48% of organizations introduced AI without redesigning the workflows or roles it sits within Deloitte AI Institute, Enterprise AI Trends in 2026 · ~3,700 professionals · June 2026

The tool arrives. The work underneath it stays the same as it always was.

Finding the Hidden Bottleneck

I saw this clearly on an engagement at Epsilon. The team was building a report for a business user whose role was marketing to prospects and customers, a straightforward file surfacing leads and account records to act on. When we asked what that person did with it once it arrived, the rest of their process became clear. Every time the report came in, they spent real time refining it further: segmenting records, flagging exceptions, applying rules that lived only in their own head, before the file was any use to them at all.

That extra effort is not a complaint. It is a signal. It means the person doing the work already had a clear idea of what a better version of that file would look like. They had just never had a way to build it in. We took their segmenting and flagging logic and built it into how the report itself was generated, pulling from data that had not previously lived in one connected place. The report that arrived the next week came in already categorized. Their job shifted from redoing that sorting each time to checking it, making small periodic adjustments, and then spending most of their time acting on what the report told them. They kept full control over the file. They just spent almost none of it on work the system could now do for them.

That is the pattern of workflows worth rebuilding: often a hidden opportunity, not the loudest complaint in the room. It is the one where someone has quietly built their own workaround, because the tool in front of them stopped one step short of the decisions they needed to make.

Ask what a person does with an output after it arrives on their desk, and the real workflow often starts right there, one step past where anyone had reason to look before.

Finding It Without Guessing

Finding it does not require guessing. Two questions tend to surface it:

Two questions that surface it
  • Where does this output go next, and who is waiting on it once it leaves this desk?
  • Is there a step someone repeats by hand every time, one they would clearly rather not, if the system already knew the rules?

Identifying these workflows happens when someone asked what happens after the handoff, rather than assuming the file was the finish line. The workflow worth fixing is often the hidden one, unnoticed while everyone talks about something louder.

Prepared by Forjic Technology. This is the Workflow Clarity dimension of the Forjic AI Readiness Survey. For the full analysis behind all five dimensions, see Access Is Not Adoption. Source: Deloitte AI Institute, Enterprise AI Trends in 2026, approximately 3,700 professionals, June 2026.