Is AI right for this problem?
Name the decision AI should support, who it affects, and the business value it should move. Then ask the honest question: is AI even the right tool?
Diagnostic 04
Know if your AI use case is ready before you build it.
Only 7% of consumers forgive AI more than they forgive a person. So getting it wrong costs more than you think. This review tests the use case, workflow, data, human oversight, governance, adoption, and how you’ll measure success. You’ll get a clear call: go, fix the gaps first, narrow it, or stop.
Questions answered
Turn AI interest into a clear decision.
I start with the customer problem and the business value at stake. Then I check whether the right conditions exist, before anyone compares platforms or plans automation.
Questions this diagnostic answersView questions
Is the workflow ready?
Look at the current work, owners, exceptions, data, and rules. If people can’t describe the process clearly, AI won’t either.
Where do people stay in control?
Decide where people review, approve, correct, override, or escalate, and who’s still accountable when it goes wrong.
How will you know it works?
Set the measures, adoption needs, risk signals, and review points before launch, so you can prove the value before you scale it.
Scope and evidence
Check the work before you automate it.
What I look at depends on the decision and what you can share. I compare what the AI is supposed to do with the real workflow, the context it’ll have, the judgment calls people make, and the risks.
Evidence we reviewView evidence
Use-case and decision fit
The customer problem, business decision, intended users, expected value, other options, and whether AI fits the level of risk.
Workflow and process readiness
Current steps, handoffs, decision rights, exceptions, service rules, recovery paths, and places where the work is still unclear or inconsistent.
Data and knowledge dependencies
Required records, knowledge sources, data quality, access, age, source, privacy limits, and the context people use to judge a response.
Human oversight and exceptions
Review points, approval limits, escalation paths, override authority, quality checks, and the work people still own.
Governance, risk, and trust
Acceptable-use limits, security and privacy needs, customer disclosure, audit records, accountability, failure impact, and stakeholder concerns.
Adoption and measurement
User readiness, training, workflow fit, success measures, risk indicators, feedback paths, and the evidence needed to continue, change, or stop.
Deliverables and decisions
Get a readiness decision and a responsible path forward.
A list of AI risks won’t help you decide. Each deliverable tells you whether to go ahead, what has to shape it, and what evidence should drive the next step.
Readiness decision
A straight answer backed by evidence: ready, fix the operation first, narrow it, or stop.
Supports the decisionWhether to proceed, fix the missing conditions, reshape the use case, or choose another approach.
Human-in-the-loop map
A clear view of where AI can help, where human judgment is essential, how exceptions move, and who has authority and accountability.
Supports the decisionWhich decisions AI can support, which need approval, and how people respond when the system is unsure or wrong.
Operating constraints and guardrails
Documented workflow, data, knowledge, governance, privacy, quality, and trust needs that must shape the implementation.
Supports the decisionWhich boundaries and controls must be in place before the use case can operate responsibly.
Sequenced implementation priorities
A practical order for operating changes, a focused pilot, adoption support, measurement, review, and any justified expansion.
Supports the decisionWhat to change first, what the first release should include, and the evidence that earns the next round of investment.
Engagement fit
Use this review before AI becomes a build commitment.
I’ll need access to the people, workflow, data, and knowledge around the use case. And you’ll need to be open to hearing that AI isn’t the right move yet.
Strong fit
This is likely a good fit when
- An AI proposal has support, but the workflow, governance, adoption needs, or success measures are still unclear.
- The use case affects customer service, success, or operations across teams, systems, or decisions.
- A sponsor can involve process owners, frontline users, data and knowledge owners, risk partners, and technical teams.
- You’re willing to delay, narrow, or reshape the idea when something important is missing.
Limited fit
Another path may work better when
- You only need a vendor picked or a chosen tool implemented, with no readiness decision.
- AI is expected to cover for a broken workflow without changing ownership, process, or decision rules.
- The process owners, frontline users, data owners, risk partners, or accountable leaders can’t take part.
- Success means launching, usage, or automation volume, not better work and better customer outcomes.