Quick answer. First wins are high-volume, low-risk, well-documented tasks: summarising, drafting, classifying and extracting. Save decisions, judgment calls and anything regulated for humans — and never automate a process that's currently broken.
The quick-win matrix
| Department | Automate first | Keep human |
|---|---|---|
| Marketing | First-draft copy, repurposing, summaries | Strategy, brand voice sign-off |
| Sales | Email drafts, CRM notes, lead enrichment | Negotiation, relationship calls |
| Support | FAQ deflection, ticket triage | Complaints, billing disputes |
| Finance | Data extraction, report drafting | Decisions, sign-off, figures (verify) |
| HR | JD drafting, CV screening shortlists | Hiring decisions, sensitive cases |
| Operations | Document summaries, routing | Exceptions, judgment calls |
The test for a good first project
Score any candidate task against four questions. Three or four "yes" answers means start there:
- High volume? Worth automating only if it happens often.
- Low risk? A mistake should be cheap and recoverable.
- Well-documented? Clear inputs and a clear definition of "good".
- Measurable? You can show time saved or quality held.
What not to automate
- Legal, medical or financial decisions — see hallucination by industry.
- Anything needing accountability or human judgment.
- A process that's currently broken — fix it first, or AI scales the flaw. This is the top business mistake.
From first win to scale
Prove one task, measure it, then expand — the full sequence is in the 4-phase starter guide. Set guardrails before you grow with governance basics and the data privacy checklist.
Picked your first task? Use the match engine to choose the right model and the calculator to size the cost.