We are in the phase of a new wave of technology with AI where most of the discussion centers on the technical: security and governance, model capability, data management, process design, agentic AI management, and cost sensitivity. This is normal in the early stages of a new technology disruption. The first questions are always: how can we technically take advantage of this given our current estate and state of play? But as the dust settles and as the ‘wow’ factor dies down, I believe that we will enter the next, more significant, phase: change management and organization re-engineering.
I recently commented on a diginomica piece on LinkedIn, focused on AI token costs, saying:
I think there’s also a point that’s sometimes overlooked in these discussions: changing behaviors, business models and ways of operating. When SaaS came along there was also a lot of talk then about pricing concerns because of the switch from CAPEX to OPEX – which only died down when people understood that SaaS enabled things like continuous engagement with customers/employees, different ways of delivering products and services, as well as engineering organizations in a way that allowed them to take advantage of continuous upgrades.
I think at the moment so much of what I see with enterprise AI is still the technology piece being slapped on, without a recognition that AI will change organizations in ways we often don’t yet understand.
The core of my point is that whilst AI is clearly upending a lot and that there are valid concerns around costs and value, I think that to really take advantage of its capabilities we need to think beyond the technology itself and focus attention on the fundamental changes that organizations and people are going to achieve when businesses learn how to operate differently because of AI. That’s not to say that there haven’t been extensive discussions around threats to jobs and skills – of course there have. In fact many companies are already arbitrarily slashing jobs because of so-called ‘AI efficiencies’ (reader: I don’t think anyone would say cloud computing made things substantially cheaper, rather it changed the model to capture more revenue).
I think given the wealth of experience we have gained in recent technology shifts (cloud, mobile, social, etc), we should and could be having a more sophisticated discussion about how we think about people, work and organizational change, as more AI tools become accessible. Ultimately, no-one is going to receive the desired benefits of AI if we continue to put lipstick on a pig and hope for the best. Not only that but we have seen a backlash from consumers against companies that don’t thoughtfully consider their use of AI and ignore their most valuable asset: people.
With this in mind, it was with great pleasure that I got to sit down and discuss AI change management with Celonis’ Kerry Brown, who is the vendor’s Transformation Evangelist. Brown is more than qualified to speak on this topic, given her extensive experience in change management across multiple companies and industries, including but not limited to: Baker Hughes, Coca-Cola Enterprises and SAP (where she also ran point with the user groups).
Opening our discussion, Brown said:
I very much think that if the goal of technology as a catalyst is to change the way we work, those outcomes come when people do their jobs differently. My litmus test for anything is: how do people’s jobs change?
Quite right. And much of what Brown argues about how we should be thinking about this when it comes to AI centers around the idea of: allowing people to openly share their success, without fear of eliminating their current ‘job’. And asking, what is your special sauce? The latter may sound twee to some, but I think it really captures the human value-add that many people allude to in discussions around AI – and Brown has plenty of experience and examples to back it up.
Plus ça change
When ChatGPT first broke onto the scene nearly four years ago, it was hard to deny the ‘magic’ feeling that came with its user experience. It felt like talking to a computer, in natural language. It’s interesting how, before ChatGPT, the Turing Test — the idea that a machine counts as intelligent if it can converse indistinguishably from a person — was widely held up as the benchmark for artificial general intelligence. That benchmark has largely been set aside now, arguably because we’ve passed it. In other words, humans have looked at ChatGPT and said: ‘Okay, we can talk to you naturally, but we are still special and we know that fundamentally’. Whether or not that’s naive remains to be seen, but it speaks to the fear many have when it comes to the adoption of AI at work.
The speed of change with AI and this ‘magic experience’ is driving a lot of the psychological anxiety amongst users at work. As Brown said:
People are people, and the way people react to change is pretty consistent. A lot of the truisms I’ve developed over the years are still true. For example, an equation I came up with years ago: change management equals expectations plus accountabilities, and surprises in either of those – whether you’re at CEO level or an individual contributor – is what leads to distress and anxiety.
I think it’s fair to say that there are plenty of surprises in expectations and accountabilities with AI, if we are assessing change management according to this equation. Brown added:
The magical thinking of this – the magical speed of what it does – means it can be done much faster. It gives an individual or a team the opportunity to digest and do something radically different, quickly. That unpredictability amplifies the stress and the discomfort, but it also changes the change curve.
In other words, the speed and democratization of what you can do with AI throws everything into a bit of chaos. With cloud computing we did have marketing and sales teams adopting tools on the sly and IT departments having to play catch up, but by and large these were manageable change programs that eventually saw everyone working towards the same outcome. Equally, cloud and mobile allowed people to carry out process work more efficiently, wherever they were. However, the process was still key.
With AI, there’s shadow AI, but there’s also this sense of not knowing where the line is and a fear around disclosing whether or not to bring results to the fore. Questions over governance and accountability are there, of course, but there’s also fear around: if I’ve figured out how to do huge chunks of my job with AI, what does that mean for my job? Will I be praised or will I be ousted? Brown added:
What I see is – back to expectations and accountabilities – if I feel under threat, my likelihood of wanting to tell you exactly what magic I’ve done goes down. Have I done my job away? Will I make myself redundant, or obsolete? Will my coworker make me look obsolete? There are a lot of trust issues, including who owns the win.
AI permeates outside of the organizational structure and boundaries – it gives you the ability to reach into different places. So when you look at “does it affect my job, does it affect our jobs, how does that come to be,” the trust aspect is one part.
From senior leaders: what do I expect from you, what’s going to be different, how is it going to be different, what do I want to see happen, how are we going to measure it?
Brown gave a comparison that I think is probably the most useful one I’ve heard so far, talking about the building of trust needed when everyone suddenly started working from home during COVID-19 lockdowns. There were, in hindsight, wildly inappropriate attempts by organizations to monitor and ensure employees were at their desks at all hours of the day. When, really, what organizations should have been asking themselves is: if my employees’ outcomes are the same with reduced commute time and more flexibility during the day, what does this mean for how we operate? And, what is being lost and gained here?
Brown said:
Now remote work is completely normal; people know how to measure it and see that the right things are happening. We have a much more fluid way of working — people get things done at different times, the cadence of the workday has changed — but remote work isn’t an untrusted “is everyone cheating?” anymore.
Both groups have confidence that work is getting done and that recognition can happen. That’s the moment of discomfort I feel most parallel to now, and that was about building trust. The same trust applies now: I’m investing in this — are you using it? I can measure if you use it, because I can measure your token usage. Am I using it enough, not enough? If I’m not using it, am I going to get in trouble? That same trust-building around what the new patterns are hasn’t been normalized yet.
Building trust in change
To Brown’s point, there has to be a level of willingness in organizations to experiment openly. The lack of trust in AI is understandable – we’ve all seen hallucinations or false outcomes – but too often there’s a view that one mistake should mean throwing the baby out with the bathwater. Companies love to proclaim that they’re open to new ideas or new approaches, whilst being quick to judge when things aren’t perfect the first time around.
Yes, the stakes are high with AI. Putting something out publicly generated by AI with a whole host of nonsense on it can set a company back, or using AI in a way that compromises data sharing principles could land you in legal trouble – but that doesn’t mean you shouldn’t be testing it out in ways that have lower consequences. Equally, employees and leaders alike should be allowed to fail in the open and bring results to the table, without fearing their job will be at stake or that they may be dealing with fragile egos on the other side of their experimentation. If humans really are special, then they should not fear new ways of doing things. And this is what Brown is keen to get across. She said:
If I look at the human piece of it, I think about myself. I know we’re looking at how to use AI in jobs, and I think: what can I do more of, what can I do less of? What matters is what I bring that is uniquely me. If I’m an optimist, I think where we really matter – where we make a difference – becomes our special sauce, our superpower.
She goes on to provide some advice:
The guy who ran strategy for Coke was my mentor way back, and he said: ‘whatever the three things you’re already best at, become exceptional at them; whatever you’re bad at, make sure they don’t bite you – but don’t waste time becoming merely good at what you’re bad at’.
Maybe this is a place where AI can get me better at the things I don’t want to do or am not best at, so I can become exceptional where my special sauce and superpower are. That allows people to exercise their agency and their gifts, rather than feeling replaceable.
This is where Brown and I are wholly aligned. I’ll use myself as an example. There is a world in which I look at AI and I think ‘great, I can get it to write for me all day every day and hopefully something will stick’. But that’s short sighted, of course, as it takes away the value of what I really do: which is talk to people, dissect claims, attend events, discern valuable information from less valuable information, connect people, and bring trusted judgement to an audience. If I was to just let AI take over all my writing, then where is my value?
Instead, I’ve been experimenting with AI to connect data sources together, to transcribe interviews and clean up transcripts, to research, to test headlines, to challenge my writing for new ideas, and to help me prepare for interviews. A lot of these tasks were manual before, were time consuming and meant I spent less time on the things that matter above. Now, I can probably do twice as much of the valuable stuff, more deeply, because I’ve got AI supporting me with tasks that were just grunt work.
And this is where Brown argues the discussion with AI and change management needs to move towards. It needs to go beyond ‘Oh, AI is stripping me of my work’ and head in the direction of ‘Now AI is doing those tasks, what else can we do now that’s unique to us?’. She said:
When you go home, look around at your peers, and if they’re still working while I’m done at three and they’re going to be at it until six, how do we reshape and reform those jobs? What’s happening now with AI is that your job may have shrunk by 13%, mine may have shrunk by 37% in terms of the tasks I can do differently – but what do those tasks also generate in terms of new tasks? We don’t yet know what can go away. And this goes back to fear and trust: do I tell you that I’ve managed to figure out how to make a whole bunch of my job go away?
Trying to do more with the same
Brown recently attended a Gartner event, where she saw a tech company with limited funding, on a freemium model, discuss how they’d used AI to improve their customer satisfaction ratio and convert customers to a paid model. She said:
They couldn’t add headcount, so they used AI to create a better-answer capability on their website – and it turned out to be better than a lot of the new people they were hiring, who were trying to drink through the firehose. Their ability to answer customer questions got better, and their ability to convert from free to fee got better.
So people could do the human things, versus bringing in a new person, dipping them in all this knowledge, and having them answer those questions. They actually grew their business by putting AI in and creating capacity they couldn’t have created with headcount.
Brown said that there’s often an assumption – or even worse, an expectation – that AI will allow a business to do more with less. When, perhaps, what a company should be thinking is: how can we use AI to do more with the same? She added:
That was a really clear place where people were encouraged to do something that didn’t threaten them. There’s an intersection point of “what do I want to do, what do I share,” and the sweet spot is where doing good is rewarded, doesn’t embarrass or threaten me or my peers, and is measurable in a way that gets repeated and fostered.
I don’t think large organizations necessarily have the nimbleness to adjust rapidly.What can be done differently, what do we do when this group has 13% capacity? Culturally that comes back to the organizations saying “we’re going to cut to 10% of our workforce” — well, why 10? Should it be 11, or seven?
Should it be 32? What exactly should it be? That comes back to the intersection of trust and capability measurement — and the measurement is only going to get better. And then the leadership becomes the differentiator.
None of this means every task displaced by AI will be replaced by more meaningful work, or that every organization will choose to reinvest the capacity it creates. Employees are rational to be cautious when some companies are announcing workforce reductions before the promised productivity gains have even been demonstrated. “Do more with the same” therefore has to be a leadership commitment, not merely an optimistic interpretation of what AI might enable.
My take
What I enjoyed about this conversation with Brown was how she approaches her thinking with regards to AI change management – it’s neither carrot nor stick, but rather a mature, sophisticated approach that incorporates experimentation, capturing human value, reducing shame and fear, as well as fostering an environment that allows people to do more of what they’re good at. Rather than seeing AI as a tool to replace people towards a very average median. As I wrote in my piece earlier this year on the importance of friction, if every organization is an agentic enterprise with all frictionless experiences: how do you build loyalty and how do you differentiate? I’d argue you are on a fast track to being very mediocre.
And for employees that are feeling apprehensive or cautious about AI – either through fear of trying, or fear of exposing what they’re capable of doing now – this also helps frame the thinking. What are you really good at? And does AI give me the chance to become exceptional at those things? As Brown said:
If you can identify what makes you special, what makes you different, then the confidence to amplify that with AI becomes more acceptable.
The challenge for leaders is to create an organization in which people feel safe enough to reveal that value – and where becoming more productive doesn’t simply make them more expendable.