Thought 01 · Digitalization
A maturity framework for digitalization
Across X all kinds of projects, it’s always critical to achieve as much clarity and common ground as possible with all stakeholders. Technical and business teams don’t always share the same language, and the messy reality of the day-to-day operations doesn’t help to create a clean picture.
After finding myself in this situation repeatedly, I gradually developed and refined this framework. It is deliberately simple, but I’ve found it useful for several purposes:
- building a shared vocabulary with domain experts about the digital maturity of our processes, helping us improve our understanding of the current situation
- identifying realistic next steps and setting clear expectations about what each one requires
- making progress visible
Just make it visible.
It won’t solve all your problems, but making data visible in the right place can make a huge difference. The hardest work here is breaking information silos and consolidating what’s relevant.
A screen showing stock levels in front of a machine. A dashboard that pulls from three systems that never talked to each other.
Data living in silos, requiring an analyst to extract, or only visible to someone who already knows where to look.
A nudge at the right moment.
No need to monitor dashboards: the system proactively notifies you when something needs attention.
Stock drops below a threshold. A delivery is late. An anomaly appears in production output. You find out immediately, not at the end of the week.
Dirty or incomplete data and noisy systems. Too many false alarms—and everyone ignoring them—is worse than no alerts at all.
A nudge with a plan.
The system doesn’t just notify you; it gives you context and a recommended action. It helps, but you still decide.
Stock is low. The system identifies the supplier, checks lead time, and drafts a purchase order for your approval.
Jumping here before Alert is reliable. If you don’t trust the diagnosis, you won’t trust the recommendation.
Let it do its thing (but keep an eye on it).
The previous stages have built the foundation. Now the system acts autonomously on well-defined, high-volume, low-ambiguity processes. Humans handle exceptions and set the boundaries.
Routine purchase orders below a threshold, with known suppliers and within budget, are processed without human intervention. Edge cases are escalated.
Automating when human input is essential. Skipping stages and automating bad processes. Sleeping at the wheel: it still needs monitoring.
The way I see it, you can skip a stage, but not the work of understanding what happens there.
A good example: a company tried to implement autonomous forklifts on the shopfloor, directly from completely manually driven ones. The vendor had a sophisticated system that required a lot of structure and standards we were not even close to being able to provide. It never went anywhere. Eventually I led the development of a more pragmatic product that evolved as we learned. By the time I left, we reached the ‘Assist’ stage, where each forklift driver had a screen telling them when and where to pick up containers and where to deliver them. The product had become a critical part of the process and created strong foundations for future automation.
I’ve intentionally left AI out of the framework because it doesn’t affect its core. It changes what’s possible at each stage, or how fast you can move through them, but it doesn’t remove the lessons implicit in each stage.
With AI it can be even more tempting to jump directly to ‘Automate’, but more often than not, starting with a more manual approach can expose missing data, undocumented exceptions and the judgments people apply without necessarily being able to articulate them. That knowledge becomes the context necessary to make AI workflows reliable enough for production.
The framework isn’t intended as a rigid sequence for an entire company. A single process can contain tasks at very different stages. The useful question is not “How digitally mature are we?” but “What would the next useful level look like for this particular task?”
Next thoughtWhere AI fits in messy business processes