AI Adoption Roadmap Prompt: Decide What to Automate First

Most AI adoption advice stops at "start with a high-impact use case." Useful right up to the moment someone asks which use case. That's the gap an ai adoption roadmap prompt fills: it scores every workflow in a department, ranks them by automation fit, and tells you what to automate first, second, and never.

The pages that rank for this query are consultant essays. They describe five stages of maturity, warn about analysis paralysis, and recommend "modest, intentional investments." None hand you a way to actually rank your own workflows. You leave knowing the shape of the journey and still guessing where to start.

A prompt closes that gap because it applies the same rubric to every candidate. No pet-project bias. No loudest-voice-in-the-room effect. Just a score.

Why the ranking pages don't help you choose

Read the top results and a pattern shows up fast. TechCXO's roadmap for AI adoption sorts companies into five personas and tells 60–70% of them they're stuck in analysis paralysis. True, probably. But it contains no weighted scoring mechanism and no way to evaluate which internal workflow warrants investment first. The advice is directional: "start with a specific use case that makes sense."

Wizeline's four stages of AI adoption does the same at the technical layer, mapping the path from prompting to RAG to LLMOps. Helpful for architects. Silent on prioritization.

The closest anyone gets to operational is AI Success Lab's prompts-into-processes post, which tells you to "identify 3 weekly-repeated prompts" and schedule them. That's a heuristic, not a rubric. It can't compare a daily 40-minute report against a weekly 3-hour reconciliation and tell you which one pays back faster.

What you can do with this prompt

• Inventory every recurring workflow in a department, from raw notes or a process doc
• Score each one on volume, repeatability, error cost, and system sprawl
• Rank workflows into automate / assist / leave-manual bands
• Sequence a quarter-by-quarter rollout instead of a big-bang plan
• Write a one-page initiative brief for the top-ranked candidate
• Defend the priority order to a skeptical budget owner with numbers

Anatomy of the prompt

The reason a raw ChatGPT question returns mush is that it has no rubric to apply. The fix is a fixed structure: describe the department, name the scoring dimensions, lock the output shape.

Variables
  {{department_description}} — what the team does + its recurring workflows
  {{known_volumes}}         — rough frequency/time per workflow, where known

Prompt
  Role: automation strategist scoring workflows for AI fit.
  Task: score each workflow 1-5 on volume, repeatability,
        error cost, and system count. Composite = weighted sum.
  Rules: use only stated facts; write "Not stated" for gaps.

Output contract
  1. Scored inventory table (workflow | 4 scores | composite | band)
  2. Ranked automate/assist/leave-manual list
  3. Quarter-by-quarter sequence with the reason for the order

Put the output contract last. Models weight the most recent tokens, so a contract that sits above a long pasted process doc gets diluted. Claude honors a ## Output format heading more reliably than an inline "return a table" note; GPT-4o needs the columns restated on the final line or it drops one. Small detail, and it's the difference between a clean table and a wall of prose.

1. Gather inputs

Pull a plain description of the team and its recurring work into {{department_description}}. Add whatever volume data exists, even rough, into {{known_volumes}}. Don't pre-filter — the point is to score everything, including the workflows nobody wants scored.

2. Run and read the scores

Run the prompt and check the composite scores before the narrative. If a workflow you assumed was the obvious first pick lands in the assist band, that's the prompt earning its keep. Gut feel over-weights visible pain and under-weights quiet, high-volume drudgery.

3. Sequence, don't dump

A ranked list isn't a plan. The sequence step matters because the top-scored workflow sometimes depends on cleaning up a lower-scored one first. Read the reason column, not just the order.

Prompt-craft patterns that make the scores trustworthy

Two patterns do most of the work here.

Anchored scoring beats open scoring. Ask for a 1–5 score with no definition and every workflow lands at 3 or 4. Define the anchors — "5 = runs daily or more, 1 = runs quarterly or less" — and the spread appears. Anchors are what stop the model from being politely non-committal.

Volume:      5 = daily+   3 = weekly   1 = quarterly-
Error cost:  5 = customer-facing/financial   1 = cosmetic

Force a "Not stated" token. If {{known_volumes}} is thin, a model will happily invent frequencies. One line fixes it: "If a value isn't stated, write Not stated and score volume as 2." Now missing data reads as missing instead of quietly fabricated. Worth the extra sentence.

Variables you'll set

Two inputs, and both are things you already have. That's the test a workflow-scoring prompt has to pass: if assembling the input means doing the analysis first, the prompt failed before you pasted anything.

Where a single prompt stops

A one-shot prompt scores workflows well. It's weaker at the parts around the score: turning the ranking into a defensible sequence, writing the initiative brief for the top pick, and re-scoring next quarter as volumes shift. Those are separate jobs that reuse the same inputs, which is exactly where a structured pack beats a lone prompt.

Start free if you're testing one workflow. The Workflow Automation Scorer is a single prompt that takes a {{workflow_description}}, an {{error_cost}}, and the {{systems_involved}}, and returns a 25-point scorecard with an automate / assist / leave-manual verdict. Good for pressure-testing the method on one process before you commit to the whole inventory.

Getting started

1. Write a plain description of the department and drop it into {{department_description}}.
2. Add any volume or frequency data into {{known_volumes}}; leave gaps as "Not stated."
3. Run the prompt and sort by composite score, not by narrative.
4. Read the reason column on the sequence — order isn't always the same as rank.
5. Pull the top-ranked workflow into an initiative brief before touching tooling.
6. Re-score after a quarter; volumes and error costs move.

Once you've picked what to automate, the next question is what to build first inside it — the same prioritization discipline shows up in this product roadmap prioritization prompt. And if you're measuring whether an automation actually landed with the team using it, the feature adoption analysis prompt applies the same scored-inventory habit to usage data. Both pair naturally with the Department Automation Roadmap Playbook when a rollout turns into an ongoing program.