Skip to content
revot
revot

Workflow automation3 min readBy Revot

Map the workflow before you automate it

Automating a process nobody has written down makes its problems faster. A practical guide to mapping a workflow from trigger to completion — and what the map tells you before you build.

In short

  • Write down the workflow from trigger to completion before choosing a tool.
  • Capture people, systems, data, handoffs, exceptions, volume and cost — exceptions matter most.
  • Give every step and every exception a named owner.
  • Use the map to decide what to automate, what needs human approval, and how you will know it failed.

Most automation projects begin with a tool. Someone has seen what an automation runtime or an AI agent can do and wants to point it at a process that hurts. Sometimes that works. More often, the automation faithfully reproduces a process nobody had fully written down — including its gaps — and now the gaps happen faster and more quietly.

Mapping the workflow first is not bureaucracy. It is the cheapest way to find out what you are actually automating, where it breaks, and which part is worth fixing first.

Start at the trigger, end at completion

Pick one process and define its edges. What starts it — an enquiry, an order, a call, a form, a date? And what does “done” mean — the customer has an answer, the invoice is paid, the record is updated? Everything between those two points is in scope. Everything else can wait.

What to capture

For each step between trigger and completion, write down:

  • Who does it, and who they hand it to next
  • Which system it happens in — including email, spreadsheets and phone calls
  • What data moves, and whether it is sensitive
  • How often it happens and at what volume
  • How long it takes, and how long it waits before someone picks it up
  • What can go wrong, and what happens then

The people who do the work every day are the only reliable source for most of this. Process documents describe how work was meant to happen. The people describe how it happens now.

The exceptions are the process

The happy path is usually easy to automate and usually not where the cost is. The cost hides in exceptions: the enquiry with missing details, the order that does not match stock, the approval that needs a second opinion, the customer who replies to the wrong thread. Each exception is a decision someone currently makes by hand, often from memory.

List them. For each one, note how often it happens, who handles it, and what they need to know to resolve it. This is where most automation projects succeed or fail: an automation that handles the happy path and silently drops everything else creates more work than it saves.

Give every step an owner

Mapping often reveals steps nobody owns: the shared inbox everyone assumes someone else is watching, the spreadsheet that is updated “when there’s time”. Automating an unowned step does not create an owner. Before building anything, agree who owns each step and each exception queue. An automation should hand problems to a named person, not to a channel nobody reads.

What the map tells you

With the workflow on one page, the decisions get clearer:

  1. Which steps are high-volume, rule-based and low-risk — good candidates to automate first.
  2. Which steps are consequential — moving money, changing records, speaking for the business — and should keep a human approval step.
  3. Which steps need judgement on unstructured information, where a model might help inside clear limits.
  4. Which problems are not software problems at all, and need an owner or a policy instead.

Then build it to fail safely

Once you build, the map becomes the test plan. Every exception needs a path. Every external call needs a timeout and a retry policy. Steps that send messages or take payments need to be idempotent, so a retry never sends or charges twice. Inputs and webhook signatures should be validated where data enters. And every run should leave a structured log with a correlation ID, so that when something does go wrong, you can see exactly what happened and to which records.

Finally, decide what success looks like before you start — less time per case, fewer re-keyed records, faster first response — and how you will measure it against today’s baseline. Without a baseline, “it feels faster” is the best evidence you will have.

In short

Map the work, find the exceptions, name the owners, then automate the parts the map says are worth it. It is slower for the first week and much faster for every week after.

Next step

Start with the process that costs you the most.

Seven short questions about how the work happens today. No sales deck — the first conversation starts from your answers.