Hey Reader,

Welcome to the 28th edition of the 3-2-1 (check out previous issues here).

One idea from me first, three links that go deeper on it, then two things that have nothing to do with work. Let's get into it.

---

1 Idea from Me

Opinions are free. Recommendations cost something.

"Heads up, customer complaints about onboarding are up." Then the Slack thread goes quiet (a thumbs-up emoji appears), and the problem is yours now. If you lead a team, you know what I mean.

What you want is what I wrote about 4 weeks ago: "here's the problem, here's what's causing it, I looked at three possible solutions, and here's the one I think we should pick, and why". Wouldn't that be pleasant?

Holding back or jumping in

In our client work, we've seen two ways in which early- and mid-career employees struggle to bridge this gap.

Some hold back: they don't trust their own view enough to put it in front of anyone. One program participant asked rushed questions "from a place of fear... so I don't make a mistake." Another held back clarifying questions "to avoid seeming like I am not understanding the ask."

Others jump in: they have a view and no process for testing it or landing it. An engineer in one program told me he built a reputation of being "the bull in the china shop." He said, "I ended up getting discounted before I could even open my mouth." Another engineer in the program nailed the gap when he told us he realized he had "a tendency to communicate really well... but only in my own thinking model."

Two weeks ago, I wrote about the process we discovered that fixes both, because it gives the first group a way to build a view they can defend and the second group a way to find out they're wrong before the room does.

Today, I want to talk about practicing that process, and doing so with enough stakes to make the learning stick.

Where did the reps go?

Greg Shove and Bernard Coleman, recent podcast guests, both worry that AI is taking the drudgery early-career people used to build judgment on. Greg's answer was simulators, the way we train pilots before they carry passengers, and then he caught himself: a simulator gives you the reps, but nothing happens to you when you crash it. Bernard likened this to simulations and role plays, which he called "manufactured adversity," and his point was that you need to remember the pain for the lesson to stick.

So how do we help early career employees feel the pain of making a decision that doesn't pan out, without tanking the company?

What if we coached our direct reports to make recommendations they publicly staked their reputations on? Not predictions, and not measured on how they panned out. But documented thinking, measured on how sound the process was?

There's a spectrum from opinions to decisions

Professor Sir Andrew Likierman, who wrote Judgement at Work, separates opinions from decisions. Opinions are informal and personal. Decisions are formal, and someone has to deliver them. Most of your team's work sits between those two. They can recommend.

Let's define a recommendation as an opinion you're willing to stake your reputation on. The bar is high, because you have to get multiple stakeholders' support to drive a decision, and the stakes are real because the decision is backed by your thinking.

So now imagine this as a manager: you encourage and coach your team to follow a process for developing high quality decisions and document their thinking as they go. They need to do the hard work of collaborating across functions, aligning multiple stakeholders, and communicating in a way that lands with different audiences to get those recommendations approved as decisions. Either way, you have a perfectly documented thinking rep to evaluate and identify areas to improve their judgment.

Evaluate the thinking

Many business decisions take a year or more to play out, so evaluating on results rewards luck and you basically teach people to hedge.

Likierman told me the best evidence for judgment is a systematic process, because outcomes can't prove that good judgment was exercised. There's too much randomness to isolate.

So review the thinking. Thinking is hard to see on a page, and I miss it too. Most of us with experience can't name our own process, so we see the answer and miss the steps. But we've been studying this for two years and have seen what steps work and which to avoid.

Your move

The next time someone brings you a problem, ask: what would you recommend?

Then have them write one page before they answer:

• the problem and how they know it's the problem,
• what they're recommending,
• whose view they went to get,
• what would change their mind, and
• who needs to decide.

If AI drafted the first pass, fine. The page they write is what they must be able to defend in a room of decision makers. It must show they checked it.

Have them read the page with one peer to pressure test first, then with you, and look for three things:

1. Did they get a view from someone outside their team, especially someone likely to disagree?
2. Did they lead with their recommendation, name trade offs and what they left out?
3. Did they determine what the person who needs to decide cares about most?

You can go more granular with the nine behaviors, but these three questions will get you the 80/20 of whether or not your team is executing judgment at a high level.

Hit reply and tell me which of the three gaps your team has most. I'll send you my thoughts.

---

3 Things for Work (in L&D)

Shaping Early-Career Success in the Age of AI (UT Austin McCombs, with KPMG, Jul 2026)

• The free write-up of a study that ran in HBR this summer. KPMG and McCombs gave 523 early-career professionals real business tasks with an AI agent. Half beat the AI, a quarter matched it, a quarter did worse, and nothing on their resumes predicted which group they landed in. The researchers' fix is our fix: evaluate how people work with AI, not just the deliverable.

• 🏋 Effort ≈ 6 min read

Strategic decisions: When can you trust your gut? (McKinsey Quarterly)

• The science behind "evaluate the thinking." Daniel Kahneman and Gary Klein spent years disagreeing about intuition, then agreed on when it can be trusted: a world regular enough to learn, and fast, clear feedback on your calls. Most business decisions offer neither, which is why the review of the page has to supply the feedback the outcome won't.

• 🏋 Effort ≈ 15 min read

Decision Journal (Farnam Street)

• Read it for the practice, not the template: write down what you think and why at the moment you decide, then come back later, so you can tell when you were lucky from when you were right.

• 🏋 Effort ≈ 6 min read

---

2 Things for Life

How to Criticize with Kindness (The Marginalian)

• Daniel Dennett's four rules for disagreeing with someone, borrowed from the game theorist Anatol Rapoport: restate their position so well they say "thanks, I wish I'd thought of putting it that way," list what you agree on, say what you learned from them, and only then say a word of criticism.

• 🏋 Effort ≈ 5 min read

Will Durant on forgetting

"Forget mistakes, forget failures, forget everything, except what you're going to do now and do it. Today is your lucky day."

• Widely attributed to the historian, though nobody seems able to say where he wrote it. After an edition about reviewing your own thinking, a reminder of what the review is for. You look back once, take what you need, and the rest is Sunday.

• 🏋 Effort ≈ 10 seconds

---

That's it for this week - enjoy your Sunday!

I'll be back in two weeks ✌️

Andrew

P.S. 👉 If you lead a talent or L&D team and want to practice this with peers rather than alone, Al Dea and I are running the Talent Leader Lab, an eight-week peer incubator. The founding cohort kicks off Nov 4, and 6 seats are left.