The expectation placed on product managers in 2026 is significantly different from the expectation that existed three years ago. PMs are now expected to lead AI feature development, evaluate AI output quality, write requirements that accurately describe AI behaviour, and communicate AI strategy to stakeholders who range from board members to engineers to end users. None of this was in the job description three years ago. Most of it was not taught in the product management courses and frameworks that most PMs were trained on.
The gap between what is expected of PMs and what most PMs were prepared for is real and growing. Closing it does not require turning your PM team into data scientists. It requires developing three specific levels of AI literacy, in a specific sequence, using a practical training approach that connects learning directly to the AI work your team is actually doing.
This article describes that approach — including what to teach, in what order, and how to structure the training so that the learning sticks.
What AI Literacy Actually Means for a Product Manager
The confusion about what AI literacy means for PMs is itself a significant problem. Many companies send their PMs to data science courses, AI engineering bootcamps, or machine learning fundamentals programmes — and find that the PMs come back more confused rather than more capable.
The reason: those courses teach the technical foundations of how AI works. But PM AI literacy is not about how AI works technically. It is about how to work effectively with AI as a product manager — how to make good decisions about AI features, write good AI product requirements, evaluate AI output quality from a product perspective, and communicate AI strategy in a way that builds confidence rather than confusion.
These are different capabilities from data science literacy. They require different training content, different pedagogical approaches, and different success criteria.
PM AI literacy has three levels. The levels are sequential — each one builds on the previous — and the appropriate investment in each level depends on the PM’s role and the degree to which their work involves AI.
Level One: Foundational AI Literacy — Every PM Needs This
The first level of AI literacy applies to every product manager, regardless of whether they are directly responsible for AI features. It is the baseline vocabulary and conceptual framework that allows a PM to participate meaningfully in AI-related conversations — in meetings with engineering, in discussions with leadership, in conversations with customers about AI capabilities.
What to cover:
The vocabulary. What is a prompt, a model, a context window, a token, a hallucination, a fine-tuned model, a retrieval-augmented generation system? Not in the technical depth that an engineer needs — but at the level that allows a PM to understand what an engineer is describing and to ask informed follow-up questions.
The categories of AI. What does AI do well? What does it do poorly? What types of problems are good candidates for AI solutions, and what types are not? This understanding is the foundation of AI opportunity identification — a core PM capability.
The failure modes. How does an AI system fail? What does a hallucination look like in the context of a product feature? What happens when a model is trained on outdated data? What are the failure modes specific to the use cases most relevant to your product?
The user perspective on AI. How do users experience AI features? What creates trust in AI output and what destroys it? What are the common user behaviours when AI fails — override, abandonment, workaround? This user perspective is what connects AI technical knowledge to PM product thinking.
How to deliver it:
A single day of focused training — half day of facilitated content, half day of practical exercises applied to your actual product. Not online modules. Live, interactive, facilitated by someone who has worked directly on AI product development. The practical exercises should be specifically designed around your product’s AI opportunities, not generic AI scenarios.
Success criteria:
Every PM can explain, without reference to materials, what your company’s AI features do, how they work at a conceptual level, what their failure modes are, and how you are measuring their performance. Every PM can participate in an AI-related meeting with engineering or leadership without needing concepts explained from the start.
Level Two: AI Product Development Literacy — For PMs Who Own AI Features
The second level applies to PMs who are directly responsible for AI features — who are writing AI product requirements, evaluating AI output quality, and making prioritisation decisions about AI initiatives.
What to cover:
Writing AI feature requirements. How is an AI PRD different from a standard PRD? What specific sections need to be included — user job, AI-specific behaviour, output quality criteria, fallback UX design, evaluation methodology, risk assessment? What are the most common AI PRD failure modes, and how do you avoid them?
AI feature evaluation. How do you define what “good” looks like for an AI feature in concrete, testable terms? How do you design a test set that allows you to evaluate AI output quality systematically rather than impressionistically? How do you interpret model performance metrics — precision, recall, confidence scores — from a product perspective?
Fallback UX design. For every AI feature, there is a state in which the AI cannot produce a confident answer or produces a clearly incorrect answer. How do you design the user experience for that state? What are the patterns that maintain user trust when AI fails, and what are the patterns that destroy it?
AI prioritisation frameworks. How do you apply RICE or ICE scoring to AI features, accounting for the additional factors that AI introduces — data feasibility, failure cost, evaluation complexity? How do you sequence AI initiatives across the automation, augmentation, insight, and experience categories?
How to deliver it:
Two to three days of intensive training, spread across a quarter to allow application and reflection between sessions. Day one covers AI PRD writing, with participants drafting an AI PRD for a real product initiative and receiving detailed feedback. Day two covers evaluation methodology, with participants designing an evaluation framework for a real AI feature. Day three, several weeks later, reviews the application of both skills in real work and addresses the questions that emerged.
Success criteria:
Every PM at this level can write an AI PRD that an experienced engineering team can build from without significant clarification. Every PM at this level can design an evaluation framework for their AI features and interpret the results. Every PM at this level has a clear process for escalating AI quality concerns and making confident decisions about whether an AI feature is ready to ship.
Level Three: AI Strategy Literacy — For Senior PMs and Product Leaders
The third level applies to Heads of Product, senior PMs, and CPOs who are responsible for the AI strategy of a product or a product portfolio — not just individual AI features but the overall AI direction, investment prioritisation, and competitive positioning.
What to cover:
AI opportunity identification. How do you systematically map the AI opportunity landscape for your product? How do you categorise opportunities by type, prioritise them by ROI-to-risk ratio, and sequence them to build capability while delivering business value?
AI competitive analysis. How is AI shifting the competitive dynamics in your product category? Which competitors are using AI to build sustainable advantages and which are building commodity features? What does this mean for your AI investment priorities?
AI investment justification. How do you build a business case for an AI initiative that will hold up to board scrutiny? What metrics and evidence do you need, and how do you present uncertainty honestly while maintaining investment confidence?
AI governance and risk. What are the ethical, regulatory, and reputational risks of your AI initiatives? How do you design a governance framework that manages those risks without creating so much friction that it prevents execution?
How to deliver it:
A combination of structured sessions and applied work. One to two days of facilitated content — ideally including participants from commercial and engineering leadership to build shared strategic language — followed by a structured assignment to produce or review the company’s AI strategy document using the frameworks from the training.
Success criteria:
The product leader can present the company’s AI strategy to the board with confidence, including the opportunity prioritisation rationale, the key risks and mitigations, and the metrics by which AI investment will be evaluated. The product leader can make confident AI investment prioritisation decisions without always needing external advisory input.
What Makes AI Training Stick — and What Does Not
The most common reason AI training does not produce lasting behaviour change is that it is disconnected from the work the participants are actually doing. Generic AI courses cover AI concepts in the abstract. Abstract concepts do not change how people approach specific decisions.
AI training sticks when it is anchored to real work. Specifically: when the exercises use real products, real features, and real data. When participants leave each session with a work product they can use immediately — an AI PRD for a feature currently in planning, an evaluation framework for a feature currently in development, a section of the AI strategy they are responsible for. When the facilitator has worked on real AI products and can connect the training content to specific, recognisable patterns from practice rather than from theory.
The format that produces the highest retention: one session per month, each with a specific work product output, each building on the previous. Three sessions over a quarter is more effective than three days in a row — because it creates spaced repetition and allows participants to apply and reflect between sessions.
Ready to build AI product management capability in your team? Book a strategy call at elvanis.com to discuss a custom AI training programme. We design and deliver AI literacy training for product teams across MENA and the UK.
Sally Abas is the founder of Elvanis and leads AI Advisory and Training programmes across MENA and globally. She has trained 150+ executives and product leaders.
Related reading: The Executive’s Guide to Leading AI Change | AI Strategy for SaaS Companies in 2026
