Tools arrive faster than judgment.
McKinsey's August 2026 State of AI report found that 80% of respondents say AI has improved their individual productivity, while 37% attribute at least some EBIT impact to AI, about the same as a year earlier.
The distance is people: asking the right thing, checking what comes back, and owning the call.
The first layer of the work is the first layer of the learning.
Junior engineers learned by writing the first version, breaking something and finding out why. AI writes much of that first version now, and the experience that turns output into senior judgment has thinned.
How to develop senior engineers when AI writes the junior work
Specification, verification and conviction.
Three behaviors decide whether AI helps: specifying the problem clearly, verifying what comes back, and owning the decision. RCM ThinkLabs develops and measures all three over time.
Decisions with incomplete information and real tradeoffs.
In RCM ThinkLabs, engineers work through an immersive narrative where characters have their own interests and information is incomplete. There is no script to repeat, so people use the skill.
The measurement is never shown, so they decide instead of performing. Each session takes up to fifteen minutes and ends with private feedback.
The engineer gets the practice. The leader gets the opening.
Each engineer gets a record of their own growth from day one. Their leader gets an opening for each one-on-one through RCM Advisor, so coaching and mentoring start from what the person did.
The purpose is development. It does not rank people against each other.
What changed for participants.
Production data from Wingbrace, an advanced engineering team, over 39 days, measured against each person's own opening baseline.
Questions from technology leaders
What is AI workforce readiness for a technology company?
It is whether people can specify the problem, verify what AI returns and own the decision, measured from what they do rather than from course completions or tool usage. RCM ThinkLabs builds and measures those behaviors in short recurring sessions.
How do you prepare engineers to work with AI agents?
Tool training is necessary and dates quickly. The durable part is judgment around the tools: asking why, checking the output, noticing what an answer left out. RCM ThinkLabs develops those behaviors through practice with private feedback.
How do you keep senior engineering judgment when AI writes junior code?
Rebuild the repetitions on purpose, with frequent decisions that carry tradeoffs and feedback. The full argument is on our page about developing senior engineers when AI writes the junior work.
Does RCM ThinkLabs replace technical training or mentoring?
No. Technical training teaches tools, and mentors transfer context no tool holds. RCM ThinkLabs tells leaders where to look, so mentoring time goes where it changes the most.
Where do the results come from?
From Wingbrace, an advanced engineering team, over 39 days: 84% of regular participants improved on a full read, and the share of questions that asked why rose from 1 in 17 to 1 in 7. Results elsewhere depend on participation, feedback and manager reinforcement.
Sources. McKinsey, The State of AI in 2026: On the Road to ROI, August 2026, 1,719 respondents in 97 countries. Wingbrace figures are RCM ThinkLabs production data from an advanced engineering team, Days 1 to 12 against Days 28 to 39, measured against each person's own opening baseline. If anything here is out of date, tell us and we will correct it.
Keep judgment ahead of the tools. Measure what it builds.
Comparing options? See how RCM ThinkLabs compares with Workera, BetterUp and other alternatives.