AI and Commerce Disconnect - Intelligent Agents Meet Stone Age Shopping Carts
How AI Agents Will Force Commerce to Evolve or Die
Continuing exploration of the future of software in the age of AI. First essay of the series can be found here Software evolution from digital filing cabinets to cognitive partners with AI.
You can ask ChatGPT to write a sonnet in the style of Shakespeare with a Freudian twist about Schrödinger's cat existential crisis but try buying a sweater that actually fits your Seattle lifestyle, and you'll discover the internet is still fundamentally designed for humans who enjoy being their own research assistants. The same digital infrastructure that promises AI-powered everything forces you to click through seventeen pages of flight options while mentally calculating baggage fees—and still watch prices change right before you hit submit.
This disconnect exemplifies what I explored in my analysis of software evolution Software evolution from digital filing cabinets to cognitive partners with AI: we've built sophisticated interfaces while ignoring cognitive bottlenecks. Current e-commerce platforms are “digital filing cabinets” with prettier checkout flows, not cognitive partners that understand purchase intent. Unlike my theoretical framework of true cognitive automation, today's commerce systems remain stubbornly stuck in the HTTP-and-HTML era of human-dependent processes.
I recently conducted experiments testing AI for personalized shopping and travel booking. The results revealed something fascinating: while AI can masterfully synthesize information and provide sophisticated analysis, the commercial internet remains a patchwork of programming language specifics and platform quirks rather than accounting for human psychology.
This isn't about creating more polite chatbots. We're witnessing the early stages of commerce's own EPIC transformation—currently stuck in the Panic phase as companies realize chatbot theater won't deliver the promised AI revolution. The companies that understand this transition first will define the next decade of digital commerce. For details on EPIC framework - The EPIC Cycle: When Winnie-the-Pooh Meets Silicon Valley Economics
Mainframe with color terminal, Commerce Edition
The scale of this infrastructure mismatch is staggering. The AI in e-commerce market is valued at $9.01 billion in 2025 and is projected to reach $64.03 billion by 2034 [1], yet these same systems force users back into manual browsing when they actually want to purchase something.
This represents the same metaphoric pattern I identified for cloud transformation of business apps: expensive digital secretaries organizing filing cabinets. Commerce platforms promise intelligent shopping but deliver configurable recommendation websites that can't process complex queries like "minimalist aesthetic + sustainable brands + Seattle climate + business travel requirements + $80-150 budget" the way AI agents naturally handle such requests.
When AI tried to help fictitious "Sarah Kim" buy transitional fall workwear for her Seattle-based product manager lifestyle, everything fell apart. Despite perfect user profiling and clear brand preferences, the AI could only provide generic price ranges like "$80-120 at COS or Uniqlo" instead of "here's the exact blazer in your size for $89.99, available for delivery Thursday."
Travel booking proved even more revealing. What should have been straightforward—Seattle to Maui, family of four, arriving around 4PM—devolved into marketing-speak about "starting from $179" prices that no actual human could book. The AI found historical averages and expired promotional codes but couldn't access real-time inventory and pricing.
Translation: While AI has achieved remarkable sophistication understanding human intent, commercial infrastructure remains optimized for humans who enjoy manually converting vague suggestions into actual purchases. This is cognitive automation's opposite, interface sophistication masking system ignorance.
The Three-Tier Commerce Transformation
This transformation follows the same stratification pattern I've identified across technology roles. Commerce platforms are evolving into three distinct categories, as I noted in my previous essay Software evolution from digital filing cabinets to cognitive partners with AI:
AI-Native Commerce Platforms —business systems built with intent-based fulfillment from the ground up. These platforms learn appropriate purchase behaviors from data and context rather than requiring human navigation through product catalogs. Companies building these systems today will capture first-mover advantages similar to early SaaS pioneers.
Hybrid Intelligence Commerce adds AI capabilities to traditional architectures while preserving configuration complexity and manual browsing requirements. These systems provide immediate improvements but retain the technical debt of rule-based shopping experiences.
Legacy Augmentation includes traditional e-commerce with chatbot overlays—the endangered middle offering minimal transformation because they preserve manual configuration limitations while adding conversational complexity.
The infrastructure problem spans three distinct layers never designed for AI agent interaction. The data access layer carries legacy "luggage" with proprietary schemas across different companies. The transaction layer remains optimized for human form-filling rather than AI decision-making; there's no API for "just buy Sarah the best blazer and deliver it Thursday." The user experience layer provides primitive personalization that can't process the multi-dimensional preferences AI agents handle naturally.
The direction of change
I think that unlike previous automating and improving user interaction with more appealing experiences, commerce will move into fully enabling AI agents.
Visa's Intelligent Commerce initiative exemplifies this shift, providing "integrated APIs and a commercial partner program to AI platforms, enabling developers to deploy AI commerce capabilities securely and at scale" [2] specifically designed for AI agents rather than human browsers.
This represents more than just improved user experience. This signals the emergence of "intent-based commerce." Rather than organizing information around product categories designed for human navigation, commerce will reorganize around AI's ability to understand and fulfill complex, contextual human intentions. Sarah Kim doesn't want to browse 247 blazers; she wants one specific blazer matching her professional requirements and personal aesthetic.
Gartner predicts that 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024 [3], representing one of the fastest enterprise technology adoption curves in recent history. But as I've explored in my analysis of how product managers evolve into AI-human mediators, The Evolution: From Strategic Leaders to AI-Human Mediators, this transformation requires professionals who can bridge algorithmic objectivity with human psychology—exactly what commerce needs.
The fundamental question shifts from "how do I research this purchase decision?" to "who should I delegate this purchase decision to?" Companies that figure out optimal delegation will win the next round of digital commerce. This isn't about replacing human judgment—it's about eliminating tedious cognitive research tasks AI handles better while preserving human control over final decisions.
The first mover advantage - massive
The company that cracks AI-agent-optimized commerce first won't just gain market share—they'll redefine entire categories. Imagine Amazon designed for AI agents rather than human browsers. Instead of searching through thousands of products, you'd specify intent and receive curated options with real-time availability and pricing.
Major companies recognize this need. Visa's collaboration mentioned earlier with "Anthropic, IBM, Microsoft, Mistral AI, OpenAI, Perplexity, Samsung, Stripe and more" [2] demonstrates industry-wide acknowledgment that current systems weren't designed for AI agent transactions. Payment infrastructure providers understand that this transformation requires rebuilding commerce from the ground up. It is interesting how casinos and payment processors jump first onto software and math scientific and practical software accomplishments. Apparently, physical presence of money creates visceral drive to progress!
Early adopters will build genuinely superior customer experiences while traditional retailers respond by adding conversational interfaces to existing infrastructure, creating AI integration theater. Customers will quickly distinguish between actual intent-based commerce and chatbot marketing, just as they learned to differentiate between genuine cloud transformation and expensive database hosting.
The convenience gap will prove impossible to bridge through incremental improvements. All things being equal, humans prefer experiences requiring minimal cognitive effort—we're evolutionarily predisposed to conserve mental energy. Research confirms that "exerting cognitive effort is aversive, and people avoid it whenever possible" [5]. The brain is expensive for the body to operate, accounting for 20% of the body's total energy consumption despite representing only 2% of body weight [6], so anything that reduces thinking load gets embraced rapidly.
Back to the Future
The internet is approaching its most significant architectural shift since the transition from text-based to graphical browsing. So far, software has been automating mechanical tasks from cognitive perspective like browsing, searching and payment.
The shift to AI agentic commerce will require big changes in how companies think and what software they use. It is a great opportunity for software engineers and product leaders. Unlike Y2K or cloud migration that required armies of engineers, program managers and consultants - with CompTIA reporting over 102,000 new IT jobs added in 2021 alone for "cloud computing migration, application integration, process automation" [4] - AI commerce transformation will be more selective, creating opportunities for those who understand both technology and human psychology.
Companies in online commerce function will be frantically developing "AI stuff", so learn it, lead it. We are moving back from the end to end purchasing experience perspective. Those of us with more mileage remember how shopping, especially in smaller stores, was. Owners knew you, they knew what you like, what you may want today because it is Friday and your child's birthday is coming up or you had the flu last week, or you that your daughter loves stuffies and that cute bear is just what she needs. This is just not possible with current commerce! With AI, especially with emotional and neuro capable AI, it can be!
Human nature remains constant for all times. We love attention and care. We will pay to have it. AI is here to stay. My word to my fellow engineers and product managers: you can't beat it so you may as well lead it!
References
[1] Precedence Research. (2024). Artificial Intelligence in E-commerce Market Size and Growth 2024 to 2034. Available at: https://www.precedenceresearch.com/artificial-intelligence-in-e-commerce-market
[2] Visa Corporate. (2025). Enabling AI agents to buy securely and seamlessly. Available at: https://corporate.visa.com/en/products/intelligent-commerce.html
[3] Gartner. (2025). How Intelligent Agents in AI Can Work Alone. Available at: https://www.gartner.com/en/articles/intelligent-agent-in-ai
[4] CompTIA. (2021). Cyberstates 2021 - IT Workforce Report. Available at: https://connect.comptia.org/content/research/cyberstates-2021
[5] Kurzban, R., et al. (2022). Rewarding cognitive effort increases the intrinsic value of mental labor. Proceedings of the National Academy of Sciences. Available at: https://www.pnas.org/doi/10.1073/pnas.2111785119
[6] Kuzawa, C.W., et al. (2014). Metabolic costs and evolutionary implications of human brain development. Proceedings of the National Academy of Sciences. Available at: https://www.pnas.org/doi/10.1073/pnas.1323099111
This analysis represents observations from practical AI agent experiments combined with pattern recognition from previous technological transitions and established frameworks for industry maturation. While the timeline for full adoption may vary, the direction of change appears inevitable and accelerating.