Draw the current workflow before you automate any of it
A director of operations at a warehouse wanted to automate a receiving process everyone described as simple. Drawing every step and handoff on a whiteboard revealed the exceptions the simple story left out.
Draw the current workflow before you automate any of it
A director of operations at a third-party logistics warehouse wanted to automate inbound receiving. When a truck arrived, someone checked the load against the paperwork, logged it, and released it to the floor. Everyone described the process the same way, in about one sentence. Truck comes in, we check it, we log it. It sounded like an easy thing to automate.
Before he approved anything, he made the receiving team draw the actual process on a whiteboard. Every step. Every handoff. Every place where something could go sideways. He told them not to draw the clean version. He wanted the real one, including all the annoying parts they usually just handled without thinking.
The whiteboard filled up fast, and the one-sentence process turned out to be about twenty steps with a dozen branches.
The exceptions are the whole job
The clean story was truck comes in, we check it, we log it. The drawing showed what actually happened. Sometimes the paperwork did not match the load, and then a person had to call the supplier, and sometimes the supplier disputed it, and then a different person had to decide whether to accept the discrepancy or reject the whole shipment. Sometimes a pallet was damaged, which was its own branch with photos and a claim. Sometimes the truck showed up without an appointment, which broke the sequence entirely. Sometimes the item was a controlled product that needed a second signature.
Each of those branches was rare on its own. Together they were most of the actual work. The “check it and log it” happy path was the part that needed no human judgment, and it was the only part the automation plan had accounted for. If they had built for the happy path and shipped it, the system would have handled the boring case beautifully and dumped every exception, unlabeled, back onto a confused human who no longer had the manual process to fall back on.
That is the trap with automating an undrawn process. The parts everyone forgets to mention are exactly the parts that are hard, and hard is where the value and the risk both live. You cannot design for an exception you never wrote down.
A head of claims at an insurer once told me almost the same story. Her team scoped an automation for straightforward claims and discovered, only after drawing it out, that “straightforward” was defined by the absence of about fifteen specific complications, none of which were documented. The definition of the easy case was a list of exceptions nobody had ever written down. Drawing the workflow was how they found the list.
Drawing it is the cheapest discovery you can do
There is no special tool required. A whiteboard, the people who actually do the work, and a rule that you draw the real process and not the tidy one. Put a box for every step, an arrow for every handoff between people or systems, and a branch for every “unless” and “except when” someone says out loud.
Two things fall out of the exercise. First, you see which steps are mechanical enough to automate and which are judgment calls that need a human, so you can scope the automation to the mechanical parts and route the rest to people on purpose instead of by accident. Second, you find the handoffs, the moments where work passes from one person or system to another, because those are where things get dropped and where an automation has to be especially careful.
The warehouse director scoped his project around the drawing. The automation took the clean receiving path and, just as importantly, detected each exception and routed it to the right person with the context they needed. The exceptions stayed with humans because the drawing made it obvious they had to. The system helped instead of hiding the hard part.
How we approach it at Density Labs
Early in every AI Opportunity Assessment, our fixed two-week engagement at $2,500, we get the people who do the work in a room and draw the current process on a wall, exceptions and all. We are not looking for the happy path. We are hunting for the branches, because the branches decide what an AI can safely take over and what it must hand back to a person. A process you have not drawn is a process whose hard parts are still invisible, and invisible hard parts are what turn an automation into extra work.
Before you automate a workflow, draw it. Every step, every handoff, every exception. The parts you would have skipped are the parts that matter.