Most businesses confuse tool access with transformation. Real AI adoption starts with a clear end state and a blank piece of paper, not a Copilot rollout.
What you will learn:
The article – Don’t let AI happen to you with Ed de Minckwitz
John Emmerson: Today I’m here with Ed de Minckwitz, Director of Public Policy UK at ServiceNow and Senior Policy Fellow at Policy Exchange. You’ve spoken about AI as an organisational transformation engine. What does that actually look like to you when you get past the strategy and into legacy systems, resistance and competing priorities?
Ed de Minckwitz: I think people have very different concepts of what AI adoption actually looks like. At one end of the spectrum, it’s giving people Copilot and hoping that your business becomes more productive and profitable overnight. At the other end, it’s sitting down with a blank piece of paper, thinking: how would I totally reinvent my business with these new tools that are available and continuing to improve? Within that spectrum, there are realities to be contended with: legacy tech, human nature, and actually some existential questions about what a business actually is and how we all want to work. AI is this huge transformational engine at all layers of the economy. But I feel that as a society, as business leaders, as governments, we’re not fully grappling with either the full extent of the opportunities on offer or some of the challenges and questions that underpin those.
John Emmerson: What do you think is stopping that?
Ed de Minckwitz: Human nature is a part of it. We are creatures of habit and we’re also enterprising creatures — we want to work, we want to endeavour, we want to feel the satisfaction of having done a thing, created a thing, offered a service. And perhaps business leaders in the abstract talk about automation and efficiencies, but by nature not everyone goes to work thinking: how could I do much less of this all the time? Also, particularly in small and medium-sized enterprises, most people are flying the plane, not thinking about how to re-engineer and build the next generation of aircraft. I see that in government all the time. It’s not that people don’t come up with ideas when you engage them on how things could be done differently — it’s just that the day job is so consuming, they don’t have the bandwidth, the resource, the senior cover, or indeed sometimes the space to say: we’re going to stop how we did it before and start doing it in another way. And that transition can be a barrier as well.
John Emmerson: The British Chambers of Commerce published research recently — about 56 per cent of SMEs have adopted AI, but when you drill into that, 96 per cent of that 56 per cent defined adopting AI as using ChatGPT, Gemini or Claude. Only around four to six per cent were looking at deep organisational AI transformation projects or agentic workflows. Is that what you’re seeing?
Ed de Minckwitz: Absolutely. And the pace of change is really challenging. In my think tank work, I find it fairly easy to write in the abstract about the pace of change — I quote Dario Amodei and the compressed century, I try to be conservative about some of the timelines. I noticed Kanishka Narayan, the minister, in an interview saying that artificial general intelligence could be with us within five years. I’m always trying to check myself on whether it’s closer to Sam Altman’s accelerated timeline or Yan LeCun’s more measured 20 or 30 years. Either one of those timescales is dramatically rapid in the context of historical technological transformation. We really are in the foothills. And just in the two years I’ve been at ServiceNow and the two years before that talking and learning about AI, the rate at which we’re seeing transformation is extraordinary.
Ed de Minckwitz: And that pace is itself a blocker. When do you press go? When do you say: I’m going to adopt today’s technology, versus thinking: hang on, I’ve just seen something new, and it surely won’t be long before there is an enterprise-grade, safe, secure version of that. How do you adopt technology in a way that means you can iterate with it and not find yourself with a new type of legacy tech problem in short order?
John Emmerson: You described the civil service previously as a primitive large language model. Do you think mid-market organisations show up differently?
Ed de Minckwitz: I said that partly to try and excuse civil servant behaviour sometimes. Governments get frustrated with the civil service without recognising that the civil servants are the prompt engineers — they’re responsible for the data they put in and the tasks they set. There are similarities in the huge volumes of text moving through layers of interpretations until an answer emerges. But I think in mid-sized organisations, the problem is more fragmentation. You’ve got knowledge sitting in spreadsheets, inboxes, individuals. And I think you’ve got more agency, one would hope, in each of those individuals. The thing is that in both civil servants and private business, we are in some ways conditioned to behave like the large language model. And actually, you can break out of that and express your agency a bit more.
John Emmerson: Where should accountability for AI adoption sit in an organisation?
Ed de Minckwitz: I think it has to have a top-down element. Everyone in a leadership role has a responsibility to be asking: how could these tools be applied to what we’re trying to do? We are beginning to see firms emerge that are disrupting sectors and forcing that top-down pressure. But I also think that everyone has a responsibility to make themselves more productive. And anecdotally — and I confess I’m a little guilty of this myself — I think a lot of us who are AI-native have worked out quite how much time and effort AI can save us. But how many of us have honestly turned around to our seniors and said: I’m at least 40 per cent more productive because of these tools. Could you give me 40 per cent more work? Personally I just crack on — it’s sort of my nature. But my business hasn’t yet turned around to everyone and gone: why don’t you all do 20 per cent more, because we’ve given you these tools? But I think I’m beginning to see that pressure put on people to say: could we do this function with the same amount of people, but do 150 per cent more? That’s coming, I think, this year.
John Emmerson: I’ve spoken to organisations doing exactly that — keeping headcount at 150 but completely transforming how the business works to ten-times the output. Do you think that will identify the winners versus the losers?
Ed de Minckwitz: Totally agree. We will change our habits. We will recognise that some of the mundane tasks can be done for us, that that means we can spend more time doing the more interesting things and that we can get things done a lot quicker. But there are still bottlenecks — things being sped up but hitting bottlenecks, whether in infrastructure or in colleagues who are not yet as fast. We’ve definitely seen businesses streak ahead. I’ve heard people in the software world in the last three months say it’s not even 10x — it’s 100x, 1000x in terms of the amount of code that can be generated. That will just begin to come through the economy. But there’s a jagged edge as well — it’s not uniform. Some people will be resistant. Some people quite like the way they work. And so businesses have a choice: to what extent do you say everyone’s got to get on board, these new tools are going to speed us up? Because the tools themselves are quite intuitive. The skill is changing the way of working, changing your own practices — as much as it is learning how to use the tools.
John Emmerson: If you had to write a one-page playbook for a senior leader on how to use AI well this year, what would you include?
Ed de Minckwitz: At the top: start with the end state. What’s missing sometimes with AI is clarity about where you want to get to. What do you hope this technology can achieve for you? Try and have that clear direction of travel. Then: carve out time for all staff to spend learning and understanding the technology. I don’t think that time is wasted at all. Something we’re really encouraged to do at ServiceNow is play with it — ask it what should this look like, what else could be done, because it’s not until you can see what it can do that you can work out the end state. Then: try and put some soft metrics to it. Start challenging people: how much more could you do with the same amount of people? We’ve all had access to it now, it’s become ubiquitous, but are we actually challenging people to use that?
Ed de Minckwitz: Then: be relentless in looking around you and seeing what other people are doing. Because of the pace of change, someone can come up on your blind side very quickly to disrupt your sector. You’ve got to have your eyes about you all the time. Next: do a real audit of those mundane functions. What is your time actually being spent on? What are the value-driving hours in your day and in your staff’s day versus the ones that are just consuming cost? And finally: there’s got to be a resilience piece. This is preparation for disruption. Be aware that there are going to be leaps — new models, new technologies, new use cases that are going to disrupt all the time. So that resilience and agility, and the mindset of resilience and agility, is really important as well.
John Emmerson: Final question: what is the one thing you would advise a leader in a mid-market organisation to do right now?
Ed de Minckwitz: Carve time out to experiment with the tools. Just ask the tools what’s available, play with them, really try and stretch what they can do. Because it’s in that frontier innovation — and by frontier I don’t mean building large language models, I mean pushing the boundaries of this new technology in your sector — that the value is found. You know your business and your sector better than anyone else. You are therefore the best placed to work out how to bend these tools to those use cases. That’s how we’re all going to innovate together: keep experimenting, keep playing with it.
Ed de Minckwitz describes a spectrum of AI adoption running from one end — handing people Copilot and hoping for productivity gains — to the other end — sitting down with a blank piece of paper and asking how you would completely reinvent your business with the tools now available. The blank page end of the spectrum is where real transformation happens. Most organisations are nowhere near it. The vast majority are at the Copilot end: tool access without workflow redesign, generic capability without organisational change. The BCC research he cites supports this: around 96 per cent of SMEs that describe themselves as AI adopters are using off-the-shelf tools like ChatGPT for content and knowledge work. Only around four to six per cent have moved into bespoke, deep AI workflows.
Ed de Minckwitz identifies a genuine tension that many mid-market leaders face: committing to today’s technology knowing that better tools are months away. If you adopt now, you risk creating a new form of legacy tech in short order. If you wait, competitors may pull ahead. His position is that this tension is real and cannot be resolved by waiting — it can only be managed by building agility into your operating model so that you can iterate as technology improves, rather than treating any single platform as a permanent solution. The organisations making progress are the ones that have made peace with this: they start with a use case, prove value, and keep the architecture flexible enough to swap in improvements as they come.
Ed de Minckwitz makes a candid admission: many AI-native workers have already become significantly more productive — he estimates 40 per cent in his own case — but most have not told their managers or pushed for more work. Most businesses have not yet turned around and asked why people are not doing proportionally more. He frames this as competitive pressure that is coming regardless: the firms that start asking that question and restructuring expectations around what a team can achieve with AI will pull ahead of those that treat AI productivity gains as individual working habits with no business capture. The implication for mid-market leaders is direct: the productivity is already there in your organisation. The question is whether your management structure is designed to see it and use it.
Ed de Minckwitz argues that the skill is not learning to use the tools — the tools themselves are quite intuitive. The skill is changing the way you work. Most AI rollouts focus on access and training: getting people through a Copilot induction, explaining what the tool can do. This solves a knowledge problem but not a behaviour problem. Behaviour change requires people to redesign their daily workflows around what the tool now makes possible, and that requires time, permission and usually structured space to experiment. Organisations that carve out protected time for people to genuinely play with and redesign their work around AI are the ones where adoption actually sticks. Those that treat it as an IT project with a training module at the end are the ones where adoption plateaus.
His framework covers six elements. Start with the end state: define where you want to get to before you pick a tool or a vendor. Carve out time for staff to experiment and learn — this is not wasted time. Put soft metrics on ROI: challenge teams to quantify what they can do with AI that they could not do before. Stay relentless about watching competitors: at this pace of change, disruption can come from a blind side very quickly. Audit mundane functions: identify where time is being spent on work that AI could do, and redirect that capacity. And build resilience: assume there will be further model releases and capability leaps, and design your operating model to absorb those without starting from scratch each time.
Ed de Minckwitz is direct that the competitive pressure is already real and will intensify. He cites examples from the software world where AI-enabled teams are producing at 100x or 1000x the output of teams working in traditional ways. While this is most visible in software and coding, the same dynamic will move through every sector. The organisations that pull ahead will be those asking: could we do this function with the same number of people but do 150 per cent more? That question is coming whether or not businesses choose to ask it themselves. The difference between asking it proactively and having it asked by competitive pressure is the window of time you have to adapt on your own terms rather than being forced to react.
Ed de Minckwitz is broadly supportive of government investment in AI and of the clarity shown by the UK government on the opportunity. But he is critical of passivity: putting large pots of money up for grabs is not the same as actively understanding and governing the technology. His view is that government needs more activism — understanding the impacts of AI more deeply so it can steward the transition better, particularly for the people whose work is being disrupted. He draws a parallel to the responsibility that business leaders also have: the disruption will be real and sometimes traumatic for individuals, and both government and businesses have a duty to manage people through it rather than simply accelerating the change and leaving people to adapt alone.
Andy Lambert co-founded ContentCal, sold it to Adobe, and spent four years there as it became part of Adobe Express. He saw teams split between...
Watch video
Andy Lambert is a co-founder of AcademyAI, which measures and builds workforce AI capability, with a focus on...
Read more
Across 22 interviews with executives since we started the aibl newsletter, one question has kept surfacing. What...
Read moreGet ahead with the most actionable insights, playbooks and real-world AI use cases you can adopt right now, in your inbox every week
