Skip to content

12 Skills: Where People Shine and Where AI Helps

First PublishedByAtif Alam

Part 1 of 2 in the Using AI Well series. Part 2 turns these skills into a practical playbook for working with AI.

I work in site reliability engineering and AI now shows up in more of my daily work. That raised a question: When I make a good decision, what am I actually doing that AI isn’t?

Much of the public conversation frames this as humans versus AI. I think that’s the wrong question.

AI is a tool, and the useful questions are where it adds the most value and which skills we should sharpen ourselves.

David Hume put it well in 1739: “Reason is, and ought only to be the slave of the passions.” Deciding what matters is still our job.

It turns out there are twelve skills behind good judgment, and we use them far beyond big decisions: when we diagnose a problem, plan, learn something new, prioritize, review someone’s work, explain an idea or work with people, to name a few. Mostly, we use them without noticing.

Philosophers have been describing them for 2,400 years, but you don’t need the philosophy to use them. Below are the twelve skills, how people and AI compare on each, where the skills show up in everyday work and, for the curious, where the ideas come from.

The 12 skills: where each one helps, and where it slips

Section titled “The 12 skills: where each one helps, and where it slips”

These 12 skills are the building blocks of judgment work. Decisions are where they all come together, but each one shows up in everyday work too.

Probability is only one lens among them. AI models learn from statistical patterns in data, which makes them strong at some and weak at others. Neither people nor their tools are reliably good at all twelve.

Here is where each tends to be strong and where each tends to slip, based on the research listed at the end of this post.

SkillHumansAI today
Causal reasoning
”If I do X, Y will happen, because…”		Mixed. Can test causes by trying things, but overrate how well they understand how things workMixed. Fine on well-known causes, weak at inferring new ones from data
Logic and consistency
Checking that an argument holds together		Mixed. Reason better when a problem matches their experience, but seek evidence that confirms their view, not tests itModerate, improving. Tripped up by irrelevant details or rewording
Pattern recognition
”This looks like that incident last year.”		Strong with fast, regular feedback (expert intuition), but see patterns in noise without itStrong, but can invent patterns from thin data
Heuristics
Rules of thumb, like “avoid single points of failure”		Strong. Simple rules often work as well as complex models, but bring predictable biasesStrong at proposing rules and checklists
Analogy
Applying what worked in one situation to a new one		Mixed. Find deep analogies once prompted, but often miss them otherwiseStrong on familiar problems, brittle on unfamiliar ones
Domain expertise
Knowing how things fail, from hands-on experience		Strong. Hands-on, tacit knowledge of how things fail, but slow to build and can harden into rigidityMixed. Very broad book knowledge, no hands-on experience
Values and ethics
”Even if it works, should we do it?”		Strong on stakes and accountability, but often decide by gut feeling, then rationalizeWeak. No personal stakes; answers can shift with phrasing
Goal-setting
Deciding what matters before deciding what is likely		Strong at deciding what should matter, but can pick narrow goals that backfireWeak at deciding what should matter
Systems thinking
Seeing dependencies, feedback loops and side effects		Mixed. Know unwritten local constraints, but have poor intuition for feedback loops and delaysMixed. Maps systems well, misses unstated constraints
Counterfactual thinking
”What would happen if we didn’t do this?”		Mixed. Can test “what ifs” in the real world, but hindsight makes the past look predictableWeak at verification. Generates “what ifs” it can’t test
Metacognition
Noticing your own biases: “What would change my mind?”		Mixed. Can correct for bias by deliberately considering the opposite, but miss their own and overrate weak skillsWeak at catching its own errors without outside checks
Social reasoning
Reading incentives, trust and group dynamics		Strong. Read tone and body language quickly, but blame personality and overlook the situationMixed. Good on many theory-of-mind tests, brittle on variations, no tone or body language

“AI” here means general-purpose language models such as ChatGPT, Claude and Gemini, as of 2026. Ratings in both columns are rough judgments drawn from the research and ideas in this post, not scores. AI capabilities change quickly, so treat the right-hand column as a snapshot.

The takeaway

Section titled “The takeaway”

The point isn’t that humans win, or that AI does. It’s that people and their tools are strong in different places.

That difference is exactly what makes them useful together, as long as you know which to rely on for what.

That’s the subject of Part 2: Sharpen, Delegate, Partner, which turns these twelve skills into a practical playbook, with examples from SRE work.

Where these skills show up

Section titled “Where these skills show up”

Decisions are only one place these skills matter. Here are some of the kinds of work that draw on them, and the skills each leans on most.

ActivitySkills it draws on
Diagnosing a problemCausal reasoning, pattern recognition, counterfactual thinking
PlanningGoal-setting, systems thinking, heuristics
Learning something newAnalogy, metacognition, pattern recognition
Designing and buildingSystems thinking, heuristics, domain expertise, causal reasoning
PrioritizingGoal-setting, values and ethics, counterfactual thinking
Assessing risk and forecastingCounterfactual thinking, systems thinking, pattern recognition
Reviewing someone’s workLogic and consistency, domain expertise, metacognition
Explaining an ideaAnalogy, logic and consistency
Research and investigationCausal reasoning, logic and consistency, metacognition
Creative problem-solvingAnalogy, counterfactual thinking
Negotiating, teaching and mentoringSocial reasoning, analogy, values and ethics
Working with peopleSocial reasoning, values and ethics

Curious where these twelve skills come from? The rest of this post traces them back through 2,400 years of philosophy.

Where these skills come from

Section titled “Where these skills come from”

Thinkers from ancient Athens to medieval Baghdad and Cairo, and on to modern Europe and America, kept arriving at the same five ideas, often centuries apart. Each one supports some of the skills above.

Five ideas from philosophy feeding into the twelve judgment skills

Five ideas, and the skills each one supports.

1. Goals. Reason is a tool, not a compass

Section titled “1. Goals. Reason is a tool, not a compass”

Before any reasoning starts, someone has to decide where to go. Reasoning can tell you how to get there; it can’t choose the destination.

  • Aristotle (4th century BCE) called the skill of judging what is good, and acting well on it in real situations, practical wisdom. He saw it as different from cleverness.
  • David Hume (1739), quoted at the top of this post, put it bluntly: “Reason is, and ought only to be the slave of the passions.” Goals come from what we care about, not from logic.

Skills it supports: values and ethics, goal-setting.

2. Reason comes in many kinds

Section titled “2. Reason comes in many kinds”

There is no single “reasoning engine.” Different problems call for different kinds of thinking.

  • Ibn Sina (11th century) described intuition as the ability to grasp the missing link in an argument instantly, without working through each step. He noted that people differ greatly in it, an early account of expert pattern recognition.
  • Charles Sanders Peirce (19th century) named three modes of inference: deduction (applying rules), induction (generalizing from examples) and abduction (guessing the best explanation).

Skills it supports: causal reasoning, logic and consistency, pattern recognition, heuristics, analogy, counterfactual thinking.

3. Knowledge is more than information

Section titled “3. Knowledge is more than information”

Knowing a fact, knowing how to do something and knowing a person are different kinds of knowledge.

  • Al-Ghazali (11th century) pointed out that knowing the definition of something is not the same as experiencing it, just as knowing the definition of intoxication is not the same as being intoxicated.
  • Michael Polanyi (1966) made the same point in modern terms: “we know more than we can tell.” Much of what experts know can’t be fully written down.

Skills it supports: domain expertise, social reasoning.

4. The mind fools itself

Section titled “4. The mind fools itself”

Good thinkers build in ways to catch it.

  • Socrates taught that wisdom begins with knowing what you don’t know.
  • Ibn al-Haytham (11th century), a pioneer of experimental science, argued that anyone seeking the truth should challenge everything they read, and suspect their own judgment too.

Skills it supports: metacognition, counterfactual thinking.

5. The system shapes the outcome

Section titled “5. The system shapes the outcome”
  • Ibn Khaldun (1377) explained the rise and fall of dynasties as a cycle: group cohesion builds power, power brings comfort, comfort erodes cohesion. It is a feedback-loop model of whole societies, written centuries before “systems thinking” had a name.

Skills it supports: systems thinking.

Sidebar: what is a cause? One question runs from the 11th century straight into today’s AI research.

  • Al-Ghazali (c. 1095): when fire touches cotton and the cotton burns, we only observe the two happening together. We never observe a necessary link between them.
  • Hume (1740s): made a strikingly similar argument. We see one event follow another, and our minds supply the connection out of habit.
  • Judea Pearl (2018): described a “ladder” of causal thinking: seeing patterns, doing (predicting the effect of an action) and imagining what would have happened otherwise. He argued that most machine learning is stuck on the first rung.

Telling real causes apart from things that merely happen together is still one of the hardest problems for AI.

Sources and further reading

Section titled “Sources and further reading”

Thinkers in this post

Section titled “Thinkers in this post”

Want to go deeper?

Section titled “Want to go deeper?”

These thinkers explored the same ideas and are worth reading next:

  • Al-Farabi, who treated logic as the grammar of reason. SEP
  • Ibn Rushd (Averroes), who defended real causes against al-Ghazali. SEP
  • Suhrawardi and Mulla Sadra, on direct versus acquired knowledge. SEP: Suhrawardi, SEP: Mulla Sadra
  • Francis Bacon, whose “Idols of the Mind” were an early catalog of biases. SEP
  • John Dewey, How We Think (1910). SEP
  • Herbert Simon, on “good enough” decision-making. SEP, “Bounded Rationality”
  • David Lewis, Counterfactuals (1973). SEP
  • Hubert Dreyfus, What Computers Can’t Do (1972)

Research behind the comparison table

Section titled “Research behind the comparison table”

Humans

AI