Before you can improve, digitize, or document a process, you need to know how it actually runs today. Not how it's supposed to run. Not how it's written up in the QM manual from 2021. But how your people really execute it every day - including the Excel sheet that officially doesn't exist, and the shout across the hallway that drives half the order flow.
This article compares the common process discovery methods honestly: what they deliver, what they cost, and where they fail in practice. And it shows a newer approach that combines the depth of the interview with the scalability of software.
Why as-is capture is the most important step - and the one most often skipped
Most digitization and improvement projects start with the to-be. A target process gets drawn, a tool gets selected, a workflow gets configured. The as-is gets shortcut along the way: "We know how things work around here."
The problem: in most companies, nobody knows it completely. Management knows the process as it was introduced three years ago. The team lead knows the official version. The clerk knows the version that actually works - with all the exceptions, edge cases, and workarounds that have accumulated over the years.
If you document the to-be instead of the as-is, you automate right past everyday reality. The typical outcome: the new system models the process the way the boss described it. The team works with it for three weeks, discovers that half of the real cases can't be handled - and goes back to the hand-maintained spreadsheets on the department share that were already working. Except now there's an additional system to maintain that nobody trusts.
As-is capture is therefore not busywork before the real project. It's the foundation every sound decision rests on: What's worth automating? Where does duplicate work happen? Which knowledge depends on a single person?
The classic process discovery methods, compared soberly
There are four established ways to capture as-is processes. Each has its place - and each has weaknesses that methodology handbooks rarely discuss openly.
The one-on-one interview: thorough, but expensive
The structured one-on-one interview is the gold standard of process discovery. An interviewer sits down with an employee and works through their daily routine: What comes in, what goes out, which systems, which exceptions, who's involved?
The strengths: In a private conversation, people tell you how things really work. Including the things they'd never say in a large group - for instance that the official approval process gets bypassed regularly because otherwise it costs two days. The interviewer can ask follow-ups, surface contradictions, go deep.
The weaknesses: The effort. Do the math realistically: preparation, one to two hours of conversation, write-up, minutes, sign-off on the minutes - that quickly adds up to half a person-day per interviewee, on the interviewer's side. With 40 employees, that's weeks of work. Which is why in practice only a handful of people get interviewed - the managers. And with that, you're back to the to-be instead of the as-is.
On top of that: the quality depends entirely on the interviewer. And the results end up in minutes that someone has to consolidate into consistent documentation. That step gets dropped astonishingly often.
The workshop: fast, but distorted
The process workshop brings everyone involved to one table. Brown paper rolls hang on the wall, or a digital whiteboard, and the process gets modeled together with sticky notes.
The strengths: In half a day, a shared picture emerges. Interfaces between departments become visible because both sides are in the room. And the workshop creates buy-in - people have drawn "their" process themselves.
The weaknesses: Group dynamics. When the CEO or the department head is in the room, the process gets described the way it should be - not the way it is. This phenomenon has a name: the HiPPO problem (Highest Paid Person's Opinion). The best-paid opinion in the room wins, even if the clerk in the third row knows better. She just doesn't say it.
Also: a workshop averages. If sales region North lives the process differently than sales region South, the group settles on one version for the poster - and one of the two realities disappears from the documentation. And finding a slot where eight people are available at the same time is a project in itself at a mid-sized company.
Observation: close to reality, but labor-intensive
In so-called shadowing, an observer accompanies employees through their workday and logs what actually happens.
The strengths: Observation captures things nobody would mention because nobody notices them anymore. The three system switches for a single transaction. The lookup in a private notebook. The call to the colleague "who always knows".
The weaknesses: The effort is even higher than for interviews - whole days per person. And the observer effect is real: people who are being watched work differently. More tidily, by the book, without the shortcuts that shape the process in everyday life. The very thing you want to see goes into hiding. Observation also only sees what happens during the observation window - month-end closing, vacation coverage, the complaint case stay invisible if they don't happen to occur.
Self-documentation: cheap, but it loses to day-to-day business
The idea sounds compelling: employees document their processes themselves. Hand out a template, set a deadline, done.
The strengths: No interviewer needed, fully parallelizable, the expertise comes straight from the source.
The weaknesses: In practice, self-documentation almost always fails - not from ill will, but from day-to-day business. Documentation is the task that always loses against the customer on the phone right now. After three reminder emails, you end up with results ranging from "two pages of prose" to "five bullet points". Comparability: none. On top of that comes a structural problem: anyone who has been running their process for years considers most of it self-evident and doesn't write it down at all. Precisely the implicit steps that matter are missing.
Comparison table: process discovery methods
| Method | Effort | Depth | Captures the real as-is? | Typical weakness |
|---|---|---|---|---|
| One-on-one interview | High (~½ person-day per employee for the interviewer) | Very high | Yes, with a good interviewer | Doesn't scale - usually only managers get interviewed |
| Workshop | Medium (½-1 day, everyone at once) | Medium | Limited | HiPPO problem, contradictions get averaged away |
| Observation | Very high (days per person) | High | Partially | Observer effect, edge cases stay invisible |
| Self-documentation | Low (on paper) | Low | Rarely | Loses to day-to-day business, implicit knowledge missing |
The pattern is clear: the methods that capture the real as-is don't scale. The methods that scale don't capture the real as-is. Which is why as-is capture gets shortcut or skipped entirely so often in practice.
The new approach: AI-led interviews
Recently, a fifth option has emerged that resolves this trade-off: an AI conducts the one-on-one interviews - via chat or voice, 15 to 20 minutes per person, whenever it fits into the workday.
That fundamentally changes the math. The bottleneck in classic interviewing was never the employees' talking time - everyone has 20 minutes. The bottleneck was the interviewer: their time, their availability, their write-up. Remove that bottleneck, and you can interview everyone instead of a handful of managers. And that is exactly when you get the as-is instead of the to-be.
Tools like wissa.ai implement this for mid-sized companies, with a few properties that learn from the weaknesses of the classic methods:
- Snowball principle: If someone says in an interview "Max and Hanne handle invoice approval", both automatically get invited to a conversation. The discovery works its way from knowledge holder to knowledge holder, instead of someone having to guess a complete interview list up front.
- Contradictions stay visible: If management says "it runs through SAP" and the team says "it runs through a spreadsheet on the group share", both statements stand side by side with their source - instead of disappearing into workshop consensus. A human decides which one holds, with one click.
- Living documentation: The process documentation isn't created as one-off minutes; it updates itself after every conversation, with a traceable diff against the previous version. Every statement can be traced back to its interview round.
- No HiPPO problem: Every conversation is a one-on-one. Nobody has to say in front of the boss that the official process gets bypassed.
Important for European mid-sized companies: an approach like this has to be works-council-ready. Concretely, that means: participation is voluntary, no performance evaluation of individuals, GDPR-compliant hosting in the EU with a data processing agreement. Check this with every vendor - without the workforce's trust, nobody tells you how things really work, and then even the best method is worthless.
Which method for which situation?
A closing assessment without sugarcoating:
- A single, critical process with few people involved: Classic one-on-one interviews by an experienced person remain a good choice.
- Interface conflicts between two departments: A moderated workshop - but without hierarchy in the room, and with prior one-on-ones as the foundation.
- Suspected hidden effort at one workstation: Targeted observation, as a supplement, not as the main method.
- The process landscape of the whole company, 10 to 500 employees: Here the classic methods fail on effort alone. AI-led interviews are currently the only approach that delivers interview depth at this breadth.
FAQ
How long does as-is process capture take at a mid-sized company?
It depends on the method. Classic one-on-one interviews cost the interviewer about half a person-day per interviewee - with 40 employees, that's several weeks. Workshops deliver a rough picture in one to two days, but with the distortions described above. AI-led interviews run in parallel: 15 to 20 minutes per person, no scheduling coordination, and the documentation is generated along the way.
Why isn't the existing QM manual enough as as-is documentation?
Because it describes the to-be of back then. Processes change constantly - new tools, new colleagues, new workarounds. A manual that isn't systematically updated documents a state that no longer exists. Comparing the manual against reality is itself a good first step of as-is capture: where does daily practice deviate, and why?
What is the HiPPO problem in process workshops?
HiPPO stands for "Highest Paid Person's Opinion". In group settings, the view of the highest-ranking person prevails - not because it's correct, but because nobody openly contradicts it. In process discovery, this leads to the official process being documented instead of the lived one. One-on-ones without hierarchy in the room avoid this.
Will employees even participate in process discovery interviews?
Usually yes - if three conditions hold: participation is voluntary, it's visibly not about performance monitoring, and the effort is small. Most people actually find talking about their own work for 15 to 20 minutes a sign of appreciation; it's rare that anyone takes a genuine interest in their daily work. Early involvement of the works council, if one exists, is important.
What do I do with the results of the as-is capture?
The as-is documentation is the foundation for everything that follows: spotting weaknesses (duplicate work, media breaks, knowledge tied to individuals), prioritizing on solid ground what gets digitized or improved, and deriving onboarding materials for new employees. What's decisive is that the documentation gets maintained afterwards - otherwise, in two years it's the next Word template from 2019 that nobody opens anymore.
Want to capture your company's process landscape without sinking weeks into interviews? wissa.ai conducts the conversations for you - 15 to 20 minutes per person, via chat or voice, with living documentation as the result. Early beta: €149/month with 12 interview credits, interviewees free. Join the beta waitlist now.