It's been a little longer this time since my last edition. But as I have said all along, Connecting.The.Dots is about speaking when there is something new to say — and these past weeks I have been grooming a thought that won't leave me alone: the new speed of transformation is becoming the normal speed, and we need to adjust to it.
I was recently at the ABSL Summit here in Poland, and I was truly delighted to hear so many conversations moving back to Data foundations before AI — a topic I have been advocating strongly over the past year. You can feel that the realisation is coming from painful, frustrating outcomes. We are all feeling them, observing them, and trying to root-cause our way through what Gartner's hype-cycle methodology calls the trough of disillusionment.
This past week I also heard a statement — provocative, delivered on a big stage to a big audience — that has stayed with me: "A fool with a tool is still a fool." It resonated tremendously in the era of AI. We'll come back to it later.
And I have valued some thoughtful in-person interactions these past weeks, with genuinely good conversations about connecting the dots. I love that these exchanges are happening beyond the newsletter. That is where the real impact shows up.
Now, the main topic of this edition is triggered by a question I keep getting. Most leaders I meet agree with my view on building the Data foundations before AI. That Bad and Ugly Data gets amplified at machine speed is real, and we all agree on it. But it is fair to push back and ask:
"Mauro, who has four or five years to build foundations while AI is changing every week and every month?"
That is exactly why we are here — to take a question like that and dig into it properly.
It is the right question. And for a while I did not have a clean answer to it. The honest reflection that followed is this edition. It is not a retreat — not by an inch — from the "Data before AI" principle. I have heard people argue recently that the foundations are the Processes, the operating models — all fair — but at the very core, the first foundation is the "food" AI runs on. And AI's food is Data. So let me reaffirm this with my strongest conviction. What follows is not a softening of the principle; it is the refinement that finally makes it actionable. And it hit me hard these past days. I call it the Acceleration Paradox.
The paradox is this. The force that makes "four or five years of foundations" feel impossible — the exponential speed of AI — is the very same force that can compress those years — turning a five-year foundation into one or two. The thing that stole your time is the thing that gives it back. You do not resolve the tension by skipping the foundations. You resolve it by building them with AI, not just running AI on top of them. You still move fastest on solid ground; AI just makes the ground faster to lay.
To see why that matters, we first have to look at the sequence almost everyone is using. Because the failure we are watching at scale is not a technology failure. In my honest opinion — and because I observe it in reality — it is a sequencing failure.
The Broken Sequence
Watch how a typical AI initiative actually unfolds, and you will see a consistent order:
Business Problem → Technology → Data → Process.
We start well — with a real business problem. Then we do something that feels productive and is actually fatal: we jump straight to the technology options that will solve it. We run the demos. We score the scenarios and options. We fall in love with a platform — or platforms — enabled by AI (which is now 120% of all cases). Everything is technically live within weeks, we celebrate, and then reality hits. The use case is limited. Scalability isn't there. Costs are too high. Business buy-in will be impossible. ROI is too low. We stop — or, worse, we live in denial, which really means we keep going when we should stop.
Only then do we turn to the Data and the Process the platform needs to function — and we hit headwinds. The Data is fragmented across systems. Definitions conflict. Processes differ by market. Master Data is inconsistent. The Data Hedges, left untrimmed, are everywhere. By the time we have wrestled with the platform and the Data complexity together, the budget and the patience are gone. So the Process — the thing that probably needed redesigning in the first place — is left untouched, or marginally adjusted to fit the tool we already bought and locked ourselves into.
The outcome is painful to watch. We have layered new technology onto an unchanged Process, fed by Data that was never ready. In Edition #12 I borrowed Hammer and Champy's warning: adding AI to a broken Process is not transformation, it is faster failure. The Broken Sequence is the mechanism of that failure. It is how good teams, with real budgets and real talent, end up with pilots that never reach production.
Reversing the Sequence
We need to reverse it. And we need to do it fast, despite the established status quo.
Business Problem → Desired Outcome → Data → Process* → Technology
*Depending on the business problem, Process and Data may swap order — Process first, Data second. But these two are, without exception, the first steps in the sequence. Technology, without any doubt, is the last.
Start with the business problem, as before. But before reaching for a single tool, define the desired outcome — then visit the Data. Apply the Good, Bad and Ugly framework I have shared before. What Data does this business problem actually involve? Where does it live, where is it created, where does it break? You are mapping the cipher key before you try to read the message.
Then design the Process. Not the PowerPoint version — the real one, and then the one you would build from a White Canvas if nothing constrained you. This is the Perfect Process: designed around the outcome, not the org chart, not the tool.
Only now do you choose technology. And here is the quiet advantage almost no one talks about: when you arrive at the technology question already understanding your Data and your ideal Process, you learn within days — not months — whether the problem is solvable with one tool, several, or none yet. The technology decision becomes fast and obvious, because you are no longer asking a vendor to define your problem for you. This is the Virtuoso Dynamic Model in operational order: Data quality and Process excellence first, technology in service of both.
The reversed sequence feels slower at the start. It is dramatically faster overall. That is the first half of the paradox: the path that feels like a detour is the only one that compounds.
Where AI Changes Everything
Here is where the objection returns with full force. "Fine — Data, then Process, then Technology. But mapping Data and redesigning Process is exactly the multi-year work I don't have time for."
This was true. It is no longer.
The instinct is to think of AI as something that lives only at the end of the sequence — the engine that runs once the Data is clean and the Process redesigned. That instinct is the mistake, and I need to call it out explicitly and responsibly, because I want us to succeed. AI belongs at every phase:
- In the Data phase, AI accelerates the foundational work itself — profiling, matching, deduplication, classification, and the mapping of Data flows that used to take armies of analysts months.
- In the Process phase, process mining reveals the real Process as it actually runs — factually, in hours rather than through weeks of workshops — and agentic AI then proposes redesigns against it.
- In the Technology phase, AI does what we always imagined: it runs the Perfect Process in operation.
AI appears three times, not once. That is the Acceleration Paradox in full: the same technology creating unprecedented time pressure is the technology that collapses the foundational work the pressure made you want to skip. The numbers shift accordingly. A single use-case foundation you once waited months for can now be laid in weeks; the full Enterprise Master Data build that classically runs four or five years can be done in one or two. And because each phase is now AI-accelerated, the value compounds — faster Data foundations enable faster Process redesign, which enables faster, safer deployment.
The Distinction That Holds It Together
I want to be precise here, because this is where the paradox is most easily misread.
"Data before AI" does not become "skip the Data because AI." The discipline is unchanged. The sequence is unchanged. What changes is the speed — not the necessity.
And this is where that statement from the big stage comes back: a fool with a tool is still a fool. AI is the most powerful tool most of us will ever hold — but point an agent at Bad or Ugly Master Data and it will not reject it. It will amplify it, confidently, at scale, inventing its own cipher key as it goes. The tool does not save the fool; it multiplies the foolishness at machine speed. That risk has not softened. What has changed is that you can now point AI at the work of getting clean — and staying clean — far faster than before. The same tool, used wisely, on the right work, in the right order, is what makes you anything but a fool.
So perhaps the formula earns a third movement:
You cannot AI your way out of a Data problem. But you can Data your way into AI success — and now, you can AI your way into Data readiness faster than ever.
Live Proof Points
This is not theoretical. Take a Process most of us still run in weeks: supplier onboarding. From a White Canvas, redesigned end-to-end, it can run in hours — third-party Data acquired externally rather than rekeyed, specialist partners plugged in for bank-Data security and fraud prevention, agentic checks running with no human queues, a secure portal closing the loop with the supplier. Process mining sits over the top as the monitoring layer, surfacing improvement ideas continuously, with Process Owners and Data Stewards as the human orchestration layer. Human, Process and Artificial Intelligence, each doing what it does best — the Intelligence Triad, amplified.
Notice what made it possible. Not the tools at the end. The Data was understood first. The Process was redesigned second. The technology was the easy part — chosen last, and chosen fast.
Contrast that with the years it once took to build the foundations behind a model like V.A.U.L.T. — clean bank-account and vendor Master Data, mapped flows, clear ownership. That work was real, and it was slow, and it paid off. The point of the Acceleration Paradox is not that the foundation matters less. It is that the same foundation can now be built in a fraction of the time — if AI is doing the building.
The Compounding Choice
Every initiative is a choice between two compounding curves. The Broken Sequence compounds debt: complexity layered on complexity, Process untouched, Data never ready, trust eroded with every failed pilot. The reversed, AI-accelerated sequence compounds value: each foundation laid makes the next outcome faster, cheaper, and more certain.
The teams that internalise this will not just ship better AI. They will out-accelerate everyone still trapped in the old order — not by moving recklessly, but by building foundations at a speed that was impossible eighteen months ago.
The discipline never changed. The speed to deploy it has.
So the question for every GBS and SSC leader is no longer "Do we have time to build foundations?"
It is: "Are we still using AI only at the end — or have we put it to work building the foundations themselves?"