A customer sends an enquiry. Someone reads it, copies the details into a spreadsheet and asks a colleague for advice. A document comes back with a missing measurement. Another email goes out. Eventually, someone has enough information to prepare a response.
This is an illustrative example, but it shows why understanding a workflow requires more than naming its steps. The work includes the waiting, the checking and the conversations that happen between them.
Before designing an AI system, we need to understand that work.
The procedure and the working day
A written procedure gives us a starting point. The people carrying it out can explain where it works, where it becomes difficult and which exceptions they handle through experience.
An audit should bring those views together. Follow a real piece of work from the moment it arrives to the point where someone considers it complete. Ask what information each person needs, which tools they use and how they decide what happens next.
An informal check may carry important judgement. A spreadsheet may hold information that another system never receives. A process may depend on one person remembering to chase an approval.
These details influence what a useful system must do.
Look at the handovers
Time spent doing a task and time spent waiting for the next step are different things.
A team may prepare a document quickly, then wait several days for someone to review it. Automating the preparation helps with one part of that process. Making the review visible and giving it a clear owner addresses another.
The audit needs evidence for both. Where possible, examine a sample of completed tasks and record the active work, the waiting and any correction needed. State the limits of that sample rather than turning a few observations into a confident prediction.
Understand the exceptions
The usual case is only part of the job.
What happens when a customer leaves out information? Who decides whether a request needs specialist advice? What should happen when two documents disagree?
These questions help identify where software can apply a clear rule, where AI might assist and where a person needs to make a decision. Some improvements may simply require a clearer process or better information.
The people doing the work should help review those conclusions.
An audit should leave something useful behind
The result should be a process map the team recognises, a record of the main difficulties and a prioritised set of opportunities. It should also identify dependencies, unresolved questions and a suitable first pilot.
That gives the business something concrete to review before committing to development.
At InnoWeb, this is the approach we want to build our partnerships around: understanding a real workflow, agreeing what could improve and developing a system around the people who use it.
The starting question is straightforward. Which process in your business would you most like to make easier?