Guide · AI in Institutions
What to define before automating a workflow
Automation projects often disappoint because they were never defined properly. Before anyone builds, pin down the task, the volume, the exceptions and who is accountable.
The short answer
Define five things before automating a workflow: the exact task and its trigger, how often it runs and how long it takes by hand, what goes wrong and how often, who reviews the output, and what happens when the automation fails. If you cannot answer all five, you are not ready to build.
The task, precisely
“Automate reporting” is not a task. “Every month, combine twelve partner spreadsheets into the donor template and flag totals that do not match” is. A precise task has a trigger, inputs, an output and a definition of done.
Volume and time
Count how often the task runs and time it by hand. The value of automation is mostly volume times minutes, and a task that runs twice a year rarely justifies a build.
Exceptions
Every process has cases that do not fit: a missing file, an odd format, a value that makes no sense. Estimate how often they occur and how long each takes to resolve. Every exception is a manual run you still pay for, and a high exception rate usually means the process should be fixed before it is automated.
Accountability and failure
Name the person who reviews outputs and the person who is alerted when a run fails. An automation that fails silently is worse than the manual process it replaced.
Put numbers on it
The automation ROI calculator takes these inputs and subtracts review time, exceptions, maintenance and running costs. In its illustrative worked example, a task run 120 times a month costs about USD 900 by hand; after every cost is subtracted it saves about USD 574 a month, and the build pays for itself in about seven months.
Automation ROI calculatorOpen the free tool What this guide does not cover
Some tasks should not be automated at all, especially decisions about people. Those need human judgement, whatever the arithmetic says.
Responsible for this guide
Examples in this guide are illustrative, not client results. Figures come from the free tool’s worked example; change the inputs in the tool to see your own.
Related services
The same method, delivered with you: scoped in writing, with a named principal and a fee agreed before any work begins.
AI agents and workflow automation
We automate one recurring task end to end, with error recovery, run logs and a tested handover, before touching the next one.
See the serviceServiceAI-readiness and AI-governance assessment
We map how one team or institution already uses AI, where data and decisions are exposed, and which governance gaps matter most. Diagnostic recommendations are kept separate from implementation, so you decide what to fund next.
See the serviceMore from AI in Institutions
6-minute readBuild an AI inventory with owners, evidence and a review cycle 
6-minute readFrom an AI-use policy to decisions staff can actually follow 
7-minute readHow to evaluate an internal knowledge assistant before release 
6-minute readCalculate automation value after exception handling and running costs What needs to move forward?
Tell us the decision, the challenge or the opportunity. We reply with a scoped approach, a named principal and a fee before any work begins.
No charge to submit an enquiry.