Introduction

Data and AI jobs often feel like a messy deck of cards. Every few years, a new role gets added, another one disappears, and two others somehow merge. And suddenly, you’re left wondering: what’s the real difference between a Data Scientist, an ML Engineer, and an AI Engineer? And when did the Data Analyst and the Data Engineer merge into a new card, the Analytics Engineer? And more importantly: who keeps changing the cards while the game is being played?

In this guide, we turn the Data & AI organization into a poker game. Each role becomes a card with its own power, seat at the table, and market value. The goal is simple: understand why the deck keeps changing, what each card does, and how to build the right hand for your strategy, team size, and stage of maturity.

Enough talking, let’s first see what forces keep reshuffling the deck 🎴!

---

Shuffling the deck 🔀

Data and AI jobs are not created randomly. They usually appear when the technology stack, the way teams work, or the problems companies want to solve change enough to require new skills and new ownership.

I personally group the forces reshuffling the deck into three main ones:

• The stack keeps changing: IT has always gone through cycles of bundling and unbundling. At one point, companies built and managed almost everything themselves, which required large technical teams just to keep systems running. Then cloud platforms bundled much of that complexity. Tomorrow, the stack may become more modular again. Every time this cycle turns, roles disappear, merge, or specialize.
• Data/AI is becoming a product: Data and AI are no longer side projects scattered across the company. More companies now treat their Data/AI stack as a core product, just like their ERP or website. That means real users, ownership, roadmaps, and measurable outcomes. Roles naturally move closer to product and business.
• AI is merging ML and software engineering: Classic ML was heavily focused on maths, models, understanding research publications, and hyperparameter optimization. LLM systems bring much more software engineering into the picture. This overlap is exactly why new roles such as AI Engineer are emerging.

---

The Four Suits

Now that we’ve seen the forces reshuffling the deck, here are the cards we’ll be playing with today.

At first, it still looks like a mess of titles and shifting roles. So let’s bring some structure to the table.

We split the deck into four suits, each representing a different family of roles in a Data & AI team: first, strategy and governance; second, platform and architecture; third, analytics and product; and last but not least, machine learning and AI.

---

♥️ Strategy and Governance

In poker, before anyone plays a hand, the rules must be clear and the table must be trusted. In the Data & AI world, this suit brings order, alignment, and credibility. These roles decide how the game is played, what counts as a win, and how to keep players accountable.

♥️ Chief Data / AI Officer: The Ace of Hearts

• Role: Defines the company’s Data & AI strategy, operating model, governance, and ROI.
• Powers: Executive alignment, investment prioritization, budget ownership, operating model, governance, and accountability.
• Frameworks: Roadmap, data catalog, DAMA practices, compliance frameworks.
• Proportion: 1 per company (or 1 per Business Unit in very large groups).
• Stack value: 🔹🔹🔹🔹🔹 5/5

And in real life? The job title varies with context. Some companies go “CDO” to stress governance and data value. Others go “CAIO” to stress AI adoption. In practice, both hats can cover the same responsibilities, just with a different emphasis.

♥️ Data Governance Lead: The King of Hearts

• Role: Implements data governance, quality, security, compliance, catalog within the company or a specific business scope.
• Powers: Stewardship model, governance, policy enforcement, risk assessment, data lineage.
• Frameworks: Quality and governance frameworks, data management tools (e.g. Collibra, Alation), etc.
• Proportion: 1-2 per business data domain or per company.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? The need for this role is rising fast with regulatory pressure. Frameworks like GDPR in Europe, BCBS 239 in banking, or privacy laws like the California Consumer Privacy Act (CCPA) make governance not just best practice, but necessary for legal or regulatory compliance.

♥️ Data Steward: The Jack of Hearts

• Role: Ensures data definitions, quality, and usability within a specific business scope.
• Powers: Business glossary, quality rules, access reviews, semantic arbitration.
• Frameworks: Data catalog, anomaly detection tools, MDM.
• Proportion: 1-2 per business data domain (e.g. customer, product, finance).
• Stack value: 🔹🔹🔹 3/5

And in real life? For example, in one business unit, “active customer” might mean something different in another. The Data Steward ensures one consistent definition.

P.S. If you really know how to play poker, there are no Queens in this deck. It happens 😉

---

♦️ Platform and Architecture

No game without a table, no bets without poker chips. In Data & AI, this suit lays the technical foundation: how data is stored, moved, secured, and made available. These roles ensure that the platform is solid, scalable, and cost-efficient so the rest of the players can focus on winning hands.

♦️ Data Architect: The Ace of Diamonds

• Role: Designs data architecture, patterns, and standards, balancing cost with performance.
• Powers: Chooses the target architecture and defines security and integration models.
• Tools: ER/Studio, Sparx Enterprise Architect, Lucidchart, etc.
• Proportion: 1-2 per platform.
• Stack value: 🔹🔹🔹🔹🔹 5/5

And in real life? In some organizations, a Data Modeler supports the Data Architect. The architect sets the big picture (e.g. platform, integration, governance), while the modeler focuses on logical and conceptual models: turning business entities into schemas the architect embeds in the overall design.

♦️ Data Engineer: The King of Diamonds

• Role: Builds data pipelines and keeps the data platform running.
• Powers: Ingestion, transformation, orchestration, reliability, and cost control.
• Tools: Cloud platform ETL/ELT, dbt, Airflow, Airbyte, Spark, Kafka, SQL, Python, etc.
• Proportion: 1-8 per product or platform.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? Data is spread across many Business Units and systems (e.g. CRM, ERP, websites). The Data Engineer is the plumber of the team, connecting systems and keeping data flowing.

♦️ Data/MLOps Engineer: The Jack of Diamonds

• Role: Ensures reliability, deployment, and observability of the data platform / products.
• Powers: CI/CD, data testing, monitoring, FinOps.
• Tools: MLflow, Feast, GitHub Actions, Jenkins, Docker, Kubernetes, Grafana, etc.
• Proportion: 1-2 per platform.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? This is the scale-up card. It’s what takes you from a notebook demo to a model that real customers can use every day. Without DataOps/MLOps, systems become harder to operate and AI stays stuck in experiments.

---

♠️ Analytics and Product

In poker, the best players don’t just look at their own cards. They read the table, the odds, and the story unfolding. In the Data & AI world, this suit plays exactly that role: turning questions into insights, raw data into reliable products, and analyses into decisions.

♠️ Data/AI Product Manager: The Ace of Spades

• Role: Drives data products from discovery to run, maximizing value and adoption.
• Powers: Discovery, backlog, prioritization, experimentation, go-to-market.
• Frameworks: Scrum, SAFe, roadmap, usage tracking, documentation, associated tools (e.g. Jira, Confluence, Notion, Figma), etc.
• Proportion: 1 per product, or 1 for 2-3 smaller ones.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? Titles vary a lot across companies. Sometimes this role is called Product Owner or even, mistakenly, Project Manager. The labels change, but the mission stays the same: ensure that data and AI initiatives deliver clear business value.

♠️ Data Analyst: The Jack of Spades

• Role: Turns business questions into actionable answers.
• Powers: Exploration, visualization, KPI design, storytelling, A/B tests.
• Tools: SQL, Power BI, Tableau, Looker, light Python or R, etc.
• Proportion: 1-3 per product.
• Stack value: 🔹🔹🔹 3/5

And in real life? The definition of Data Analyst can change widely. In some companies, it’s a technical BI profile with SQL and visualization skills. In others, it’s closer to a Business Analyst, lighter on data and stronger on business context.

♠️ Analytics Engineer: The King of Spades

• Role: Bridges Data Engineer and Data Analyst, delivering analytics-ready data.
• Powers: Semantic modeling, pipelines, visualization, reliable datasets.
• Tools: dbt, tools of Data Engineer + tools of Data Analyst.
• Proportion: 1-3 per product.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? This role popped up to solve a recurring frustration: Data Analysts waiting endlessly for usable data, and Data Engineers building pipelines without understanding business context. The Analytics Engineer closes that gap, creating reliable and business-ready datasets that both sides can actually use.

---

♣️ Machine Learning and AI

In poker, once the basics are mastered, players start pulling advanced tricks: bluffing, reading micro-signals, or optimizing bets. In Data & AI, this suit is about the technical experts who push beyond reporting into predictions, optimizations, and now generative AI.

♣️ Data Scientist: The Ace of Clubs

• Role: Builds predictive, prescriptive or generative models for focused use cases.
• Powers: Model training, feature engineering, experimentation, evaluation, ML framing.
• Tools: Python, sklearn, pyTorch, TensorFlow, MLflow, etc.
• Proportion: 1-3 per product.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? In many companies, Data Scientist is a fuzzy card. Sometimes it means “fancy Data Analyst” (SQL, dashboards, light stats). Other times, it means a PhD-level researcher building models in notebooks with no production path. This ambiguity is exactly what created the need for new cards.

♣️ Machine Learning Engineer: The King of Clubs

• Role: Deploys to production, optimizes latency, cost, and robustness.
• Powers: Model serving, real-time features, GPU/CPU optimization.
• Tools: TensorFlow Serving, TorchServe, cloud platforms (SageMaker, Vertex AI), Docker, etc.
• Proportion: 1-3 per product.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? The role of MLE became prominent during the 2010s as a response to the “Data Scientist gap”. Companies realized that DS use cases (e.g. forecasting, scoring, optimization) needed dedicated engineers to make them reliable, fast, and cost-efficient in production. The MLE is the pragmatic counterbalance to the DS’s research focus.

♣️ AI Engineer: The Jack of Clubs

• Role: Builds and integrates AI systems, especially around LLMs and agentic workflows.
• Powers: LLM serving, RAG, fine-tuning, prompt design, evaluation, agent orchestration.
• Tools: LLM provider APIs, LangChain, LangGraph, LlamaIndex, vector databases, vLLM, etc.
• Proportion: 1-3 per product.
• Stack value: 🔹🔹🔹🔹 4/5

And in real life? The AI Engineer card emerged with the rise of large language models. Data Scientists were often too research-focused, and ML Engineers were more concerned with deployment of classic predictive or prescriptive models. But with LLMs came new software engineering challenges: chaining prompts, building wrappers, managing vector databases, and deploying agentic systems reliably in production.

---

Build your hand

In poker, winning is rarely about holding one magic card. It’s about playing the right combination of cards for the right situation. Data & AI teams work the same way: the best hand depends on your company size, maturity, and goals.

Let’s lay down three example hands:

1️⃣ Starter hand
~50-300 employees, single team, no dedicated platform yet

• 1 card of ♠️ Data/AI Product Manager
• 1 card of ♦️ Data Engineer
• 1-2 cards of ♠️ Data Analyst

What can this hand really do? Help a small or mid-sized company get started with Business Intelligence, establish a single source of truth, deliver its first dashboards, and show initial ROI.

---

2️⃣ Scale hand
~300-2,000 employees, one BU, dedicated analytical platform in production

What can this hand really do? Run a modern data platform for a mid-sized or large company, combine data from several sources, support dozens of dashboards and data products, and deploy Data Science and Generative AI use cases into production workflows.

---

3️⃣ Enterprise Data & AI hand
2,000+ employees, multiple BUs, formalized governance across platforms

What can this hand really do? Run Data & AI at enterprise scale across many business units and data domains. It can support large analytics estates, industrialize advanced AI use cases, and operate them reliably in production.

---

Conclusion

I’ll leave you with a few lyrics from a song 🎧 I really like:

You’ve got to know when to hold ’em,
Know when to fold ’em,
Know when to walk away,
And know when to run.

Kenny Rogers – The Gambler

And here’s the twist: in poker as in Data & AI, once you understand the cards, the suits, and how to play your hand, the win is no longer about luck. You already know how to play to win 🃏!

👉 You’ve now seen the poker cards, the roles that make up a Data & AI team. But cards are nothing without the right tools on the table. Curious about the tooling side? Data Team Maturity: Understanding Tooling Levels.