What can actually be automated in mid-market organisations

In almost every first conversation with a new client, we hear that sentence at least once. “That can't be automated.” Behind it sit different arguments: too many edge cases. Too much unstructured data. Too many individual decisions. Volumes too low. Too expensive. Too complex.

In most cases the sentence isn't true. It just sounds plausible because the organisation has worked manually for so long that any automation looks like a major project. Our experience from dozens of automation mandates: roughly 80 percent of processes considered “non-automatable” are in truth simply not yet cleanly described.

Why “can't be done” usually means “we haven't tried yet”

Mid-market processes grow. They emerge from an exception that became the rule. From an Excel sheet someone created five years ago. From an email chain that “just works”. These organically grown workflows carry organisations for a long time, and they deserve respect. But they're rarely documented, rarely consistent and rarely in a state that would allow automation without preparatory work.

When we look into a process like that, we typically find three things. First: a clear core that is 70 to 90 percent automatable. Second: a handful of genuine edge cases that should remain in the human decision frame. Third: a number of apparent edge cases that are actually just information gaps and disappear with a better input form.

The skill isn't to automate everything. The skill is to separate these three categories cleanly.

How we work

Our process automation service follows a recurring pattern that we've embedded in many organisations. It is deliberately pragmatic and doesn't shy away from uncomfortable questions.

Step 1: Make the process visible

We start by spending a short time alongside two or three people in their day-to-day. Not an interview in the meeting room, but real observation on the actual case. That way we capture what really happens, instead of documenting what the organisation believes happens. Almost always we see loops, double entries and waiting times that no-one notices any more.

Step 2: Separate the core from the edges

Then we model the process so that we describe the recurring core cleanly. Which inputs come from where, which decisions are actually rule-based, which are genuine judgement calls. We are deliberately radical at this stage about simplifying. Because every edge case built into the core makes the automation brittle.

Step 3: Automate where it pays off

The automation itself is often the smallest part. We connect existing systems, build integrations at the right places, set clear data contracts, add AI components where useful for unstructured inputs, and put people into the process at the points where they add value.

Step 4: Preserve decision points

We automate processes, but we don't disempower people. At the points where your organisation deliberately wants to decide, the decision stays with the human. The automated part only prepares, sorts, filters, documents.

What our clients get out of it

One mid-market client with applicant tracking systems automated job-listing processes that way and halved the handling time per posting. Another client was able to pre-filter recurring IT tickets with a clear context check and effectively eliminated support requests with no information attached. A third uses our automation to marry editorial approvals in their CMS with external master data sources.

The lever is the same in all three cases: as soon as the process is described, it can be improved. The claim that something “can't be done” loses its force when it meets a concrete sketch.

What this can mean for your organisation

If you're stuck in a process where you've heard the sentence “we can't do that” many times, a second opinion is worth having. We don't go in with the expectation of overhauling everything. We look precisely with you at where automation makes economic sense, where it doesn't, and which small step you can take first.