How we decide what to automate
Most automation disappointment comes from automating the wrong thing well. Here is the test we apply before we quote anything.
Four questions, asked out loud
1
Can you name the process in one breath?
If it takes a paragraph to describe, it is several processes. We split it or we stop.
2
Is the volume real?
We want the count per week and the people it currently occupies. A painful process that runs eleven times a year is not a candidate.
3
Do the rules exist?
Automation encodes a rule. If two experienced people in your team decide the same case differently, the rule is not written yet. That is a workshop, not a build.
4
Could it be approved out of an existing operational budget?
On its own, with no programme around it and no steering committee attached. If it needs an innovation budget to survive, it is too big for a first step.
What we need from you to scope it
Volume
How many documents per day or per week, how much time they take now, which teams touch them.
Variability
Which formats exist: scanned paper, PDFs, emails, images, structured files. How many major customer or supplier formats are in scope.
Fields
Which fields have to come out: quantities, item numbers, references, dates, reason codes.
Validation and exceptions
What makes a document valid or invalid, which cases need a person, who approves them.
Entry point
Where the data gets keyed today, which API or process step the automation would use, plus whether the first flow posts final entries or prepares them for someone to confirm.
None of that needs a project to answer. It is one conversation with the person who does the work today.
We say no more often than we say yes
-
The exchange is clean and structured, system to system. That belongs on your existing middleware, ION or plain EDI. A second layer would add only cost.
-
The volume is high but the rules are simple. A lighter integration will beat an agentic platform on cost, so we say so.
-
The volume is low and a person handles it well already.
-
The process is about to change, so we would be automating a moving target.
-
The real problem is master data quality, which automation amplifies rather than fixes.
Workato earns the recommendation only when the work needs interpretation, judgement, exception routing plus a governed write-back to your system of record. If it does not need all four, the cheaper tool is the right answer.
Saying this early is cheaper for you than discovering it in month three.
What a first engagement looks like
One process. One production flow. An explicit list of what is in scope plus an explicit list of what is not. Time and materials against a bounded estimate, so you can stop.
What is deliberately not in a first build: every format variation, every country or business unit from day one, full automation of everything adjacent to the process. The Not-in-Scope list is a feature.
Typically a first production flow lands in roughly six to eight weeks. That is directional rather than a guarantee.
The second process is a decision you make after the first one has a result, not a commitment you make now.
