Book Brew 186: Your AI Pilot Has Been Circling the Airport for Six Months

In The Good Place, Mindy St. Claire lives in the Medium Place: a beige afterlife reserved for a person whose moral score landed squarely between eternal paradise and eternal pain.

She gets warm beer, a television that plays only one terrible movie, and the kind of meh that makes a dentist waiting room feel aspirational.

A surprising number of AI pilots (tests) businesses have implemented have moved in with her.

They were approved for 30 days. Six months later, three employees use them (poorly), two forgot the login, and leadership keeps saying, "We are still evaluating." The test-run has become the Medium Place of business technology.

Chidi Is Running the Steering Committee

Chidi Anagonye can turn a simple decision into an ethical hostage situation. Give him an AI pilot with privacy concerns, unclear metrics, and six opposing stakeholders, and the business may achieve immortality before it achieves implementation.

I understand the hesitation and totally understand that the AI tools are changing quickly. A poor decision can expose data, waste money, frustrate employees, or automate a process that already needed an exorcism.

Endless evaluation creates its own damage.

• People build shadow workflows.
• Early adopters improvise rules.
• Skeptics collect evidence that the project was always nonsense.
• The business pays for a tool that has the operational authority of an ugly Medium Place decorative throw pillow.
  
  

“Great planners who don’t execute their plans go nowhere.” - Ray Dalio, Principles

Planning produces value when it moves a decision forward. A pilot should gather evidence, expose risks, and teach you what the real workflow requires.

A Sandbox Is a Controlled Environment

A demo inside a sandbox can show that the tool can perform a task while everyone is watching and the example data has showered, dressed nicely, and remembered its manners.

Daily operations happen outside of the nice, neat sandbox and can present their own challenges:

• client exceptions,
• half-completed forms,
• cranky spreadsheets,
• confidential information,
• exhausted employees,
• and the process everyone swears is documented even though the official procedure is a screenshot from 2023 with a red arrow pointing at a button that moved last year.

Damien made a video you should watch: Rebuild Your Business or Job as AI-First.

“To begin with the end in mind means to start with a clear understanding of your destination.” - Stephen Covey, The 7 Habits of Highly Effective People

Give the Human an Actual Job

Some human-in-the-loop plans assign a person to click Approve after the AI finishes. Congratulations, your workflow now has a human-shaped mouse accessory.

Real workflow oversight requires:

• time,
• standards,
• skill,
• and authority.

The reviewer (aka the squishy human) needs to know:

• which errors matter,
• when to reject the output,
• and where to escalate the deeply weird stuff.

Four hundred approvals before lunch will turn any thoughtful checkpoint into a button-mashing side quest.

This is where our work usually starts.

In our custom-made Next-Level AI Workshops, we work with your team to select a real process, define the AI role, define the human role, and build the first review loop.

With Fractional COO/CTO support, we can carry that work further into ownership, permissions, metrics, handoffs, and implementation across the business.

Land the Plane

(did you know Damien is an actual pilot??)

Every pilot needs five decisions before anyone starts testing:

1. Name the owner. One person collects feedback, resolves questions, and moves the decision forward.
2. Name the business problem. Use a concrete description such as, "Client intake takes four hours and creates duplicate work."
3. Choose the evidence. Track time, errors, turnaround, client experience, or the number of employees whispering profanity at a spreadsheet.
4. Schedule the review. Put the decision date on the calendar while everyone still remembers why the pilot exists.
5. Use three outcomes. Adopt the workflow, revise the experiment, or retire it. Eternal evaluation sends the pilot back to Mindy’s house with another case of warm beer.

Are you prepared for landing?

Ponder This

1. What AI pilot have you implemented recently? How’s it going??

Books

• Principles - Ray Dalio
• The 7 Habits of Highly Effective People - Stephen Covey