For years, online shopping meant searching, comparing, clicking “Buy” and paying.
Now AI agents are starting to do more of this work for us.
You can give an agent a simple goal:
“Find running shoes under $150 that can arrive by Friday.”
The agent can search, compare products, build a cart and increasingly help complete the purchase.
This is called agentic commerce.
What does an AI agent need to do?
A shopping agent needs several capabilities:
Find products and compare offers.
Build a cart and place an order.
Pay securely with the user's permission.
Different protocols are being developed for these tasks. They are not all direct competitors.
They often work at different levels.
MCP connects AI agents to external tools and data.
UCP supports agentic commerce between shoppers, businesses and payment providers.
ACP supports commerce and checkout flows between agents, users and businesses.
AP2 focuses on secure and authorized agent payments. It was originally introduced by Google and moved to the FIDO Alliance in 2026.
The big idea: agentic commerce is becoming a stack, not a single protocol.
A reality check
In September 2025, OpenAI launched Instant Checkout in ChatGPT.
It showed that an AI assistant could move beyond recommendations and help users buy products directly.
But the first version also showed how difficult agentic commerce is.
A good shopping experience needs more than a “Buy” button:
Multi-item carts
Inventory and delivery
Returns
Merchant integration
Payment security and user authorization
The lesson is simple: a protocol can work well while the overall product still fails to offer enough value.
From discovery to the full shopping journey
The next step is not only about recommending products.
AI platforms are starting to cover more of the journey:
Google’s Universal Cart is an example of this direction. It lets users collect products from different merchants
across services such as Search, Gemini, YouTube and Gmail.
The goal is clear: the AI interface can become a place not only to ask what to buy,
but also to prepare and execute the purchase.
The real challenge: trust
The hardest question is not “Can an AI agent shop?”
It is “Can we trust it to act for us?”
Did the user really authorize the purchase?
Did the agent respect the budget and other limits?
What if the price changes?
Who is responsible if the agent makes a mistake?
This is why secure authorization and traceable payments are becoming so important.
The system must connect the user's intention with the agent's action and the final transaction.
What should online businesses do?
Make product information clean, accurate and machine-readable.
Keep prices, stock, variants and delivery information up to date.
Do not depend on only one protocol or AI platform.
Make products easy for agents to understand and compare.
Protect the customer relationship, even when an agent handles the transaction.
Build strong trust, security and customer service.
The strategic question
Traditional e-commerce was built around people visiting websites.
Search, mobile and marketplaces already changed that model.
AI agents may change it again.
If an agent can compare thousands of products, build a cart and eventually pay,
the customer may no longer need to visit a merchant's website.
So the real question is not only:
“Will AI agents shop for us?”
It is:
“Who will control the customer when the customer stops shopping directly?”
The bottom line
Shopping is moving from clicks to goals.
We describe what we want, and agents increasingly handle the work.
The winners will not necessarily be the companies with the most impressive AI demos.
They will be the businesses whose products are easy for agents to find, understand, trust and buy.
Mastering Data Analysis in 2026: From basics to job-ready
Mastering Data Analysis in 2026: From basics to job-ready
From far away, some fields look difficult and even a bit scary. But when you really take the time to explore them, they start to make sense. You begin to see a clear logic, almost a kind of simple beauty in how they organize things. Data analysis is undoubtedly one of those fields.
Demand for skilled data analysts will rise through 2026; however, they will become more specialized in their skills. Companies are moving away from needing someone who can create charts, and instead towards someone who can actually ask appropriate questions, explore the available data and develop concrete decisions from that data. This itinerary has been prepared to help you meet those expectations through meticulous planning, while avoiding false claims that you must have a complete understanding of deep learning capabilities in order to become a successful data analyst.
Why a roadmap in 2026?
The world of learning data has grown very fast. There are many platforms, hundreds of courses, and all kinds of certificates. And yet, many learners still don’t know where to begin.
It’s tempting to follow the trend and jump straight into the use of a given tool or language. But in doing that, we often forget something important: the basics of data analysis. These foundations, both theory and practice, are clear, structured, and very valuable on their own.
This guide traces a progressive path, structured across five levels. Each level builds on the previous one. Every skill has a reason to exist in the day-to-day work of a data analyst.
Level 1 Foundations: understand before you calculate
Everything begins with the data itself. Before writing a single line of code or opening a dashboard, an analyst must know what they are working with.
Data types are the first cornerstone. Structured data (relational tables), semi-structured (JSON, XML), or unstructured (free text, images) each format calls for different methods and imposes specific constraints. Confusing one type with another is one of the most common mistakes which could have an expensive cost.
Then comes descriptive statistics, which is far more than a set of formulas to memorize. Understanding why the mean can be misleading in the presence of outliers, intuitively understanding what variance reveals about the dispersion of a phenomenon, distinguishing a normal distribution from a skewed one, these are analytical reflexes built through practice and curiosity.
Finally, Excel and spreadsheets remain, despite everything, a reference tool in the vast majority of organizations. Mastery of pivot tables, conditional formulas, and aggregation functions is an immediately marketable skill.
Level 2 Programming and data: speaking the language of machines
Here the work becomes more technical but also more liberating. Because programming means breaking free from the limitations of graphical interfaces. It's worth noting that in the age of AI and vibe coding, programming is no longer a barrier like it used to be.
SQL remains the universal language of the analyst. Joins, aggregations, subqueries, window functions. Mastering SQL allows you to extract information from any relational database with surgical precision.
Python has established itself as the second pillar. With pandas, numpy, and matplotlib, it enables manipulation of datasets of any size, automation of repetitive processing, and production of customized visualizations. It is also a gateway to more advanced analyses at level 4. The Excel and Python course I developed on LinkedIn Learning illustrates precisely how these two tools complement rather than oppose each other.
R, optionally, offers a particularly powerful statistical environment with ggplot2 and the tidyverse. It is widely present in academic circles and in certain sectors such as healthcare or finance.
Level 3 Analysis and visualisation: turning data into meaning
This is the heart of the profession. An analyst who knows how to extract data but cannot stage it for decision-makers accomplishes only half their mission.
Data visualisation It is a communication discipline. Choosing the right chart type for the message to be conveyed, respecting the principles of visual perception, building a dashboard that guides the eye toward what matters all of this is learned. Power BI and Tableau are the industry standards. Seaborn and Plotly allow fine-grained customization in Python.
Statistical tests are the tool that distinguishes a real trend from an artifact of chance. Considered as the hardest parts to master, but really worth it and make all the difference. The Student's t-test, chi-squared, ANOVA, the notion of p-value and confidence intervals these concepts allow the analyst to answer a fundamental question: "Is what I'm observing statistically significant?" Without them, any analysis remains fragile.
Dimensionality reduction with Principal Component Analysis (PCA) at the forefront is a theoretically rich and practically powerful technique. It allows simplification of complex datasets, visualization of latent structures, and preparation for subsequent analyses. PCA is indispensable. Other methods such as t-SNE or MCA (for categorical variables) complete the toolkit.
Level 4 Analytical modelling: going further with regression
This level marks the boundary between the data analyst and the data scientist. And it is important here to clarify what I mean by modelling within the scope of this roadmap.
Business metrics are often underestimated in technical curricula. An analyst must understand what a KPI truly measures, how to build a cohort analysis, what a conversion funnel analysis or customer lifetime value (LTV) estimation reveals. This business vocabulary is what allows the analyst to be understood by non-technical teams.
Regression, in its linear and logistic forms, constitutes the acceptable boundary with machine learning for a data analyst. It enables modelling of causal relationships, making measurable predictions, and evaluating model quality through indicators such as R², residuals, or the confusion matrix. Ridge and Lasso regression add a useful regularization dimension when data is noisy.
A deliberate boundary: advanced machine learning (random forests, gradient boosting, neural networks) and deep learning fall outside the scope of this roadmap. Not because they are without interest, but because they belong to a different profession "the data scientist" with its own theoretical and computational requirements.
Level 5 Agentic AI for data analysis: the new frontier
The final level reflects a profound shift in what is expected of a data analyst in 2026. Artificial intelligence is no longer a tool reserved for data scientists, it has become a daily lever for the analyst who knows how to use it with precision.
Prompt engineering is the entry point. Knowing how to formulate clear, structured, and context-rich instructions to a large language model (LLM) allows an analyst to interrogate datasets in natural language, generate transformation code, interpret statistical outputs, or draft analytical narratives at a speed that was unimaginable just a few years ago. This skill is not about replacing analytical thinking, it is about amplifying it.
Beyond isolated prompts, agentic workflows represent the next step. An agentic system is capable of autonomously chaining a sequence of tasks: extracting data from a source, cleaning it, running an analysis, generating a report, and sending it to the right stakeholder all without manual intervention at each step. Platforms such as n8n, LangChain, or Claude with tool use, make this orchestration accessible to analysts who are not software engineers.
Mastering this level means understanding when to delegate to an agent, how to design a reliable workflow, and how to maintain human oversight where judgment and interpretation are irreplaceable.
What this roadmap does not say
It does not say you must learn everything before you start analyzing. A competent data analyst is built through action, on real data, facing concrete problems. This map is a guide, not an exhaustive prerequisite.
It does not say that machine learning is out of reach or without interest either. It simply says it belongs to a different territory and mastering the skills figuring in the roadmap is already entirely sufficient ambition to build a solid career as a "Data anayst".
A final thought: the mindset that makes the difference
Beyond technical skills, the best data analysts share a common trait: disciplined curiosity. Curiosity about data, about the phenomena it reveals, about the stories it tells. And the discipline not to be swept away by seductive correlations at the expense of causal rigor. Another point I always tell my students, and which I’ve included in my LinkedIn profile banner, is to learn how to learn.
In 2026, data is everywhere. But people capable of reading it with method, questioning it with intellectual honesty, and translating it into value for organizations they remain rare.
The first four episodes gave you the theory. Now it's time to act.
We explored how AI evolved from prediction to autonomy, we learned about different types of agents, we discovered the Model Context Protocol that connects agents to real tools, and we discussed what this means for the future of work.
But theory without practice is just information. To really benefit from agentic AI, you need to see the opportunities and know how to use them.
Why the LLM brain matters so much
At the heart of every agentic system is the Large Language Model. This is the brain that makes everything work. But the LLM doesn't just generate text. In agentic AI, it plays many different roles:
Perception: The LLM reads and understands data from different sources. It can analyze emails, documents, images, time series data, or customer feedback. It sees patterns that humans might miss.
Decision-making: Based on what it perceives, the LLM decides what to do next. Should it fetch more data? Should it call a specific tool? Should it ask for human help? These decisions happen in real-time, making the agent truly autonomous.
Reasoning across domains: The same LLM can work with financial data in the morning and computer vision tasks in the afternoon. It can analyze sales trends, process images, or understand natural language. This flexibility is what makes modern agents so powerful.
The LLM is not just one component. It's the intelligence that ties everything together.
Orchestration: Making agents work together
A single agent can do useful work. But real power comes when multiple agents collaborate.
This is orchestration. One agent handles data collection. Another does analysis. A third makes decisions. A fourth communicates results. They work together like a team, each doing what it does best.
But someone needs to coordinate this team. That's the orchestrator, often another AI agent that:
* Assigns tasks to the right agents
* Manages the flow of information
* Handles errors when something goes wrong
* Ensures everyone works toward the same goal
Good orchestration turns simple agents into powerful systems.
Communication: How agents talk to each other
For orchestration to work, agents need to communicate clearly.
In traditional software, programs pass data in fixed formats. But AI agents are different. They use natural language, structured messages, and shared memory to exchange information.
This creates new possibilities:
* Agents can explain their reasoning to each other
* They can negotiate and compromise
* They can share context and learn from each other's mistakes
* They can work together even if they weren't specifically designed to
The Model Context Protocol we discussed in Episode 3 makes this communication standardized and reliable.
Opening to external tools: The real game changer
An agent that only thinks is limited. An agent that can act in the real world is transformative.
This is where external tools come in. Through protocols like MCP, agents can:
* Access databases and retrieve information
* Send emails and messages
* Update spreadsheets and documents
* Call APIs and web services
* Control other software and systems
* Make purchases or schedule appointments
The key insight: you don't need to build these connections yourself. Modern agentic platforms already connect to thousands of tools. Your job is to recognize which tools solve your specific problems.
The course uses short videos to show real opportunities. Each video focuses on one practical scenario where agentic AI creates real value. The format is simple: watch a quick lesson, see the opportunity, try it yourself.
Here's what the course covers:
Understanding perception and decision: How LLMs analyze different types of data and make smart choices. You'll see examples with time series forecasting, image analysis, and document processing.
Setting up communication: The patterns that make agent collaboration work. You'll learn when agents should share information and when they should work independently.
Connecting to external tools: Practical demonstrations of agents using real tools: databases, APIs, automation platforms, and productivity apps. You'll see which tools matter most and how to connect them quickly.
Recognizing opportunities: The most important skill is knowing where agentic AI can help. The course shows you how to look at your daily work and spot the patterns where agents can make a difference.
This isn't about theory. It's about seeing possibilities and knowing how to capture them.
Starting today
You don't need to wait for permission or perfect conditions. You can start experimenting with agentic AI right now:
Pick one repetitive task you do every week. Ask yourself: could an agent handle this? If you're not sure, that's okay. Try anyway.
Use existing platforms. Tools like Claude with computer use, n8n, or Zapier already have agentic capabilities built in. You don't need to code from scratch.
Start small. Don't try to automate your entire workflow on day one. Automate one small piece. Then add another. Each small win builds your confidence.
Think about the full cycle. Remember: perception → decision → action → communication. When you design an agentic solution, consider all four parts.
Share what you learn. When you discover something that works, tell your colleagues. Teaching others helps you understand better.
What makes this moment special
We're at an interesting point in AI history. The technology is powerful enough to be genuinely useful, but still accessible enough that anyone can learn it.
Five years ago, building agentic systems required deep technical expertise. Today, you can create powerful agent workflows with visual tools and natural language instructions.
But this window won't stay open forever. As agentic AI becomes mainstream, the early advantage goes to people who start learning now.
The tools are ready. The opportunities are real. What you do next is up to you.
Previous episodes of this series for the conceptual foundation
Try Claude.ai or similar platforms to experiment with agentic capabilities
This series started with theory. It ends with an invitation to practice. The best way to understand agentic AI is to use it. Start small. Try things. Build something.
** Start small.
** Try things.
** Build something.
The future belongs to people who act, not just people who understand.
When AI Shapes Young Minds: The Cognitive Risks of Early, Unfiltered Use
When AI shapes young minds: The cognitive risks of early, unfiltered use
Artificial intelligence is reshaping how children learn. From homework helpers to instant tutors, AI is present earlier in childhood than many of us expected. That access brings benefits: faster feedback, new explanations, and new ways to practice. It also brings a less obvious risk, one that quietly changes how young people learn to think.
Why this matters
Learning is not only about acquiring facts. It is the process of building mental tools: the patience to wrestle with a hard question, the persistence to try multiple approaches, and the habit of testing an idea by making small mistakes. When those habits are replaced by instant answers, the brain misses crucial training.
The cognitive risks of early, unfiltered AI use
Weakened problem-solving muscles. Reasoning strengthens when learners try, fail, and adjust. Instant solutions can short-circuit that cycle.
Lower tolerance for ambiguity. If an answer is always a click away, children may stop learning to hold a question and explore multiple possibilities.
Surface-level understanding. AI can produce correct outputs without the student internalizing the underlying logic.
Convergence of thought. AI trained on large data can subtly nudge many students toward similar reasoning patterns, shrinking diversity of approaches.
Dependency and decreased curiosity. Over-reliance on tools encourages seeking quick answers rather than asking deep questions.
Not all AI use is harmful “context and design matter”
AI is a tool. It can accelerate discovery when used to extend human effort rather than replace it. The risk appears when children use AI as a substitute for thinking, not as a partner for it.
Good use: AI offers hints and nudges while the learner still does the core work. Harmful use: AI supplies full solutions that the learner copies without reflection.
Practical rules for parents and teachers
Here are clear, practical steps to help children benefit from AI while protecting cognitive growth.
Require first attempts: Ask students to show their own thinking before using AI. A short sketch, rough notes, or a recorded explanation is invaluable.
Use AI as a coach, not a copier: Configure tools to provide hints, not final answers. Prompt children to try once, then request a hint if stuck.
Teach verification skills: Show how AI can be wrong and how to check outputs using logic, examples, or trusted references.
Encourage reflective prompts: After using AI, have the student explain in their own words what changed and why.
Limit easy access during practice: For tasks designed to build thinking, discourage AI use until the student has practiced independently.
Mix human feedback with AI feedback: Human coaching that focuses on process, not just correctness, preserves cognitive development.
Prioritize oral examinations for accurate assessment: Unlike multiple-choice tests or project-based evaluation, oral exams create a direct exchange that eliminates guesswork and minimizes opportunities for cheating. They also reveal each student's genuine understanding, depth of reasoning, and individual contribution.
Curriculum and policy suggestions
Schools and districts should not only adopt technology; they must define how it is used. Practical policies include:
Designated periods for “AI-free practice” during which students must work without assistance.
Assignments that require process documentation (drafts, logs, or audio reflections).
Teacher training on how to scaffold AI as a learning partner.
Assessment methods that reward reasoning steps, not just final answers.
How to explain this to kids
Simple language works best: “AI can be a smart friend, but friends shouldn’t do your homework for you. practicing and making mistakes helps your brain grow.” Build small rituals: try 15 minutes alone, then 10 minutes with a tool, then five minutes to explain what you learned.
Quick checklist for today:
Ask your child to do one homework problem without tools.
Discuss the steps they took.
Then invite them to use an AI assistant for hints and compare results.
Conclusion
AI will continue to be a helpful presence in children’s lives. The choice we face is not whether to use it, but how. If we design learning environments that force thinking before assistance, teach students to verify and reflect, and pair AI with human guidance, we can preserve and even strengthen the core habits of reasoning that last a lifetime.
Why entrepreneurs are joining the Microsoft Excel world championship
Most people think of Excel as a simple tool for budgets or reports, the kind of software that hides somewhere on a crowded desktop. Yet every year, something unexpected happens: Excel steps into the spotlight. Not as a productivity app, but as a full-scale competition watched by thousands around the world. Yes, an actual world championship where speed, logic and creativity collide in thirty intense minutes of problem solving.
It may sound surprising at first, but the atmosphere is closer to an esports arena than a quiet office. Competitors face fast-paced scenarios where they must recreate shapes, build models, break down logic puzzles or combine formulas with astonishing precision. Every five minutes, the lowest score is eliminated. The pressure is real, the energy is electric and the crowd reacts to formulas the same way others react to a perfect chess move or final-second goal.
How the competition works
Throughout the year, players join online challenges that serve as gateways to the main event. These sessions are open to anyone and give newcomers a taste of the intensity behind Excel esports. The best competitors move forward to a playoff structure, and only a small group earns a seat at the live finals in Las Vegas, where a packed audience watches them think faster than most of us type.
What truly stands out is that the finalists are far from anonymous. Many are founders, consultants or business owners who spend their days solving real problems for real clients. Andrew from Australia, Michael from Canada, Di from Ireland, their profiles look more like LinkedIn success stories than stereotypical gamers. And yet, when the countdown begins, they become athletes of logic, transforming spreadsheets into a field of strategy and instinct.
Why entrepreneurs belong in this arena
This championship is more than a contest; it is a reminder of what modern leadership looks like. Today’s founders navigate data, automation, financial models and rapid decisions every single day. Mastering the tools behind those decisions is not optional anymore. Excel may seem simple at first glance, but at high level it becomes a playground where creativity meets discipline.
When entrepreneurs step into this competition, they send a strong message to their teams and clients: “I understand the tools we use. I can think fast, adapt quickly and solve problems under pressure.” There is something refreshing, even inspiring, in seeing business leaders roll up their sleeves and compete alongside analysts and students. It breaks the myth that CEOs only delegate technical work. Here, they prove they can lead through skill, not only through title.
A path open to anyone willing to try
You do not need to be a genius or memorize every function to get started. The championship welcomes people at many levels. The online battles offer a chance to learn, experiment and progress. By repeating challenges, you slowly build the intuition that the best players share: the ability to see patterns, think in formulas and transform uncertainty into structure.
Whether you are an entrepreneur, a student or someone simply curious, this competition offers a rare opportunity to join a community that values intelligence, curiosity and technical mastery.
At its core, entrepreneurship has always been a race against time. You identify a problem, design a solution and adjust faster than everyone else. The Excel World Championship celebrates exactly that mindset. It transforms spreadsheets into a story of resilience, strategy and passion.
If this world inspires you and you want to strengthen your technical abilities even further, you may enjoy one of my LinkedIn Learning course dedicated to combining the power of Python with Excel. It is designed to help professionals automate tasks, analyze data more effectively and bring their spreadsheet skills to the next level.
