Day 1b · Critical Thinking · 11:30 AM EST
The 10%
who thrive
What separates people who use AI well from those who get buried in it.
15 min + Q&A
Katherine
critical thinking · QA instinct · 5-year outlook
The question isn't "can you use AI?" — it's "can you tell when it's wrong?"
The reality of AI use today
52%
of Claude conversations involved active
human collaboration — not hands-off delegation
Anthropic Economic Index, Jan 2026
↗
Augmentation — what that means
Human actively guides the AI: steering prompts, reviewing outputs, making final calls. Contrast with automation (45%) — where users delegate tasks with minimal back-and-forth or review. (Remaining 3% unclassified.)
Most people still actively guide the AI.
This workshop is about staying in that majority.
The five-layers-removed SVP
CEO
→
VP
→
Director
context loss
Manager
→
IC
→
Work
Middle managers who've lost ground-level context make bad calls. AI does the same thing when you let it run unchecked.
Recurring theme for the 2 days: AI imitates humans.
From your own team · Pre-workshop survey
You need the context
loaded yourself
"I don't want the team to become mindless, thoughtless people that rely solely on Claude and aren't providing a distinct creative, strategic human perspective."
— Survey feedback
📖 Read the source, not the summary
Before this workshop, every survey answer was read individually — not just the AI summary. That's the context that makes the AI useful.
🚀 The 10% who thrive
They combine AI speed with strategic perspective and data QA. CEOs manage people smarter than them all the time. You're training to drive the smart ship — not become the ship.
📖 How a serious engineering org thinks about it
Humans remain accountable
for all artifacts
Bryan Cantrill, Oxide Computer — RFD 0576
↗An "artifact" is anything you create and ship: a report, an email, a client deck, a code commit. If AI helped make it, you still own it.
⚖️
Responsibility
"…bear responsibility for the artifacts we create, whatever automation we might employ to create them."
🔬
Rigor
"…can help us sharpen our own thinking by pointing out holes in our own reasoning."
🤝
Empathy
"…we must keep in mind our empathy for that human, be they the one who is consuming our writing."
👥
Teamwork
"…not use LLMs in such a way as to undermine the trust that we have in one another."
⚡
Urgency
"…that pace must not come at the expense of our responsibility, rigor, empathy and teamwork."
"…if used recklessly or thoughtlessly, they can have the opposite effect, replacing crisp thinking with generated flotsam."
Merriam-Webster Word of the Year 2025
↗slop /slɒp/ noun
AI-generated content mindlessly created and thrust upon someone who didn't ask for it.
"You can feel it when you read it — there's a rhythm to it, a particular kind of confident vagueness. That discomfort is your QA instinct working."
Handing off unchecked AI work damages trust — fast.
One slop report to a client can undo months of relationship-building.
Build the habit, not the checklist
Develop a QA instinct
Every output needs a human pass before it goes anywhere.
🔊 Read it aloud
Does it sound like a person or a press release? Slop has a rhythm you can hear.
✅ Verify the claims
Dates, stats, names. AI confabulates. A 5-second search catches most of it.
✍️ Would you own it?
If it went out with your name, would you stand behind every line?
⏭ Skippable · Live demo reference
The same prompt.
Two very different outputs.
❌ Without context
Prompt:
"Write a paragraph summarizing Q3 attribution performance for Client X"
Output:
"Client X saw strong Q3 performance, with a 22% increase in ROAS driven by paid social, and a 14% improvement in cost-per-acquisition vs Q2. The campaign reached 2.4M impressions with above-benchmark CTR."
⚠ Every number in this paragraph was invented. None were in the prompt.
✓ With context loaded + QA pass
Same prompt + actual Q3 data pasted in
- ✓Output cites real numbers → you can verify each one
- ✓Read aloud: does it sound right?
- ✓Verify the claims: dates, stats match source
- ✓Would you own it?
This is the difference between Slide 3b and the slop slide. Same tool. Same prompt. Completely different outcome based on what you brought to it.
Research findings · Why autonomy raises the stakes
The closer AI gets to autonomous,
the more this matters
⚙️ The Automation Trap
Scale AI's analysis of Manus (2025): fully autonomous agents completed just 2.5% of complex professional workflows end-to-end. The other 97.5% needed human intervention somewhere in the chain.
"…as various tasks are stitched together without any human review in between them, the risk of errors compound, and any time saved in the process is lost by needing human intervention to fix the end product."
— Debevoise Data↗
🧠 The Trust Trap
Microsoft Research, CHI 2025 — surveyed 319 knowledge workers across 936 real work tasks using GenAI. Key finding:
"Higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking."
— Lee et al., CHI 2025↗
"The rules that we set up for agentic AI systems usually cannot capture the nuanced social and cultural contexts that experienced employees rely on when deciding that not following a policy is actually the right course of action."
— Debevoise Data, Nov 2025↗
Where this is going
In 3 years, most routine knowledge work
will be AI-capable.
Dario Amodei, Anthropic CEO
↗
😰 Who struggles
- ·Stops at Tier 1 — pastes output, never edits
- ·Hands slop off to clients or teammates
- ·Tries to out-produce rather than out-think
- ·Gets replaced by anyone with the same prompt
🚀 Who thrives
- ·Treats AI as augmentation
- ·Maintains QA instinct
- ·Leads with judgment & strategy
- ·Builds trust through transparency
Neither column is about the AI getting better. It's entirely about what you do with it.
Where competitive advantage actually lives
70% of competitive advantage
is people and process. Not tech.
IDC / Workera AI Workforce Report 2026
↗You're sitting in the 70%.
You don't have to win the AI race. You have to build the human layer that knows what to do with it.
After the workshop
Normalize talking about AI
in your team
🏆
Share what workedWins normalize use and spread patterns across the team.
💥
Share what failedFailures build judgment faster than any tutorial.
📅
Stay on top weeklyCompetitive advantage is time-sensitive. This field moves weekly.
🔄
Ask for feedback — and make it safeIf something you shared didn't help, ask. "Hey, was that useful?" creates a culture where people iterate instead of just tolerating slop quietly.
After this workshop: we'll have an ongoing Slack channel for exactly this.
For managers and senior contributors
New hires are coming in
with different habits
The gap isn't AI skill — it's critical thinking under AI pressure.
Managers across the industry are noticing a widening gap — not in AI skill, but in the critical thinking that makes AI skill useful. It's most visible in people who built their foundations before AI was ubiquitous.
🎯 Hire for judgment
Pull up a Claude-generated client performance summary with one wrong number — e.g., credits the wrong campaign for a lift. Ask: "Does this look right to you?" Someone who skims and says yes tells you something. Someone who asks to see the raw data tells you something very different.
🗂️ Onboard for context
Have a new analyst build one client report manually before giving them AI access. Not as a ritual — because you cannot QA what you don't understand. An analyst who's never pulled attribution data by hand cannot catch it when Claude gets the attribution wrong.
📣 Model it publicly
At your next team standup: share a prompt you used that gave a confidently wrong answer, and show how you caught it. Not as a cautionary tale — as a normal part of your workflow. "Here's what I asked, here's what it said, here's what was actually off." Senior people modeling this changes what new hires think "good" looks like.
Most AI onboarding gives people tool access. Almost none of it teaches them when not to trust the output. That gap is yours to close as a manager.
Day 1b · Complete
Critical thinking isn't
a soft skill. It's your main skill.
Everything else in the next two days builds on this.