When Process Mining Fails

I've seen process mining succeed. I've seen it unlock millions in savings, fix broken flows, and empower teams to build solid process improvement strategies.

But I've also seen it fail. Not because the tech wasn't good enough or the data wasn't clean enough, but because the people who needed the insights couldn't get them without going through someone else.

The Unicorn Model That Worked (Until It Didn't)

In a previous role, I led a team that delivered process mining capabilities across a large global organisation. We embedded directly with the business units. We sat with them, listened to their problems, and surfaced insights in the moment. We turned these insights into numbers for business plans, and armed them with everything they needed to affect real change.

We were multi-domain specialists who could bridge the gap between data and process. Our team combined technical skill with real-world experience across areas like Finance, Manufacturing, and Supply Chain. We acted as internal consultants, trusted by the business because we spoke their language and understood their challenges.

And it worked. Really, really well.

But senior management saw it differently. "This isn't scalable."

They weren't wrong. This operating model relied on having a rare kind of person, one who understood both the data and the domain. And you couldn't replicate that everywhere, especially across dozens of business units.

The Instinct to Scale like other Data Analytics Capabilities

Once a few process mining projects succeed, the instinct is often to scale process mining like any other data analytics capability: split out the roles, industrialise the workflows. Data engineering here, dashboard design there, requirements gathering up front.

On paper, this kind of operating model might look scalable. But it subtly shifts the focus. You start treating dashboards as the end deliverable, measuring success by how many you've built. The real goal gets lost.

Process mining isn't about delivering visualisations; it's about uncovering insight, driving conversations, and prompting action. It's about giving teams the evidence and numbers they need to push for strategic change. Without that mindset, you end up delivering dashboards that rarely get used because the operating model has missed the point entirely.

The challenge is finding a way to scale without stripping out the in-the-moment rabbit-trail insight that made those early projects succeed, and that those rare, multi-skilled individuals made possible.

Enter the Business Champion

So we tried another operating model: train someone from within the business to use the process mining tool themselves, someone we called a process mining champion.

They weren't data engineers and analysts. They were team leads, ops managers, subject-matter experts with an understanding of the value process mining could unlock.

The idea was that by giving them access and training on the tooling, they could self-serve their way to insight. Sometimes it worked. But often, it didn't.

Why?

Because most process mining tools are still fairly technical. Making meaningful changes or adapting an analysis often requires a deep understanding of the underlying data structure and modelling approach which is too much to expect from someone who doesn't have this experience nor the time to become a part-time data engineer and analyst.

And in hindsight we were really just trying to replicate another version of the unicorn model, hoping to find those rare individuals who could bridge the same gap we had, but just embedded in the business.

So the gap remained. The process expert had the context to see the insight, to follow the subtle patterns, the exceptions, the odd loops that didn't belong. But they couldn't get to the insight without help.

When Insight Requires a Rabbit Trail

Some of the best discoveries we made came from people following rabbit trails.

"Why are refunds up this month?"
"That's odd, why are these orders skipping QA?"
"Could this be linked to the recent system update? I wonder if the timing lines up?"

These weren't questions on a requirements document. They were questions sparked in the moment by curiosity, by context, by someone who knew the process inside out.

And when those people could pair their intuition with real data, that's when the magic happened.

But only if they could explore it themselves (or whilst riding a unicorn 😉).

When exploring insights requires a rare breed...

Can AI Close the Gap?

This is where we think AI can make a real difference.

Imagine an operations manager saying:

"Show me the orders where customers complained before anything was dispatched."
"Show me where we're having to fix or rework the same invoice more than once."
"Where should we focus next to get things done faster?"

No complex queries. No data prep. Just natural questions turning into powerful, flow-aware insights.

If done right, this doesn't just make process mining easier, it finally puts it in the hands of the people who actually know the process. It also provides them with a conversational partner they can bounce ideas off in real-time.

Not by scaling analysts. Not by forcing every team to hire individuals with a rare mix of skills. But by using AI to act as the translator, the invisible analyst between process context and insight.

Final Thought

Process mining was never meant to be a back-office tool. Its power lies in unlocking insights from the people closest to the work, but that only happens when the tooling becomes an extension of their own thinking.

In my previous role, we tried to bridge the gap with those rare individuals who had both data and domain expertise, both in our process mining team and out in the business. And while that approach had real impact, it didn't scale.

The push to scale like any other analytics function (by splitting out roles and industrialising workflows) was understandable, but it missed the point entirely. Process mining relies on curiosity in the moment, not a predefined set of requirements pushed through a drawn-out delivery workflow.

Back then, we didn't have today's AI capabilities to help bridge that gap at scale. But now we do, and that shift is shaping what we're building at Flow Myna. We're designing a service that brings the power of process mining directly to the people who live and breathe their process every day. Not by dumbing it down, but by enabling curiosity. Helping process experts explore, find answers, share insights and develop process strategies without needing a data team to translate for them.