Flowgrammer

AI Automation Partner Scorecard

Ten questions and an evidence rubric for comparing an AI automation partner before approving a project.

Download the scorecard (PDF)

Use this scorecard to compare an agency, consultant, freelancer, or internal build team before you approve an automation project.

Score each question from 0 to 2:

  • 0: no clear answer or evidence;
  • 1: partial answer, but important details are missing;
  • 2: specific answer with a named owner, example, or document.

The ten questions

Question Score 0–2 Evidence to request
Can they define the operational outcome and current baseline? A before-and-after metric and how it will be measured
Do they map the workflow before selecting tools? A workflow map showing inputs, decisions, handoffs, and exceptions
Are human judgment and escalation explicit? A list of decisions the system cannot make alone
Can they explain data access, security, and failure modes? Permissions, data flows, retention, and failure handling
Are testing and acceptance cases defined? Normal, incomplete, conflicting, and escalation test cases
Who owns accounts, code, documentation, and credentials? A written ownership and handover plan
What support exists after launch? Response expectations, maintenance boundaries, and change process
Can they show relevant and verifiable proof? A case study, system walkthrough, or approved reference
Is the quote tied to scope and risk? Deliverables, assumptions, exclusions, and change rules
Will they tell you when automation is the wrong answer? A clear example of work they would leave human or fix manually first

How to read the score

  • 0–7: pause. The project may be carrying too much delivery, ownership, or measurement risk.
  • 8–14: continue discovery, but resolve the missing evidence before signing.
  • 15–20: the partner has supplied a usable basis for comparison. Confirm references, scope, and ownership before approval.

A high score does not guarantee a successful project. It means the partner has made the important work visible enough to evaluate.

Questions to ask in the first call

Ask the partner to walk through one real workflow. Listen for the baseline, the exception path, the human review point, and who owns the system after launch. If the conversation stays at the level of tools and features, ask what the team will do differently on Monday morning.

Flowgrammer's approach is documented in How to Choose an AI Automation Agency and the AI Success Audit.

Continue with Flowgrammer

Book an AI Success Audit call