The next phase of artificial intelligence in commerce is being created right now — and not necessarily by the companies that speak the loudest about AI or invest the most in new tools. It is being shaped by a small group of practitioners who work not with abstract forecasts, but with real production environments: integrating new systems, testing their effectiveness, rebuilding workflows, and being the first to see where artificial intelligence truly creates business value — and where it remains merely a technological promise.
Oleh Halat, a recognized expert in AI-integration strategy and E-Commerce Business Analyst at ALEKO Products in Kent, Washington, belongs precisely to this group. His work spans the areas that are already defining what AI-powered commerce is becoming: agentic commerce protocols, conversational discovery as a replacement for keyword search, a methodology for evaluating where AI creates measurable business value, and the workflow infrastructure required for AI-native operations.
And while the industry focuses on which AI tools are worth buying, a completely different question is coming to the forefront: what will the next generation of AI-powered commerce look like? And, along with it, the question of who is actually shaping that generation.
The answer is becoming increasingly clear. The next generation of AI-powered commerce will not be shaped by the loudest AI vendors, nor solely by the largest retail players. It will be shaped by a small group of operational practitioners who are already building this future in real production environments: implementing agentic systems, defining evaluation methodologies, identifying workflow realities, and learning from what actually works in real commercial settings. Halat is one of these operators. His research portfolio, early implementation of new AI systems, and the foundational artifacts he has created — ACIAF and Cosmoflows — describe both where commerce is headed and how it gets there.
For most observers, “AI in commerce” still means tools — chatbots, recommendation engines, generative pricing, automated product descriptions, or personalized email campaigns. In this view, artificial intelligence remains an add-on to an existing commercial system: another layer on top of existing operations, sales channels, catalogs, search, and analytics.
Halat sees the fundamental shift differently. The next phase is not about adding more AI tools to existing operations, but about rebuilding operations so that they become AI-native. This means that artificial intelligence stops being a separate tool and gradually becomes part of the operational architecture of commerce itself.
This shift is defined by three specific trends, and all three are happening at the same time. The first is agentic commerce — commerce in which AI agents can conduct transactions on behalf of people. The second is conversational discovery, in which natural-language queries replace keyword search as the primary mechanism for product discovery. The third is AI-as-infrastructure, where AI becomes as invisible and essential as a database, CDN, or payment gateway.
None of these trends is purely speculative. Each of them is already being implemented by retailers that understand where commerce is heading. That is why Halat views the next five years not as a period of AI experimentation, but as a period of structural transformation in e-commerce operations.
“Agentic commerce is not a future scenario, but a shift in transactional protocols that is already taking place. Retailers that are building today’s operations as if AI agents will be conducting transactions tomorrow will be ready when it happens. Most are not building their operations that way,” the analyst says.
This statement by Halat explains why the future of AI-powered commerce cannot be reduced to the automation of individual tasks. If AI agents become new participants in the transaction process, then not only does the interface change, but the very structure of commercial interaction as well.
So what shifts should we expect in the next phase? First and foremost, the customer’s path to a product will change. For many years, e-commerce architecture was built around keyword search: categories, tags, product names, SEO structure, filters, attributes, and search algorithms. But generative AI is gradually changing the very nature of product discovery.
The customer of the future will no longer formulate a query through keywords, as today’s customer does. Instead, they will describe intent, provide context, outline constraints, lifestyle, budget, use case, or the problem they intend to solve with the product. The system will respond in kind: it will not simply offer products to choose from, but will engage in a dialogue with the customer, clarify the need, and generate a commercial offer in the moment — right here and now.
“Search engines are not being replaced by better search engines. They are being replaced by something fundamentally different — conversation. This distinction matters because every layer of today’s e-commerce architecture is built around keywords,” Oleh Halat explains.
This idea already has practical implications for the entire industry. The transition of discovery from keywords to conversation presents retailers with a new challenge: they must rethink not only search systems, but also catalog structure, product data, SEO, merchandising logic, recommendation systems, intent analytics, and even how the effectiveness of the commercial experience is measured.
What gives Halat grounds to speak about this shift is his practical experience implementing such solutions in a production environment. At ALEKO Products, which operates through multiple major U.S. retail channels and processes over one million website visits through its own direct-to-consumer storefront alone, Halat was among the first practitioners to implement agentic coding environments and parallel large language model systems into production e-commerce operations — as early as 2024, before most of the industry had begun to seriously consider agentic systems.
This work was not abstract. It included evaluating vendors based on measurable outcomes, integrating multiple AI systems into existing production architectures, and understanding where AI delivers measurable commercial value — and where it does not. These practical conclusions are what shape his vision of the next five years of commerce.
In this sense, Oleh Halat does not merely analyze the future of AI-powered commerce. He works in an environment where that future must first be tested, integrated, constrained, and measured — and only then turned into operational reality.
That perspective — grounded in production deployments rather than industry commentary — becomes increasingly important in the next stage of commerce
The expert is convinced: “In five years, no one will be talking about AI integration anymore. AI will become infrastructure — as invisible as a database or CDN. Companies that are investing in this invisibility now are building the next generation of commerce.”
Put simply, the successful companies will not be those that are first to add an AI feature to their website, but those that can rebuild their operations so that AI becomes an organic part of the commercial system.
Halat’s published and ongoing research forms a comprehensive roadmap for where AI-powered commerce is heading. His work explores generative AI as a replacement for traditional product search, the development of agentic commerce and the new transactional protocols it requires, the comparative economics of e-commerce platforms in an AI-native world, and the next evolution of machine-learning recommender systems beyond traditional collaborative filtering.
Each of these areas is part of a broader shift: generative AI changes the way products are searched for and discovered; agentic commerce changes the nature of transactions; AI-native platforms change the economics of technology choices for businesses.
The new generation of recommender systems transforms personalization from a reactive mechanism into a co-creative interaction between the user and the commercial system.
Together, these areas describe the operational architecture that retailers will need to build over the next five years. This research is not purely academic — it is grounded in what Oleh Halat sees every day in production deployments.
“Recommender systems have evolved from collaborative filtering to deep learning and LLM-based recommendation in less than a decade. The next iteration will not recommend products to customers — it will co-create the catalog with them in real time,” Halat says.
In other words, if today a recommendation system is still merely a mechanism for matching products, then in the near future it will become an interface between customer intent, catalog data, the retailer’s commercial logic, and AI agents that actively participate in decision-making. For e-commerce, this means a transition from a static architecture of product pages to a dynamic architecture of commercial scenarios.
Two innovative artifacts created by Halat lie at the foundation of what AI-native commerce will require. The first is the AI-Commerce Integration Assessment Framework, or ACIAF — a methodology designed for the systematic evaluation of where exactly AI integration belongs within a commercial operation. The second is Cosmoflows.app, an independent platform for workflow tracking and process automation that he designed and built.
Both tools address foundational layers that most retailers have overlooked. ACIAF gives operators a framework for disciplined decision-making around AI investments. Cosmoflows gives operators workflow visibility on which intelligent automation can later be built. Together, they describe how an AI-ready operation is actually created — not as a single AI project, but as a multi-layered architecture that the next generation of commerce will rely on.
The most popular approach among businesses today is to view AI as a set of tools: one for search, another for recommendations, a third for content, a fourth for support automation, and so on. But the real complexity of commercial operations often does not align with this logic. A business’s lack of understanding of the processes it automates and the data it uses automatically blocks AI effectiveness.
“AI-native operations require workflow-native foundations. You cannot layer intelligent automation onto opaque processes — that only automates opacity. Visibility precedes automation. Automation precedes AI,” Halat emphasizes.
That is why ACIAF and Cosmoflows should be viewed not as auxiliary tools, but as foundational infrastructure for next-generation commerce. One artifact answers the question of which scenarios make sense for AI and allows its impact to be assessed. The other makes visible the processes without which mature automation cannot be created. Together, they prove that the AI-powered commerce of the future begins not with a model, but with operational discipline.
The conclusion for the rest of the industry is direct, but uncomfortable: companies that wait to react to AI-powered commerce will not have time to react. The transition from keyword search to conversational discovery, from human customers to AI agents acting on their behalf, and from manual workflows to AI-native operations — each of these processes is unfolding within timeframes shorter than most strategic planning cycles.
The companies that will be best positioned over the next five years are those already building the foundation in production environments, using methodologies such as ACIAF for disciplined investment decision-making and tools such as Cosmoflows to reveal the operational reality on top of which AI integration will later function.
Another conclusion worth stating here and now concerns the future competitive advantage in e-commerce: its dependence on access to individual AI vendors will be minimal; instead, it will be determined by a company’s ability to rebuild its own operations.
AI will become the standard. The difference will be created not by the mere fact of using artificial intelligence, but by the quality of that use. This means integrations, process maturity, data architecture, the speed of team learning, and an understanding of where automation can truly improve business outcomes.
The next generation of AI-powered commerce is being shaped not by the loudest AI vendors and not by the largest retail incumbents. It is being shaped by a small group of operators who are building this future today — implementing agentic systems in production, defining evaluation methodologies, uncovering workflow realities, and learning from what actually works in real commerce environments.
Oleh Halat’s career is a notable illustration of how this work happens. He operates precisely in the segment where large-scale technology trends intersect with operational reality — where AI must be validated not by a presentation, but by results, and where a new system must be embedded not into theory, but into a living e-commerce stack.
The companies that will be best prepared for future changes are those that recognize the value of this operator-led foundation work and either find such specialists or develop them within their own teams. Because the next phase of AI-powered commerce will not appear suddenly. It is already being created — in production environments, in workflows, in new methodologies, in invisible infrastructure, and in the decisions of practitioners who understand that the AI of the future begins with operational reality today.
