Aerospace has readiness levels. Your AI workflows do not.

 I read my friend friend Yvette Brown and her business partner Shari Nomady’s new AI Fluency book this weekend and it is wonderful about articulating the mindset shifts and confidence that come with AI Fluency.  Loved it. A conversation we had sparked Yvette to write her ideas into the book , and that book inspires today’s JERK Report.

It echoes the Jim Rohn quote “We are the average of the five people we spend the most time with.” AI fluency spreads like that too; one person builds something that works and others get inspired.

I was thinking about AI fluency and adoption and how can we gauge it? Can I get a JERK read on it? Can I use that to make the training I give even better?

I realized the whole thing reminded me of aerospace work I used to do on technology readiness levels and the transition to production. Some of you reading this are nodding and some of you unsure what I just said.

When you make something new and it works in the lab, scaling it to production is a different problem. You can grow a new material in a petri dish and still be unable to make it at the size of a car.

Will Willoughby’s templates became the basis of the Defense Science Board’s transition-to-production work, NAVSO P-6071. He told me once, while we worked together, that in the early NASA days they could barely make some of the parts at the required dimensional tolerances. They tested every part. They scrapped more than ninety percent in order to have what they needed to get to the moon.

That gap was a big enough problem that NASA created Technology Readiness Levels to show the technology works. Manufacturing Readiness Levels, added later by the DoD, show you can make it at scale.  

AI adoption is a similar pattern.

One person can build a prompt, a skill, a gem, a workflow that makes their own work faster or better. That is a technology readiness level. One person, proven.

Sharing it across the organization so other people run it consistently, without the builder adjusting the knobs, is a manufacturing readiness level.

We could call it a workflow readiness level. WRL. Pronounce it as whirl.

Workflow readiness levels address Technology Readiness for the first 3 levels and Manufacturing or Scaling Readiness in the higher levels.

Workflow Readiness Levels

Proven for you, analogous to a technology readiness level

1. The AI workflow is built from an existing process as is, not yet refined for AI.

2. The AI workflow is refined for AI. The output is better or faster than the old way.

3. You run the AI workflow regularly and get what you need with only minor adjusting of the output. Proven for you.

Proven for others, analogous to Manufacturing Readiness Levels

4. Someone else has run the AI workflow and succeeded, but needed your help to do it.
5. Others run the AI workflow regularly without you, with only minor adjusting to the output, from a written spec.
6. The AI workflow is how the work gets done. A manager owns it, a non-builder keeps it running, and people can go on vacation without detrimental effects.

 The Jerk Layers this week

The JERK read to do on your WRL

Position

Your workflow’s WRL today.

Velocity

Is the WRL climbing week to week.

Acceleration

Is WRL climbing faster than it was

Jerk

The rate itself turns. People pass the workflow on without you. Or the subtle reversal: new people still pick it up, but fewer each week than before.

As you’re thinking about how to get the ideas out of your head alone, so others can do this too, start with this week’s practice.

Your five-minute practice this week

Pick one workflow you use AI for. Just one. Paste this into your LLM.

I built this AI workflow that works well for me. I want to move it from something that works when I run it to something the whole organization can run as a normal part of work, without me adjusting it each time. Interview me. Ask one question at a time, keep each short, and wait for my answer. Where a list of answers makes sense, offer choices and include Other. If I give a number, ask whether it is a count or a guess. Ask these in order:

1.      Name the workflow, what it produces, and who uses the output. Ask me to paste or describe the last output and the workflow prompt itself if available.

2.      Of the last ten times I ran it, how many outputs were usable without rework. What did I fix by hand.

3.      What does it depend on that only I know or control: prompts, files, data, access, a tool, a model version.

4.      Has anyone else run it. What happened, and what did I have to tell them.

5.      Who would own it if I were out for a month. Whose job or routine would change if this became the normal way to do the work.

Then answer in under 350 words:

·        Proven for me: what is my evidence it works. The Workflows Technology Readiness Level

·        Proven for others: could people who are not me run it repeatedly. Yes, no, or not yet. This is the Workflow Readiness Level

·        The gaps: what stands between this and everyone using it, weakest first. Name the weakest one.

·        The handoff: write a one-page spec another person could run it cold from. Purpose, inputs, steps, what good looks like, when to reject an output, who owns it, and what to retest when the model changes.

·        The first step this week, and how I will know it worked.

·        What is my current WRL?

Do not flatter me. If my answers are vague, say so and tell me what to bring next time.

Run it again next week and see the gap between this week’s answers and next week’s, your velocity of WRL. One sitting only gives you a position.  

If you go on vacation, does everything stop while you’re gone?

If yes, which workflow is most critical and are you starting there?

If no, then what are you doing next with your freed up time?