Issue 149  July 17th 2026

AI can summarize everything and clarify nothing.

As AI floods your work with summaries, options, and fragments, the scarce human skill is turning them into direction.

THE ARGUMENT

In the last issue of Being Designerly, the evidence was blunt: AI can improve the work without improving the worker.

This issue is about the specific muscle that atrophies first.

Synthesis.

AI is very good at giving you more. More summaries. More meeting recaps. More extracted themes. More generated options. More "key takeaways." More polished bullets from material you barely had time to read.

That sounds useful. And often it is.

But more fragments do not automatically create direction.

A dozen AI summaries can still leave a team unclear. A page of generated options can still avoid the real tradeoff. A polished deck can still fail to answer the question everyone is actually asking:

So what should we do?

That is where synthesis becomes undelegatable.

A summary makes the material smaller. A synthesis makes the meaning larger.

A summary says, "Here are the main points." A synthesis says, "Here is what these points mean together."

A summary preserves the pieces. A synthesis creates the through-line.

That distinction matters more now because AI has made fragments cheap. It can ingest the transcript, scan the research, cluster the feedback, compare the options, and draft the recap in seconds.

But someone still has to decide:

What matters? What is noise? What pattern is real? What tension cannot be ignored? What direction follows from all of this?

That is not a formatting task. That is judgment. And it is one of the easiest skills to fake with AI.

Here's the uncomfortable callback: synthetic competence often looks like synthesis. It has headings. It has themes. It has confident language. But underneath, nothing has been decided. The material has been organized, not understood.

The tell is simple:

If nothing was excluded, nothing was synthesized.

Synthesis requires leaving things out. It requires choosing the through-line. It requires naming the tension. It requires deciding what the material means now, for this audience, in this context.

AI can help you see the fragments.

But the direction still has to be earned.

THE TELL

Here's how to catch a summary pretending to be synthesis - in your own work this week. (If you recognize these on your team, forward this so you have a shared vocabulary for it.)

🚩 It lists themes but never names the tension between them.

🚩 It includes everything, because nothing was important enough to exclude.

🚩 It repeats what people said without explaining what it means.

🚩 It sounds clear, but no decision gets easier after reading it.

🚩 Everyone nods in the meeting, then asks, "So what are we doing?"

If the output made the material neater but not more meaningful, it wasn't synthesis. It was arrangement.

YOUR PRACTICE THIS FORTNIGHT

🛠 The Human-First Rep: The One-Sentence Through-Line
(no AI - do this first)

Synthesis is the most common casualty of AI-assisted work - it's what you outsource every time you ask AI to "summarize these ten documents." So this fortnight, do one synthesis task manually.

Take a messy input: meeting notes, customer feedback, research findings, competing options, a long thread, a strategy discussion. Read it. Then write one sentence:

The most important thing this material reveals is ______, which means we should ______.

Do not make it perfect. Make it directional. Then ask:

What did I include?

What did I deliberately leave out?

What tension did I name?

What decision does this make easier?

What would someone disagree with?

That last question is the diagnostic. If no one could disagree with your synthesis, you probably only summarized.

🤖 The Sparring Prompt: Attack the Through-Line
(now bring AI in - as a challenger, not a crutch)

After you've written your human-first through-line, open your AI tool and paste this:

"I'm practicing synthesis, not summarization. Here is the material I reviewed: [paste notes/findings/context]. Here is my one-sentence through-line: [paste your sentence]. Act as a sparring partner. Do not rewrite it yet. First, test the strength of my synthesis: What did I smooth over? What tension did I resolve too early? What would I have to exclude for this to actually mean something? Where is my point of view too generic? Then give me one sharper version of the through-line and one alternative that points in a different direction."

Compare what the AI attacks to what you already noticed. The gap is your atrophy risk - the meaning-making you're closest to outsourcing.

This is the PAIR loop again: Prime (you formed the through-line first), Assess (AI attacks it), Interpret (you decide what holds), Retain (the judgment stays yours).

Reply and tell me: what messy input did you run your Human-First Rep on? I read every response.

THE PROOF

This issue is less about one article and more about a pattern showing up across knowledge work.

📄 Is AI Ruining Our Skills? Early Results Are In — and they’re not good
The empirical case for why "outsource the thinking, keep the output" quietly erodes the underlying skill. Synthesis is exactly the kind of judgment-heavy work that decays when you stop doing the reps.
→ nature.com/articles/d41586-026-01947-1

📄 WSJ Intelligence: The "Human Premium" — Yahoo Finance / WSJ
Names the skills judged "non-replicable" — the ones that make meaning rather than just produce output. Synthesis sits squarely inside that premium.
→ finance.yahoo.com (WSJ Intelligence study)

📄 AI Is Splitting the Job Market — Straits Times / PwC
Roles that amplify human judgment grow 2x faster. Turning fragments into direction is that judgment, made visible.
→ straitstimes.com (AI job-market analysis)

FROM THE BOOK

I've been sharpening the "Hands" section of Undelegatable - the part about how knowledge workers create value when AI can generate so much of the artifact.

One skill keeps getting clearer:

Synthesis - Turning Fragments into Direction.

Not summarizing. Not organizing. Not making things shorter.

Synthesis is the human act of deciding what a collection of fragments means - and it matters more precisely because AI creates more fragments than ever.

The future of knowledge work will not belong to the person who can collect the most information. It will belong to the person who can turn scattered inputs into a direction others can understand, trust, and act on.

That is the work AI can assist. But it should not own.

"The real risk of AI at work isn't that it replaces your skills. It's that it replaces the reps that built them."
- from Undelegatable, forthcoming

Know someone whose team ships more decks but makes fewer decisions? Forward this - synthesis is a team skill before it's a solo one.

The human edge is not automatic. It has to be built.