Whether they run large corporations or mid-cap businesses in Europe, sit inside private equity firms or portfolio companies, busy leaders are asking the same thing about AI: how do I make sure nobody wastes the time I don’t have?
That question resonates with me! In the las twenty-five years I have been on the receiving end of a lot of recommendations from consultants, however the true value was always in implementation (this is also where most of the disconnects are). Since all my roles have been on the front lines, running organisations, units, departments, I have learned that what lasts is not a single solution or a set of use cases. What survives the operational reality across a longer time frame is a connected set of activities (i.e. a system) that are deeply ingrained in how organisations run. It has its checks and balances, executives receive clarity on performance, clear metrics, employees understand the ‘what’ and ‘how’, and after some time, if done properly…it (almost) runs itself.
On AI transformations, the question worth answering is not “what can AI do?” but “how do you build something that lasts?”
Why Most AI Transformations Fail
Firsthand observations. Don’t these sound familiar?
It is fascinating that if we start by observing the baseline scenarios prior to defining the AI transformation, the following appears:
- Leadership alignment. The industry requires the C-suite to talk about AI, however, there was little time spent together among the senior leadership team for a coordinated, facilitated, properly led discussion about it.
- Scattered pilots, no system. Employees have ideas and vendors flood senior managers with demos and use cases, but the value is not assessed and rarely compounds.
- AI treated as a technology rollout (sitting with CTOs). The tools arrive; however, the organisation is not ready or managed in a way to use them. Rarely does the operating model change to reflect this powerful set of tools. Old habits, such as requesting new tools, quietly reassert themselves, challenging any ROI business case assumptions.
- Use cases collected, not selected. Ideas pile up with no business, data, adoption, benefit filter, so nothing connects to EBITDA and ROI can’t be proven.
- No baseline. Without an honest starting point and a pause for a (short!) reflection, it is difficult to set an ambition level (status quo, and where we want to be in 12 months).
- Shadow AI penetrates the organisation. Lack of approved tools and clear communication of how employees may use AI means that people reach for their own tools (often unapproved, no guardrails), while the more cautious & risk averse employees wait, increasing the AI resistance.
- Capability left to the curious few. A handful move fast, get impatient, but the rest of the organisation never catches up.
I am convinced you recognise your organisation in some of these observations. What is important, is that none of these is a technology problem. These are leadership and operating model problems that reiterate why so many organisations, having spent millions, still don’t see a lasting impact, even though they might have deployed an impressive use case…or two.
AI Transformation Is an Operating Model Challenge, not a Technology Programme.
Start with AI Maturity, not the use cases
The usual advice is “start with your strategy.” True, but strategy can sound so abstract. I prefer something simpler: a frank, data backed conversation about where the organisation actually is, and is not. After reviewing the results (data driven; here D. Kahneman’s System 2 comes into play) comes a conscious decision about the ambition level: what to do, and more importantly, what not to do. How do we want to position ourselves? Do we want to become a ‘leading AI house’? How aggressive do we want to be? Are we regulated, or not? What does the competitive landscape look like? Where do we choose to play, and how do we intend to win?
The instrument I use for that conversation is our own (Symmetria Partners) maturity model. It deliberately looks beyond data and infrastructure at dimensions such as leadership, employee engagement/resistance, what is on the strategic agenda, the legal-entity boards, and how consistent the reporting is. From there it considers how much autonomy the organisation is ready to give its AI, calibrated to appetite, progress and security readiness. In one regulated, data-driven business we moved AI maturity by over 44% in four months, a measurable increase we could point to precisely because we had measured the starting point.
We begin with a maturity assessment and the leadership engagement around it, and from there we set an ambition level with honest timeframes. It is worth saying that most companies that started their AI journey twelve to twenty-four months ago are still at a low level of maturity. Naming where you really are is the precondition for deciding where you want to go.
Building an AI Operating Model That Lasts
Around that directional ambition, sits the rest of the system. The elements that turn the defined ambition level into something that keeps running after the AI Transformation programme:
- A scored use-case pipeline. Use cases are selected, not collected (again, a conscious decision about what is approved for a pilot and what is not), ranked on multiple criteria such as business value, scalability, risk, feasibility, data readiness and adoption. In the programme this turned 40+ ideas into a shortlist of only 3 pilots carrying over €1M in addressable EBITDA. This is a great scorecard that makes the review and approval of high potential use cases clear and organized.
- Tools and usage best practice. A shared, sanctioned toolset and clear guidance, so people build a common standards. This is absolutely needed to reduce/eliminate shadow AI usage, which is high data/information security risk.
- A sandbox for proof of concept. A safe space to test ideas quickly and cheaply before committing, without putting the business/IT infrastructure at risk.
- An AI ambassador community. People inside the business who carry the agenda and keep momentum once the external support has gone. These roles are diverse, from sharing what is happening on AI front, to increasing adoption, reducing fear.
- An internal communications cascade. Years ago, my boss told me to say everything at least six time!. AI comms needs to be tuned to the organisation’s own rhythm and culture, ideally using the existing format to blend in, and not to create an AI communication industry.
- Reporting that management and supervisory boards will recognise. Concise one-pagers, a maturity view, ROI tracker, capability development, risk profile. Teams want progress, and investors want visibility into the returns. The one-pager metrics serve both (I love it!).
- Handover and ownership. The operating model passes to the company’s own team, so capability and value don’t leave with the engagement. Having spent over 25 years in the trenches, this is the difference between a project and an asset.
The rest of the system
Some of you may say…easy, logical, nothing new here. Indeed, none of these elements are remarkable on its own. The value is in the connections between them. It is the cause-and-effect relationship among them (i.e. a system of connected activities) that creates a lasting impact. A use case produces a demo, but a steady pipeline of high value (and assessed) ideas year over year can only be delivered from an operating system. And that produces value that lasts long enough to show up in valuation. In a PE portfolio company, that distinction is the whole game.
Which brings me back to the car, and the line that started this. The reason to build the system rather than the great use case is simple: it is the only way to make sure that, in the rush to “do AI”, nobody wastes the time you don’t have.
