The 9Q Grid

9Q / Why AI pilots fail

Why AI pilots fail, and how to tell before you fund one

Most AI pilots fail before the technology does. They are mislabelled (a Q3 automation sold as a Q9 transformation), or they ask for more machine agency than the organisation's trust and governance can support. As Suhit Anantula writes in The Co-Intelligent Organisation: "Most AI failure is diagnosis failure before it becomes technology failure."

The 9Q Grid is Suhit Anantula's map of AI agency and type of work. Read the 9Q Grid.

The usual reasons, and what they miss

Data, integration, skills and change management are real. They are also symptoms. The root is usually diagnosis: the wrong quadrant for the trust available, or a story that runs ahead of the systems.

Reason 1: the pilot is in the wrong quadrant (mislabelling)

Common patterns: Q3 sold as Q9, Q1 sold as Q5, Q6 sold as "optimisation" when it is really an ungoverned leap. Check by placing the pilot on the grid with two questions — how much will the AI act, and what kind of work is it?

Reason 2: the pilot overdraws the trust budget

Trying to buy a Q9 outcome with a Q1 trust budget leaves the organisation insolvent in the currency of trust. See the AAA Licence chapter.

Reason 3: the failure mode nobody watched

Silent drift (Q3) — automation on yesterday's rules.
Optimising the wrong thing (Q6).
Recommendation capture (Q2).
Simulation capture (Q8): addicted to iteration, never commits — literally pilot purgatory.

Reason 4: the workflow should have been deleted first

Automating steps that should not exist is a common stall. Delete Before Digital thinking comes before digitising the mess.

Failure shapes: Shotgun and Drift

Two portfolio patterns make pilot failure more likely. A Shotgun scatters pilots across many quadrants so nothing compounds. A Drift looks like upward movement in the story while the deployments stay behind. Full taxonomy on Strategic Shapes.

From pilot to production: a five-question pre-flight check

Which quadrant is it really in?
Have we earned the trust it needs?
What's its failure mode, and who watches it?
Should any steps be deleted first?
What's the next quadrant if it works?

Public cases

Public case — not a client: Robodebt (Australian Government; library places it at Q3 Automator) is an AU example of automation without matching governance, drawn from the Royal Commission record. Optionally see Klarna as another public library reading.

What about AI ROI?

ROI follows position: higher-agency quadrants promise more and need more trust first. Tie measurement to what the AI does, what people stop or start doing, and what governance costs.

FAQ

Is it true that most AI projects fail?

Failure rates quoted online vary widely by study and by how "failure" is defined. The more useful question is why yours might stall.

Why do AI pilots fail to reach production?

Mislabelling, an overdrawn trust budget, unwatched failure modes, and automating work that should be deleted.

What are common reasons AI projects fail?

Data and integration issues are real, but most trace back to diagnosis: the wrong quadrant for the trust available.

How do you move an AI pilot into production?

Run the five-question pre-flight check on this page, then fund only what the trust budget can govern.

How do you measure AI ROI?

Tie it to the quadrant: what the AI does, what people stop or start doing, and what governance costs.

Helix work in this area

With the City of Mitcham, Helix narrowed 51 use cases to 3 pilots and ran a six-week AI Pod. Helix case studies ↗

Pilot stuck? A Helix Clarity Scan tells you what to fund, fix or stop.Delete before you digitise — talk to Helix Lab

Turn your map into a next move.

Use the 5Ps to explore responses to what the map reveals, test a move and carry the learning into the next loop.