AI Strategy for SMEs: Start With One Workflow, Not a Roadmap
The 60-page AI strategy deck is where SME AI initiatives go to die. Better: pick one workflow with measurable waste — high volume, hours of pain, review-friendly, digital data — and ship it in about 6 weeks. The first shipped workflow produces the strategy; the deck never produces a shipped workflow.
We've watched the same movie in enough German SMEs to name the pattern: the workshop-to-nothing pipeline.
The workshop-to-nothing pipeline
External consultants run a two-day workshop. A 60-page deck arrives: maturity models, use-case portfolios, a three-year roadmap. A steering committee forms. Twelve months later nothing has shipped, the deck is stale, and the team is more AI-skeptical than before. The deck didn't fail because the analysis was wrong — it failed because it postponed contact with reality, and reality is where all the useful learning lives. The pattern is common enough that some companies have been through it twice, with two different consultancies.
How do you pick the first AI workflow?
Score your candidates on four criteria:
| Criterion | Strong candidate | Weak candidate |
|---|---|---|
| Volume | Happens daily, 20+ times | A few times a month |
| Pain | Hours of skilled time per week | Minor annoyance |
| Review tolerance | A human checks output before it leaves | Errors reach customers directly |
| Data availability | Inputs already digital | Paper, phone calls, tribal knowledge |
Four strong marks: build it. Anything where errors reach customers unreviewed is not a first project, no matter how attractive the ROI slide looks.
Typical first wins in German SMEs
Three workflows come up again and again in our projects: email and inquiry triage with drafted replies (high volume, review built in), quote generation from unstructured requests (hours per quote become minutes), and internal knowledge search across manuals, wikis, and old project folders (no output risk at all). All three ship in roughly 6 weeks, and all three produce a number — hours saved per week — that either justifies the next project or doesn't.
What the first project buys you beyond ROI
Three things no deck delivers. Data cleanup: nothing exposes a messy product catalog or inconsistent email templates like trying to automate against them. Evaluation muscle: your team learns to measure output quality instead of debating it in meetings. Realistic expectations: after six weeks, everyone knows what needs review and what doesn't — a lesson no vendor slide can teach.
After one shipped workflow, the strategy writes itself: it's the ranked list of the next three workflows, backed by your own numbers instead of a consultant's benchmark.
We ship first AI workflows as fixed-scope projects — see AI & LLM Integration, or read how to add AI to existing software first.
Olio on LinkedIn
