MODULE 2 OF 6

The Four AI Specificities

Understanding the technical triggers that create unique management friction in AI projects.

2.1 Why AI is Different

Traditional IT projects follow deterministic logic: define requirements, build solution, deliver predictable outcome. AI projects are fundamentally different. They are governed by four inherent technical specificities that transform the nature of project friction.

"The other thing is the inherent problem of just the sheer complexity in a variation of AI. So, there's a big mindset shift from deterministic solutions to more stochastic meaning. AI introduces a lot of variations in edge cases that people don't normally understand."
— P21, Project Manager

These four specificities act as technical triggers that, when combined with cross-cultural dynamics, create the causal chains of friction we'll explore throughout this module.

170.8%
Stochastic Uncertainty → ACT
128.0%
High Expectations → INQUIRE
104.2%
Autocatalysis → STANDARDIZE
29.8%
Process Disruption Prevalence

2.2 Stochastic Uncertainty (Non-Deterministic Output)

⚡ STOCHASTIC UNCERTAINTY

What It Is

Unlike traditional software engineering where the answer is known in advance, AI solutions are non-deterministic. The same input can produce different outputs, and success is probabilistic rather than guaranteed.

"Data science is that like a lot of times, we don't know. It's different. It's like you know... We don't know the answers, right? It's not always like... It's like an engineering problem where you know you just need the requirements and then you implement something to those requirements."
— P5, Data Scientist
"With the current kind of projects that people are building with large language models... It's such a general purpose technology. It's very... you can't test for all possible things. There's many ways that it can go wrong."
— P33, AI Engineer

The Friction It Creates: Baseline Erosion

Because model performance fluctuates unpredictably due to "data spikes" or latent biases, practitioners find it impossible to maintain a stable project forecast. This creates constant firefighting.

⚡ Stochastic Uncertainty 📉 Baseline Erosion ACT (170.8%)

Cultural Dimension

Cultural backgrounds significantly mediate how individuals cope with this ambiguity. Some cultures have higher uncertainty tolerance than others, requiring project managers to align risk tolerances across the team.

"AI generally is very uncertain. So there's this question is, how comfortable are you with uncertainty?"
— P21

2.3 Autocatalysis (Technical Velocity)

🚀 AUTOCATALYSIS

What It Is

In chemistry, autocatalysis is a process where a product of a reaction acts as a catalyst for the reaction itself. In AI, each breakthrough fuels further advancements at an exponential rate. The cycle becomes self-reinforcing.

"Every day there's something new. Every day there's a new discovery. There's a new something. Everything just happening so quickly that a lot of times we just feel like constant change and constant churning. And so a mixed communication... very difficult."
— P9, Engineering Lead

The Friction It Creates: Knowledge Asymmetry

This technical velocity makes baseline alignment a primary management hurdle. Team members fall behind the current state of the art, and this gap is exacerbated in multicultural teams where communication is already complex.

"And I noticed this difference between Brazil that were we were a bit more advanced... is difficult to us to tell that way so directly..."
— P1 and P2, Brazilian Engineers
🚀 Autocatalysis 🧠 Knowledge Asymmetry TRAIN / INQUIRE

Practical Impact

The constant churning means that skills learned six months ago may already be obsolete. Organizations must build continuous learning into their DNA rather than treating training as a one-time event.

2.4 Process Disruption (Non-Linear Workflows)

🔄 PROCESS DISRUPTION

What It Is

AI development is inherently disruptive to established business processes. It requires a fundamental mindset shift from deterministic to probabilistic solutions, which often triggers systemic resistance within organizations.

"That's usually when the resistance gets a bit up and people think, oh, wait, we also need to change some parts of our processes. We don't want that."
— P29, Project Manager
"This AI machine learning solution requires a certain degree of collaboration between the sales recruiters. However, the sales recruiters themselves tend to think really competitively. So they're like, yeah, if everyone has this information. There's no reward."
— P29

The Friction It Creates: Organizational Resistance

Employees perceive AI implementation as a threat to established processes, job security, or individual incentives. This creates cultural resistance that can derail projects.

🔄 Process Disruption ⚔️ Cultural Resistance ACT (161.2%) + STANDARDIZE (116.4%)

Most Prevalent Stressor

Process Disruption is the most frequently cited stressor in the dataset, appearing in 29.8% of analytical units. It requires managers to function as "Cultural Bridges," harmonizing diverse team mentalities while formalizing new workflows.

2.5 High Customer Expectations (The "AI Knows All" Perception)

🎯 HIGH CUSTOMER EXPECTATIONS

What It Is

AI professionals frequently encounter an "Adversarial Client Gap" where stakeholders project unrealistic dreams onto a technology they perceive as a panacea rather than mathematics.

"Uh, first, you know, this is like for developers, for data scientists or researchers, this is like the customer is like the enemy because maybe he has a lot of ideas, a lot of like, the ideas that are in the cloud... and that's something that we need to put all the things on the floor and say OK. We can do this. But we can't do the other things."
— P7, AI Developer
"Client would come from his head and be like I need this and then you have to like lower his dreams and be like this is so for us we just use maps and graphs and we use a little bit of research to show that this is what's possible, this is what's not possible."
— P32, AI Engineer

The Friction It Creates: Anthropomorphic Projection

Stakeholders perceive AI as an all-knowing entity rather than a probabilistic tool. This creates a Trust-Understanding Gap where stakeholders lack patience for the iterative, non-linear development cycle.

🎯 High Expectations 🤖 Anthropomorphic Projection INQUIRE (128.0%)

Cultural Dimension

This gap is widened by cultural differences in how "value" is defined and how commercial approaches are handled:

  • Minimalism vs. Over-achieving (P13)
  • Efficiency vs. Sustainability (P1, P2)
  • Compliance vs. Business value (P40)

2.6 The Causal Chain Model

These four specificities don't operate in isolation. They create distinct causal loops that explain why traditional IT governance fails in AI projects.

Chain A: The Reactive Loop

Stochastic Uncertainty → Baseline Erosion → ACT (170.8%)

Managers trapped in continuous manual intervention, replacing strategy with firefighting.

Chain B: The Re-education Loop

High Customer Expectations → Trust Gap → INQUIRE (128.0%)

Constant stakeholder education to recalibrate unrealistic dreams with technical reality.

Chain C: The Governance Loop

Autocatalysis → Temporal Misalignment → STANDARDIZE (104.2%)

Attempting to decelerate technical velocity through documentation and phased deployment.

Chain D: The Integration Loop

Process Disruption → Cultural Resistance → ACT (161.2%) + STANDARDIZE (116.4%)

Managers as cultural bridges, mediating between diverse mentalities while formalizing workflows.

2.7 Resilience Buffers: The Circuit Breakers

The research identified 279 negative cases where technical triggers did NOT lead to friction. These reveal two critical resilience mechanisms:

🛡️ Professional Insulation

Strong professional IT norms act as a buffer against Process Disruption. By enforcing strict protocols and "ground rules," managers bypass latent cultural resistance.

🏗️ Architectural Simplicity

Practitioners neutralize Stochastic Uncertainty through strategic architectural choices (e.g., RAG frameworks). This bounds probabilistic variance within a verifiable environment.

"We need this responsibility of building the model. So we need us... that's our main concern. We need to remove this kind of biases and these kind of issues."
— P2, on ethical responsibility

📌 Module 2 Key Takeaways

🔗 Connect to Your Assessment

Your assessment results will show how your organization responds to these four specificities:

→ Take the assessment to see your profile

✍️ Reflection Exercise

Think of an AI project you've worked on or observed:

  1. Which of the four specificities caused the most friction?
  2. Did you see any of the causal chains in action?
  3. What resilience mechanisms (if any) were present?

Take notes—you'll use these insights in Module 6 when building your action plan.