Two Ways Project Managers See AI — And Why It Matters

The word “artificial intelligence” contains a hidden diagnostic. Which word do you land on first?

Do you emphasize the ARTIFICIAL of Artificial Intelligence?

Or do you emphasize the INTELLIGENCE of Artificial Intelligence?

Pay attention to how your colleagues talk about AI.

Some say “it doesn’t really understand anything — it’s just predicting the next word.” Others say “I don’t care how it works, look at what it just did.”

These aren’t personality differences. They reflect something more fundamental: a difference in understanding.

Research on AI literacy has shown that many users don’t distinguish between probabilistic systems (where the same input can produce different outputs each time) and deterministic systems, where the same input always produces the same result. That gap in understanding drives almost everything else: how much people trust AI outputs, how they verify them, and how badly things can go wrong on a project.

For project managers, this matters twice. It shapes how you personally use AI. And it shapes how your team uses it: with risks you may not be seeing.

The two camps

Think of it as two orientations, anchored to different words in the phrase “artificial intelligence.”

CAMP ONE The “Artificial” thinkers

• Understand the probabilistic/deterministic distinction
• Know outputs are pattern-matched, not reasoned through
• Hold two ideas at once: impressive and not intelligent
• Skepticism is informed, not reflexive
• Verify outputs before acting on them
• Ask: which parts of this need to be deterministic?
• Resist “wishful mnemonics” — won’t say it understands or knows
• Aware of the gap between fluency and comprehension
• Understand hallucination as mechanical inevitability

| CAMP TWO The “Intelligence” thinkers

• Relate to the output, not the mechanism
• Don’t distinguish probabilistic from deterministic
• Anthropomorphize naturally — “it thinks,” “it knows”
• More likely to trust confident output without verifying
• Mistake fluency for comprehension
• Fill in gaps — project human qualities onto partial interactions
• Often the more intellectually open, growth-oriented people
• Optimism is genuine but not grounded in how it works
• Vulnerability: can be misled or over-reliant without knowing it

---|---

Research published in Communications Psychology analyzing over 12,000 open-ended metaphors from a nationally representative U.S. sample found that Americans are increasingly viewing AI as warm and human-like — meaning the “Intelligence” camp is growing, not shrinking. Nature, 2026 →

Which camp are you in?

Five scenarios. For each one, circle the answer that sounds most like you. There are no right or wrong answers — this is about self-awareness. Tally your score using the key at the end of the article.

A = 0 points B = 1 point Maximum score = 5

Question 1

An AI tool gives you a confident, well-written summary of a project risk. What do you do next?

A Use it — it reads clearly and covers the key points.[0 pts]

B Cross-check it. Fluent writing does not mean accurate content.[1 pt]

Question 2

A colleague says the AI “understood exactly what I needed.” How do you react?

A Makes sense — it does seem to understand context pretty well.[0 pts]

B It matched the pattern of what you needed. That is not the same as understanding.[1 pt]

Question 3

You ask the AI the same question twice and get two slightly different answers. Your first thought?

A Odd — I wonder which one is right.[0 pts]

B Expected — it is a probabilistic system. I need to verify both against a source.[1 pt]

Question 4

Someone on your team says they do not trust AI because “it makes things up.” How do you respond?

A I get it — that is a real concern and hard to predict.[0 pts]

B It is not making things up — it is completing a probability distribution. The issue is using it without verification guardrails.[1 pt]

Question 5

You are evaluating whether to use AI for a critical project deliverable. What is your primary concern?

A Whether the output quality will be good enough.[0 pts]

B Whether this task can tolerate probabilistic output, or needs a deterministic check built in.[1 pt]

Add up your B answers and check your score at the end of this article.

This is not a generational divide

It is tempting to assume the “Artificial” thinkers are older, more technical, or more cynical. That is not what the evidence shows. Both orientations appear across all ages, all backgrounds, all levels of technical experience. What differs is understanding, specifically, whether someone has ever been taught how a probabilistic system actually works.

Here is the uncomfortable part: most AI literacy frameworks were built around older, deterministic systems. They were not designed to teach the probabilistic nature of generative AI. So the “Intelligence” camp is not just uninformed; they have been under-informed by the field itself.

A generative AI literacy framework from academic researchers notes that existing guidelines “fail to address” the high degree of non-determinism in large language models, because they were built around non-generative systems. arXiv, 2025 →

The anthropomorphism problem is getting harder, not easier

There is a compounding dynamic that project managers need to understand. The better AI gets at sounding human, the harder it becomes to hold the “Artificial” frame. Researchers have noted that LLMs now mimic human communication so convincingly that calls to resist anthropomorphism increasingly fall flat — because the systems themselves exhibit human-like qualities.

Gestalt psychology helps explain why. People mentally fill in the gaps when confronted with incomplete information. When an AI responds fluently and empathetically, the “Intelligence” thinker’s brain constructs a mind behind it — even when explicitly told there isn’t one.

A 2020 paper tracing anthropomorphism in AI back to a 1976 complaint about “wishful mnemonics” — terms like “understand” or “learn” applied to AI — shows this is not a new problem. The language of AI has always nudged people toward the “Intelligence” frame. Taylor & Francis, 2020 →

What this means for you as a project manager

Two practical areas: how you use AI yourself, and how you manage it on your team.

Know your own orientation

If you are in the “Intelligence” camp, the risk is not that you use AI, it is that you use it without a verification habit. On a project, the cost of a confident wrong answer that nobody checked can be significant. Build a rule: any AI output that informs a decision gets verified against a primary source or a subject matter expert.

If you are in the “Artificial” camp, the risk runs the other way. Over-skepticism leads to under-use. The tool has genuine capability. Your job is to know which tasks can tolerate probabilistic output and which ones cannot.

Diagnose your team before you deploy

Most organizations treat AI onboarding as uniform; everyone gets the same training. But a team with mixed orientations needs differentiated preparation. The “Artificial” thinkers will be frustrated by surface-level training. The “Intelligence” thinkers need verification habits built in before they start, not after something goes wrong.

For “Artificial” thinkers Give them the mechanism. Acknowledge limitations directly. Let them define where deterministic guardrails are needed. Don’t oversell — they will disengage.	For “Intelligence” thinkers Don’t dismiss their enthusiasm — it drives adoption. Instead, introduce structured skepticism gently. Build verification into the workflow before they encounter a high-stakes failure.
On your project team These two types complement each other well. The “Intelligence” thinker spots possibilities. The “Artificial” thinker stress-tests them. Managed well, the tension is productive.	In risk planning Know which team members are verifying and which are trusting. In high-stakes decisions — scope, budget, stakeholder communication — make verification a process requirement, not a personal habit.

The bottom line

AI literacy is not about being for or against AI. It is about understanding what kind of system you are actually working with. Probabilistic systems are powerful and genuinely useful in a project management context but they require a different relationship than deterministic tools. They require verification, defined boundaries, and a clear-eyed sense of where fluency ends and comprehension never began.

The project managers who will use AI best are not the most enthusiastic or the most skeptical. They are the ones who understand the mechanism — and build their workflows accordingly.

References

Metaphors of AI indicate that people increasingly perceive AI as warm and human-like — Communications Psychology, Nature (2026)

Generative AI Literacy: A Comprehensive Framework — arXiv (2025)

Anthropomorphism in AI — AI & Society, Taylor & Francis (2020)

The benefits and dangers of anthropomorphic conversational agents — PMC (2025)

Rethinking AI anthropomorphism: A holistic conceptualization — ScienceDirect (2025)

Probabilistic and Deterministic Results in AI Systems — Gaine

Mindsets and mirrors: How growth mindsets shape anthropomorphism in AI — Psychology & Marketing, Wiley (2024)

Scoring key

Add up your points from the self-assessment above. Each B answer = 1 point. Maximum score is 5.

Score: 0 – 1

“Intelligence” thinker

You lead with the output, and that is genuinely useful — it drives adoption, keeps momentum, and surfaces possibilities others miss. The risk is that probabilistic output can be confidently wrong in ways that are hard to detect. Your blind spot is verification: you may trust fluent output without a habit of checking it.

Practical step:For any AI output that informs a project decision, name one way you will verify it before acting on it — a subject matter expert, a primary source, or a second method.

Score: 2 – 3

Mixed orientation

You have instincts from both camps. You appreciate what AI can do but sense when something needs a closer look. This is productive, as long as it is intentional. The risk is inconsistency — applying scrutiny to some outputs and not others without a clear principle for when each is warranted.

Practical step:Define in advance which project tasks require verification and which can tolerate probabilistic output. Make it a rule, not a judgment call in the moment.

Score: 4 – 5

“Artificial” thinker

You understand how the system works, and that protects you and your project from misplaced trust. The risk runs the other way — over-scrutiny can lead to under-use, and you may leave real capability on the table by applying verification overhead to tasks that do not need it.

Practical step:Make a short list of project tasks where probabilistic output is fine — drafting, brainstorming, summarizing low-stakes content — and commit to using AI there without over-verifying.