How to choose the first business process to automate

Start with a workflow, not a technology

The first automation should solve a defined operational problem. “Use AI” is not a process. “Record every website enquiry, assign an owner and remind that person if no next action is logged” is.

The DGS first-automation scorecard

Score each candidate from 1 (low) to 5 (high). A promising first process scores well on value and readiness without unacceptable risk.

Do not simply add all scores. Treat low stability, poor data or high consequence as gates. A high-volume process is still a poor first project if nobody agrees how it should work.

Interactive decision tool

Is this a sensible first process to automate?

Choose one real process, then rate five factors from 1 (low) to 5 (high). The result is a decision aid, not a guarantee.

RepetitionHow often does the same process occur?Select 1–51 — Rare or highly variable23 — Regular45 — Frequent and repeatable Business valueHow much time, revenue or customer experience could improve?Select 1–51 — Minimal benefit23 — Useful benefit45 — Material benefit Process stabilityDo people agree on the steps and outcome?Select 1–51 — Unclear or constantly changing23 — Mostly understood45 — Clear and consistent Data readinessAre the required inputs available and dependable?Select 1–51 — Missing or unreliable23 — Usable with some work45 — Structured and dependable Safety and reversibilityCan errors be detected, corrected and handed to a person?Select 1–51 — High consequence or hard to reverse23 — Manageable with controls45 — Easy to detect and reverse

Assess this processReset

Your result will appear here.

Select all five ratings, then assess the process.

Your answers stay in your browser and are not submitted to DGS.

A worked example

Illustrative example: a consultancy compares social-post scheduling with enquiry assignment. Scheduling is frequent and easy, but late enquiry ownership is affecting revenue and customer experience. The enquiry workflow has clear rules, structured form data and a reversible reminder step. It becomes the better first automation even if the social task feels more annoying.

The smallest sensible implementation

A complete narrow workflow is preferable to a broad system that stops halfway. For example, automatically copying a form into a spreadsheet is not enough if no person is alerted and no follow-up status is recorded.

When not to automate

• The process changes every time because the service itself is not settled.
• Most decisions require contextual human judgment.
• The available data is inaccurate, incomplete or collected without appropriate permission.
• The expected saving is smaller than setup and maintenance effort.
• A failure would be difficult to detect or reverse.

How to prove value

Record a simple baseline before implementation: volume, handling time, delay, error count or missed next actions. After launch, compare the same measures and review exceptions. Scale only when the first workflow is reliable and produces a useful result.

← Back to Insights