Subscribe|Submit a tip|Advertise with Us

🧩 Welcome to DataPro 161. Our Expert Insight this week features Anas Riad, Data Scientist, BI Consultant, and ML Engineer at AdWay, as he maps what the next decade of AI could actually look like. From small language models (SLMs) and model routing to the rise of AI agents and the long-range pull of AGI, he breaks down what is moving from hype to real deployment, and what still needs serious guardrails around security, privacy, and energy costs. If you are building in AI, leading data teams, or trying to future-proof your career, this edition focuses on the decisions that will matter most between 2025 and 2035.

We are also bringing something new to our readers, straight from experts with first-hand industry experience. Packt has launched live AMA sessions with industry leaders on the Packt DataML YouTube channel, hosted by our Growth Lead Abhishek Kaushik. These sessions give you direct access to practical guidance, career clarity, and real world insights from voices shaping the future of Python, AI, BI, and data science.

In this edition, we have included Anas’s take on why smaller models will win more real-world workflows, how AI agents are evolving beyond demos, and what individuals and organizations should do now to stay productive without losing the human and ethical core of the work. You will find the extended highlights below.

Knowledge Partner Spotlight: Outskill
At Packt, we’ve partnered withOutskillto help readers gain practical exposure to AI tools through free workshops, complementing the deeper, hands-on, expert-led experiences offered throughPackt Virtual Conferences.

Learn AI tools, agents & automations in just 16 hours (End of Year offer)
Outskill offers a focused 16-hour, two-day live AI program designed to help professionals enter 2026 more confident, faster, and AI-ready. As part of their Holiday Season Giveaway, the first 100 participants can join free and unlock bonus AI resources including a prompt library, monetization roadmap, and a personalized AI toolkit, making this a practical on-ramp for anyone serious about staying relevant in an AI-driven workplace.

Join Tivadar Danka for an intensive, hands-on workshop on January 24, where you will build a rock-solid mathematical foundation for machine learning from the ground up. In this live session, you will go beyond model.fit() and implement core ML concepts from scratch in Python and NumPy, covering linear algebra, calculus, probability, and optimization through a full end-to-end linear regression workflow. Whether you are an aspiring data scientist, ML engineer, or Python developer transitioning into AI, this workshop connects theory to real implementation and helps you truly understand how modern ML systems work. Use EARLY50 for early access savings and reserve your seat.

Use EARLY50 for early access savings and reserve your seat.

Cheers,

Merlyn Shelley

Growth Lead, Packt

Artificial intelligence is no longer a future concept. It is already embedded in how we write, code, analyze data, design products, and make decisions. What has changed over the last few years is not the existence of AI, but its accessibility. Tools that were once reserved for large research labs and tech giants are now available on personal laptops, phones, and everyday workflows.

In a recent episode ofPackt Talks, we sat down withAnas Riad, a data scientist, BI consultant, and ML engineer at AdWay, to unpack where AI and machine learning are heading next. The discussion spanned small language models, AI agents, industry disruption, ethics, energy costs, and the long-term question everyone is asking. Are we moving toward AGI, and what does that actually mean for society?

This blog distills that conversation into a structured outlook on what the next decade of AI could look like, and how individuals and organizations should prepare.

From Hype to Everyday Utility

Machine learning has existed for years, but only recently has it become part of everyday life. The real shift happened when AI became usable by non-experts. Today, anyone can interact with powerful models on their phone or laptop without understanding the mathematics behind them.

This democratization explains the current hype cycle. AI feels revolutionary not because it is new, but because it is finally accessible. At the same time, the pace of progress is so fast that relevance feels temporary. What feels cutting-edge today can feel outdated in months.

That speed is both exciting and destabilizing. It forces individuals and businesses to think less about mastering a single tool and more about staying adaptable.

The Near-Term Shift: Smaller Models, Smarter Deployment

Despite the constant release of larger and more capable models, the next year is unlikely to be defined by dramatic breakthroughs. Instead, the focus is shifting towardefficiency.

One of the most important trends is the rise ofsmall language models (SLMs). Not every task requires a massive, multi-billion parameter model. In fact, using large models for simple tasks is often slower, more expensive, and unnecessary.

Small models excel when the task is narrow. Summarization, classification, lightweight reasoning, or structured extraction can often be done faster and cheaper with an SLM. Large models still matter for complex reasoning, multi-modal understanding, or long-context tasks, but the future is not one model doing everything.

The real change is architectural. Systems will increasingly route tasks to the right model rather than defaulting to the largest one available. This improves speed, cost, and deployability, especially for local and edge use cases.

What Changes for Users?

From a user perspective, the difference between large and small models will mostly be invisible. What users will notice is faster responses, lower costs, and AI that feels more embedded into everyday tools rather than accessed through a single chat interface.

The key shift is optimization. Instead of asking, “What is the best model?” teams will ask, “What is the right model for this task?” This mindset is essential for building scalable AI systems.

Industry Impact: No Sector Is Immune

AI is already reshaping software engineering, data science, and analytics. Code is written faster, debugging is assisted in real time, and deployment pipelines are increasingly automated. Tasks that once took days now take hours.

Beyond tech, the impact is spreading everywhere:

Healthcareis seeing early gains in diagnostics, scheduling optimization, and treatment modeling.

Financeis using AI for credit risk, fraud detection, and portfolio optimization.

Operations and logisticsare being optimized through predictive modeling and automation.

Creative industriesare seeing massive productivity gains in writing, design, video, and music.

The long-term implication is clear. AI adoption is no longer optional. Organizations that resist it will fall behind competitors who use it to move faster and operate more efficiently.

Training, Architecture, and the Rise of AI Agents

One of the most misunderstood aspects of modern AI is what it means to “use AI well.” It is not about chasing every new framework or model release. Success is measured by productivity gains, not by tool count.

👉 Continue reading the full article on the Packt Medium handle.

See you next time!