their appetite for taking risks and making bets. Usually, a good product org consciously makes bets, experiments, and learns from failures. And as an extension of that, I love to observe meetings and see how the team tackles issues and makes decisions. A lot of what I call ‘theatre’ is head-nodding and a lack of decision and accountability. Good product organisations typically have a very different atmosphere of collaboration and action.

Hmm, I would push back a bit on the premise of what is being automated. What I think we are largely trying to automate with AI are the ways we have always done things and hence, the mistakes we already know about. I don’t think we are, or can, eliminate ‘mistakes’ as a category. There are still a lot of mistakes to be made and learned from, and with AI and the speed of change, there is no shortage of them. Developing craft, and the judgement and creativity that goes with it, is not really about just doing a job — or a safe job. It’s much more about the quality of thinking that comes from the willingness to take risks and engage with new problems. I would argue that none of that has changed with AI. If anything, it has become an imperative, albeit a scary one. So here are some, perhaps counterintuitive, suggestions:

• Look for roles in organisations and/or teams that are trying something new and where the problem is still being figured out but where ‘the job’ is not clearly defined.
• In interviews, ask what is still unknown or undecided about the role instead of what you will be doing.
• If you are working with AI, spend more time arguing with the responses and keep asking better follow-up questions. See how far you can go until the logic starts to break and the model starts contradicting itself.
• Most importantly, look for and spend time with people who reason through their decisions and ideas with real, tested evidence and openly acknowledge their own biases. That’s how you learn to question and develop your own thinking and craft.

I spent most of my career in consulting before moving into entrepreneurship. Clients often told me that I understood the real problem quickly and took ownership of it, rather than simply delivering what was written in the brief. But the clearest feedback was not what they said. It was what they did next. They extended the engagement, brought me into another problem or asked for me by name. For me, that is the strongest measure of client satisfaction: when someone faces another difficult problem and chooses to work with me again.

They are right, but those are the visible results. Underneath, what they were really trusting was my judgement and sense of ownership. I never treated a client’s problem as just a list of deliverables. I wanted to understand the outcome that mattered, where things could go wrong and whose support we needed to make it work. My technical background allowed me to go deep into the problem, while my broader experience helped me see the commercial and institutional picture. Clients knew I would be honest when something was not working and stay with the problem until we found a way forward.

At the BIS, I worked in a consensus-driven culture. People came from different central banks, each with its own mandate, concerns and appetite for risk. The strongest technical argument alone was rarely enough to move something forward. My role was to understand what was holding the discussion back, translate between the technical and institutional perspectives, and find enough common ground to reach a decision.

It mostly does not. A CV can list technical expertise, utilisation and delivery outcomes. It cannot easily show how I move an idea through an institution, especially when the people involved have different interests and none of them reports to me. That kind of influence can sound political when stated directly. But at a senior level, it is often what makes execution possible.

I have handed off much of the first pass: research, summarising, drafting and prototyping. What still needs me is taste, judgement and creativity. Without human direction, more output often just means more AI slop.

I do not always know. Being early can become part of your identity, and then every rejection becomes proof that you are a visionary. That is dangerous. I look for behaviour rather than enthusiasm. Are serious people already spending money or taking professional risks to solve the problem? Are users tolerating an obviously poor alternative? Are institutions beginning to move before they are willing to say publicly that their view has changed? Technology being possible is not enough. The conditions around it also need to be ready: the incentives, infrastructure, regulation, distribution and people with enough influence to move it forward.

When I was building, I trusted control and my own ability to solve the problem. As an investor, I cannot step in and do that for the founder. I have to trust how they respond when the plan meets reality. I am less impressed by certainty now. I trust founders who can change tactics without losing direction, admit what they do not know and keep the confidence of their team, customers and investors through uncertainty.

It depends on the context. I have been a mentor, coach, and instructor, and the feedback on those roles centers on how I set people in the right direction and how useful my course and guidance were. If it is about building the team and the software for founders, the feedback focuses on my ability to quickly find and solve problems, moving things toward delivery efficiently. One of our clients was impressed by our ability to vet and screen QAs with a very specific niche skill, we managed to find the three excellent talents they needed in 3 weeks.

They are usually describing how they felt valued and listened to, where I could actually help in a sense of belonging and being genuinely supported. I usually challenge the wants of the founders as they need a bit more of a guidance and what they want is different of what they need and the symptoms are different of the cause and the root, this also goes to the people that come to me looking for coaching and mentorship.

Volunteer in the Global Ethereum Hackathon in London. I would have volunteered for another community activity. I would say arriving on time 😂 One of my biggest flaws is poor time management, but also, if I hadn’t had the knowledge to help people coming to the volunteer desk, that would have been bad and I wouldn’t have felt like I was helping and making a difference and impact to the hackers. I studied Blockchain and Smart Contracts a long time ago and kept studying intermittently while building my startup. Nowadays, I focus more on business development, so I’m not completely up to date, but it’s still a subject that interests me.

This goes in my Volunteering section and also the skills go in the Skills section. I would also add the experience to the Experiences section with more details.

Some people could immediately communicate which part they didn’t understand, allowing me to address it quickly. For others, I could see they hadn’t understood because their delivered task was inaccurate, or they relied so heavily on AI that no critical thinking was applied. Usually depending on their seniority level, I can gauge the level of support and detail I need to provide. For the juniors I coached, I always required at least 3 feedback cycles where they had to learn how to learn and become critical-thinking, self-sufficient and consistent.

That is a really interesting question. Even recently, people are taking my course on Pact — contract tests — which is still useful, but the hype of contract tests passed and now AI is taking over everything. I used to use TestCafe as well which was a framework that people discommissioned early and was also in the hype. I have learned from early days back to my parents time to not get too attached to the framework, but to the base of what was built in and what problem they solve. Programming languages come and go, frameworks as well, usually the hype is driven by the market which is driven by the developers. Stripe is a really good example, if you make easier and frictionless more people are going to use — here the users are the developers and they will be the ones choosing and doing benchmarks to choose the frameworks to be used. I always recommend to do a benchmark for new projects anyway as this changes overtime, tech is really fast about finding new frameworks and ways, but still takes a bit of time for a framework to pick up as the softwares that were built with previous frameworks are already running and it is an extra cost to do the migration, plus the fact that even a developer entering a new project that should be doing a benchmark to check what are the new most recent options, probably will skip it and use something familiar already. So the adoption of new tools, frameworks take a bit more time because of this matter. I am not going to deny, I like to know a base and a bit of everything, hard to predict the future, because again my background with my parents and experiencing all the tech cycles from an early age. I go deeper into things whose real usage and benefit I understand, but that understanding is only for that moment and can change over time. So, I always check what are the niche tools and frameworks that people are slowly starting to use, for example the last one was Mabl for Test Automation.