An investigation into what happens to brands, businesses, and you when the thing doing the shopping isn't human
Prologue: A Tuesday Morning in 2026
Picture it. Somewhere in London, a nineteen-year-old named Priya wakes up at 7:43 a.m., checks her phone before her feet touch the floor, and says — not types, says — to her AI assistant: "Find me a decent accounting app for my freelance work. Under twenty-five pounds a month. Book me a demo if they offer one."
She puts the phone down. Makes coffee. Checks Instagram.
By 8:01 a.m., the agent has scoured dozens of platforms, filtered by price, parsed terms-of-service documents that no human has read in full since 2019, cross-referenced user reviews from three separate databases, and scheduled a demo with one specific software company for next Thursday at 11 a.m.
Priya had nothing to do with the decision. She gave a goal. The agent gave a result.
Somewhere across town, in a glass-fronted office on the Embankment, a marketing director at that same software company is preparing a presentation about their new brand refresh. New colours. New logo. Emotional storytelling. A hero section on the website with a beautiful slow-motion video of people laughing at laptops.
The agent did not watch the video. The agent never visited the website. The agent read a structured data file, compared a pricing API, checked a trust score, and made a choice.
The marketing director's work is not entirely wasted. But it is, quietly and without fanfare, becoming less relevant than anyone in that glass office would like to admit.
This is the story of what happens to commerce when the buyer is no longer a person staring at a screen — but a machine executing a task.
Part One: The Funnel That Built the Modern World
To understand what is breaking, you need to understand what was built.
For most of the twentieth century, the theory of how to sell something to someone followed a simple logic. Psychologists, copywriters, and eventually marketers formalised it into a model they called AIDA: Awareness, Interest, Desire, Action. Make someone notice you. Make them curious. Make them want. Make them buy.
The model was elegant. It was also built on one enormous, unexamined assumption: that a human being was doing the perceiving, the feeling, and the deciding.
Everything followed from that assumption. Advertisements were designed to catch the eye. Window displays were arranged to trigger want. Salespeople were trained to read faces, to time their silences, to feel the temperature of a room. Even when commerce moved online in the 1990s and 2000s, the assumption held. The pop-up ads, the landing pages with countdown timers, the testimonials with smiling faces — all of it aimed at a consciousness behind a screen. All of it engineered for the emotional architecture of a human mind.
Then something started to shift. Slowly at first, then faster than anyone had predicted.
Search engines automated the awareness phase. Comparison sites automated the interest phase. Review aggregators sped up the desire phase. Each innovation compressed the funnel, took a little more agency away from the seller and handed it to the buyer. Each step was an improvement, by most accounts. But each step also quietly eroded the leverage that brands had over attention.
By 2022, marketers were already alarmed about what they called the "dark funnel" — the portion of a buyer's journey that happened invisibly, in private chats and social media feeds and YouTube rabbit holes, entirely beyond the reach of traditional analytics. You could no longer be certain where your customers had come from. The funnel, never as tidy as the textbooks suggested, was becoming harder to see.
What happened next made the dark funnel look like a minor inconvenience.
Part Two: Enter the Agent
The word "agent" has a precise meaning in philosophy and computer science. An agent is an entity that perceives its environment and takes actions to achieve a goal. A thermostat is a rudimentary agent. A chess engine is a more sophisticated one. What emerged between 2023 and 2026 was something qualitatively different: a class of AI systems capable of using language, browsing the web, comparing options, scheduling appointments, filling forms, and completing purchases — all on behalf of a user, based on nothing more than a natural-language instruction.
The technical term is "agentic AI." The practical meaning is simpler and considerably more disorienting: for the first time in the history of commerce, the entity doing the shopping does not have wants, feelings, nostalgia, or moods.
By the time McKinsey published its landmark analysis in early 2026, the numbers were already staggering. During the Cyber Week shopping period of November 2025, AI-influenced purchases accounted for approximately $67 billion in global online spending — roughly twenty per cent of all digital orders placed during that window. ChatGPT's shopping function, Perplexity's product search, Google's Gemini, Amazon's Rufus — each of these systems was, in various degrees of autonomy, making decisions that had previously required a human brain.
And the trajectory was unmistakable. Estimates suggested that by the end of 2026, between twenty-five and thirty per cent of all online purchases in the United States would involve an AI agent at some point in the decision process. By 2028, Gartner projected that figure could reach fifty per cent. In the B2B world — the world of companies buying from other companies — the figures were even more dramatic. Gartner, whose research carries the weight of institutional authority in the corporate world, predicted that AI agents would command fifteen trillion dollars' worth of B2B purchases by 2028, with ninety per cent of all business buying intermediated by AI.
These are not science fiction projections. They are the considered estimates of the world's most cautious research organisations, and they arrive with a peculiar kind of violence: quiet, invisible, and already underway.
Part Three: What the Agent Sees, and What It Ignores
Here is the detail that no one at the glass-office brand refresh meeting wants to hear.
When an AI agent is given the instruction "find me the best accounting software under £25 a month," it does not open a browser in the way you open a browser. It does not admire a hero image. It does not feel the reassurance of a professional-looking colour palette. It parses structured data — the machine-readable information that sits beneath or alongside the visible surface of a website. Prices. Availability. Ratings expressed in schema markup. Policy terms that can be extracted and compared programmatically. API response times that indicate how reliably a service delivers information.
What it ignores is, to put it mildly, significant. It ignores the storytelling. The emotional copywriting. The founder's personal journey that the brand blog has been labouring over for three years. The award-winning video. The carefully chosen sans-serif font that the design team debated for six weeks. It ignores all of it — not out of malice but out of architecture. It was not built to feel. It was built to decide.
Researchers at Merkle, one of the world's largest marketing consulting agencies, tested this in 2025 with a controlled experiment. They sent AI agents looking for specific product categories. The agents completed their research, formed their recommendations, and returned their answers. The product websites, in almost every case, had never been visited. The AI had sourced what it needed from structured databases, product catalogues, and information that existed in machine-readable formats. Marketing copy — the most expensive, most obsessed-over element of most brand strategies — had not featured in the decision at all.
"Brand loyalty," as one analysis put it with characteristic bluntness, "becomes algorithmic. Agents do not have brand nostalgia. They evaluate structured signals — price, availability, shipping speed, trust scores, and return flexibility."
That sentence deserves a moment's pause. Think about what it means. Think about Nike, whose entire commercial existence rests on the emotional resonance of "Just Do It." Think about Apple, whose products are priced at a premium that only makes sense in the context of the feelings they generate. Think about every small brand that has ever competed against a larger rival by being more authentic, more human, more itself.
Now think about what happens when the buyer has no emotional architecture to appeal to.
Part Four: The B2B World — Where Robots Meet Bureaucracy
If the consumer world is being reshaped, the business-to-business world is experiencing something closer to a reconstruction.
Business purchasing has always been a peculiarly human affair. The lunch with the account manager. The conference where you meet the vendor and shake their hand and decide, based on something ineffable, whether you trust them. The email chains that go on for weeks. The procurement committee meeting where someone raises a concern about integration and the whole deal stalls for a month. Commerce between companies has been, at its heart, a social process — governed by relationships, contracts, approved supplier lists, and the particular politics of every individual organisation.
Agentic AI does not eliminate these dynamics. But it compresses them, automates the first eighty per cent of them, and hands the remaining twenty per cent — the genuinely complex, genuinely human parts — back to people who now have much less time to do anything other than validate.
Gartner's research, published in March 2026, found that sixty-seven per cent of B2B buyers now prefer a purchasing experience that involves no sales representative at all. Forty-five per cent said they used AI during their most recent significant purchase. Eighty-nine per cent, according to Forrester, were using generative AI for independent research before they made contact with any vendor.
The sales representative — that figure of twentieth-century commerce, trained to schmooze and pitch and close — is not disappearing. But they are migrating. They are retreating to the final stage, the validation stage, the stage where a buyer who has already had their decision shaped by an AI agent wants to speak to a human being before signing something. A separate Gartner survey, published in May 2026, found that sixty-nine per cent of B2B buyers still turn to human sales representatives to validate AI-generated recommendations. The human remains in the loop. But the loop has moved.
For young people entering the workforce — for the nineteen-year-old who is about to become a procurement analyst, or the twenty-one-year-old who just got their first job in marketing — this is the world they are stepping into. Not a world where AI has replaced human judgement entirely, but a world where AI handles the research, the filtering, the comparison, and the scheduling, while humans are reserved for the moments that require trust, nuance, and accountability.
Understanding this is not optional. It is the literacy of the age they are inheriting.
Part Five: The Consumer World — Trust as Currency
Back to Priya.
Or, more precisely, to the question her behaviour raises. She delegated a financial decision — one involving a monthly payment, a software subscription, a scheduled meeting — to a machine she has been using for eight months. She did it without hesitation. She did it the way her parents would have Googled something without stopping to think about whether Google could be trusted.
This is the generational story that underlies the technology story.
Research published in 2025 by PayPal found that sixty-one per cent of Gen Z shoppers had actively used AI tools to help with a purchase in the previous year — not as an experiment, but as a normal part of how they shop. A separate study found that thirty-three per cent of people aged seventeen to twenty-five now prefer AI platforms for product research, a figure that nearly matches the thirty-seven per cent who still use traditional search engines. The crossover point — where more young people turn first to an AI than to a search engine — is not far away.
Capgemini's 2025 consumer trends report found that seventy-one per cent of consumers across all ages want AI integrated into their purchasing experiences. The pull is not just from the young. But the young are furthest ahead. They are the generation that grew up with recommendation algorithms — with Spotify knowing what they wanted to hear before they knew themselves, with TikTok serving content that anticipated moods they had not yet articulated. Delegating to an AI agent does not feel alien to them. It feels like the next logical step in a relationship with technology that has been building since childhood.
And yet. And yet.
Even among the most enthusiastic users, trust in agentic commerce is, as a June 2026 report from Checkout.com put it, "still catching up." The same research found that twenty-seven per cent of consumers trust no organisation to operate an AI shopping agent on their behalf. Twenty-four per cent said they would never delegate purchases to AI. On average, consumers in the six markets surveyed said they would allow an AI agent to spend only £177 per purchase without requiring additional approval — a figure that tells you something important about where the boundaries of delegation currently sit.
The hesitation is rational. An AI agent, however sophisticated, can hallucinate. It can misread a policy. It can confuse two similarly named products. It can be manipulated — a risk that researchers have documented with some alarm — by sellers who embed instructions in their product data, exploiting the agent's processing logic to make themselves appear as the best option even when they are not. This is not a theoretical threat. It is an active area of concern in the field of AI security, and it has a name: prompt injection, where malicious instructions hidden in product descriptions or metadata attempt to override an agent's genuine assessment.
McKinsey, in its definitive analysis of agentic commerce published in early 2026, put the trust question precisely: "Trust is built not through perfect execution but through explainability and reversibility. As autonomy increases, consumers want to understand not just what the agent did but why it behaved in that manner." Loyalty, McKinsey added, is becoming "less about sentiment and more about policy."
That phrase — loyalty as policy — might be the single most important insight about what is changing. The relationship between a person and a brand has historically been emotional. It has been built through years of consistent experience, through advertising that lodged in memory, through the texture of a product in a hand. The new relationship, mediated by an agent, is contractual. The question is not "do I love this brand?" but "does this brand's data conform to the parameters I have set?"
Part Six: AAIO — The Art of Being Chosen by a Machine
Here is where the story turns practical, and where a new vocabulary becomes necessary.
For twenty-five years, the defining discipline of digital visibility was Search Engine Optimisation — SEO. You wrote content, built links, structured your pages in ways that made a search engine's crawlers recognise you as authoritative and relevant. The goal was a high rank on a results page. A human being then decided whether to click.
Between 2022 and 2024, a new layer emerged: Answer Engine Optimisation (AEO) and Generative Engine Optimisation (GEO). As ChatGPT, Perplexity, and Google's AI Overviews began synthesising answers rather than listing links, brands needed to be cited rather than merely ranked. The goal shifted from appearing in a list of ten results to being the source an AI drew from when it composed its answer.
Now, in 2026, a third layer is forming. It has been called Agentic AI Optimisation — AAIO. And it represents, as one analysis put it, "the most significant shift in digital discoverability since the birth of SEO itself."
The distinction matters. In SEO, you were trying to convince a crawler. In AEO/GEO, you were trying to be cited by a generative model. In AAIO, you are trying to be selected by an agent executing a task with real-world consequences — a purchase, a booking, a contract. The stakes are higher. The signals that matter are entirely different.
What does an agent use to evaluate a brand? The checklist is unglamorous but precise. Clean product data with consistent attributes, global identifiers like GTINs and SKUs, live pricing and real-time inventory, machine-readable return and warranty policies, API endpoints that respond quickly and reliably, review data expressed in structured schema markup, trust signals that can be verified programmatically.
The rule — blunt, revolutionary — is this: if your data is not structured and accessible, your brand does not exist to the systems doing the buying.
Platforms are already building infrastructure around this reality. Scrunch, one of the new category of "Agent Experience Platforms," delivers AI-optimised content directly to agents at the CDN layer — the technical infrastructure through which web content flows — without disrupting the experience for human visitors. The same website, in other words, serves two entirely different versions of itself: one for people, one for machines. A dual audience has become a dual architecture.
HubSpot's research, published as part of their Spring 2026 product announcements, found that AI-referred sessions — visits or interactions that came through an AI agent — were converting at 4.4 times the rate of organic search traffic. The volume is still small. The quality is extraordinary. And the gap between volume and quality is what tells you where energy should be directed.
Part Seven: What This Means for the Small Brand, the Start-Up, and the Student
There is a paradox buried in all of this, and it is worth surfacing because it is genuinely surprising.
The conventional assumption would be that the era of AI agents favours large brands. Bigger brands have more data, more resources, more visibility in the training sets of the models that agents use. Surely the behemoths win?
Not necessarily. In fact, almost the opposite may be true.
A small brand with meticulously structured product data — clean descriptions, accurate pricing, real-time availability, honest reviews in the right schema format — can out-compete a much larger brand whose data infrastructure is a legacy mess of inconsistent information, stale prices, and marketing copy that no agent can parse into a purchase signal. The playing field is not level. But the criterion on which it is uneven has changed. It is no longer about budget. It is about data discipline.
This has real implications for young entrepreneurs. The accounting software freelancer that Priya's agent chose was not necessarily the biggest name in the market. It was the one whose data answered the agent's questions most clearly. In a world where sixty-one per cent of Gen Z already uses AI for purchase research, being the brand that agents can read is becoming more valuable than being the brand that people remember.
There is also something worth noting about the B2B trajectory for young people starting careers. Gartner's prediction that ninety per cent of B2B buying will be intermediated by AI by 2028 means that the skills valued in sales, marketing, and procurement are shifting. Understanding how structured data works, how APIs function, how AI systems evaluate trust signals — these are no longer purely technical skills. They are commercial skills. They are what will separate the people who thrive in the new economy from the people who are surprised by it.
Part Eight: The Things That Remain Human
It would be dishonest — and, frankly, not very useful — to end this analysis without noting what the data consistently shows at the edge of the agentic revolution.
The human does not disappear. They migrate.
In B2B, sixty-nine per cent of buyers still want a human being to validate the recommendation their AI agent has produced before they sign anything of consequence. In B2C, twenty-four per cent say they will never delegate purchases to AI. In complex categories — travel, major electronics, home purchases — McKinsey's research shows that agents handle research and comparison but defer to the human when the decision involves genuine trade-offs between values. Time versus comfort. Cost versus quality. Efficiency versus the ineffable feeling that a particular thing is right.
The configuration of the relationship is shifting, not erasing one side of it. Priya's agent booked the demo. But Priya will attend the demo. She will notice whether the person on the other end of the video call listens to her, whether they treat her as a freelancer with modest needs or as a future customer who might grow into something bigger. She will feel the texture of the interaction and form a judgement that no structured data field can anticipate.
The brands that understand this dual reality — that they must be machine-readable to be discovered, and human-worthy to be chosen in the moments that matter most — are the ones that will navigate the next decade with any grace.
Epilogue: The Manager Who Stared at the Video
Back in the glass-fronted office on the Embankment, the marketing director finishes her presentation about the brand refresh. The colours are beautiful. The video is stunning. Her colleagues applaud.
None of them know, yet, that the accounting software demo Priya booked this morning was booked by a machine that never watched a single second of any of their work.
The question is not whether that fact should prompt despair. It should not. The video still matters. The emotional story still matters. It matters when Priya shows up to the demo. It matters when she tells her housemate about the service. It matters in all the human-to-human moments that still constitute the majority of how trust is built and relationships are formed.
The question is whether the marketing director knows that a second audience now exists — one that needs entirely different food. Not emotion. Data. Not story. Structure. Not beauty. Legibility.
The brands that succeed in the next decade will be the ones that learn to feed both audiences simultaneously.
The buyer has left the building. But they have sent someone in their place. And that someone is reading your product schema right now.
Author's Note: Limits, Tensions, Subjective Errors, Aesopic Language, and Rhetorical Devices
Limits of This Analysis
This article is written in June 2026, at what is almost certainly the beginning of the agentic commerce story, not the middle or end. The projections cited — Gartner's fifteen-trillion-dollar B2B figure, McKinsey's automation curve, the fifty-per-cent-by-2028 consumer estimate — are forecasts, not facts. Forecasts from even the most rigorous research organisations have a habit of being wrong on timing, if not on direction. OpenAI paused its "Instant Checkout" feature in late 2025, a reminder that the road to autonomous commerce has speed bumps. Walmart found that in-chat AI purchases converted at three times lower rates than redirecting users to its own website. The article presents the general direction of change with confidence. The specific timeline should be held more loosely.
The research base leans heavily on American and British data. The dynamics in markets with different regulatory environments — the EU's AI Act, different data protection regimes, different consumer trust baselines — may play out differently. What is true for Priya in London is not necessarily true for a freelancer in Warsaw or a small business owner in Lagos.
Tensions the Article Cannot Fully Resolve
There is a genuine tension between the democratising promise of agentic commerce (small brands with clean data can compete with large brands) and the centralising risk (the AI models that agents use are built by a small number of very large companies, who make choices about which data they trust, which signals they weight, and which brands appear in which positions). The "level playing field" argument assumes that the AI models are genuinely neutral. They may not be. This is worth watching.
There is also an unresolved tension between individual convenience and collective consequence. If everyone delegates their purchasing decisions to AI agents, and those agents share similar logic and similar training, the result might be a world where commerce becomes far less diverse — where unusual products, niche brands, and genuinely innovative small businesses are systematically de-selected because they lack the data infrastructure to be legible to machines. Algorithmic homogenisation is a real risk, and it receives less attention than it deserves.
Subjective Errors and Biases to Acknowledge
The article implicitly privileges the perspective of brands and sellers — how they should adapt, what they should do. It spends less time with the consumer's experience of delegating decisions, which is more ambivalent and more complex than the optimistic reading suggests. Many people find it unsettling to have an AI spend their money. That discomfort is not irrational. It deserves more space than this article gives it.
Aesopic Language — The Hidden Messages
Every analysis of technological change contains what might be called șopârle — the little lizards, the slippery creatures that say one thing while meaning another.
When a report says "brands must win the algorithm," what it means, but does not say explicitly, is: "the companies that build the algorithms are the real winners, and everyone else is paying to participate in their system."
When an analysis says "trust is becoming the ultimate competitive advantage," what it encodes is: "trust is now something you can optimise for, which means trust is now something that can be manufactured, which means trust is in the process of becoming something other than trust."
When this article says "the buyer has left the building," what it is really suggesting, quietly, is: "the buyer has been removed from the building — and it is worth asking by whom, and why."
These are not reasons to reject the analysis. They are reasons to keep one eye on the frame as well as the picture.
Rhetorical Devices
The article uses the Priya vignette as a narrative anchor — a specific fictional person whose experience makes abstract trends concrete and emotionally accessible. It uses dramatic irony in the Epilogue: the marketing director does not know what we, the readers, have just learned. It uses bathos intentionally in the Merkle research passage — the long build-up to the discovery that marketing copy played no role in the agent's decision. The phrase "the buyer has left the building" is a deliberate echo of "Elvis has left the building" — evoking irreversibility and the end of an era while remaining light enough not to tip into melodrama.
What Can You Do Tomorrow? Five Concrete Suggestions
1. Audit one brand you care about for agent-readiness. Pick a brand — one you buy from, one you're starting, one you work for. Search for it on Perplexity or ChatGPT with a specific query ("find me [product category] under [price]"). Does it appear? Does the AI get the details right? What is missing? This is your baseline.
2. Learn what "structured data" actually means — in an afternoon. Google's free documentation on Schema.org takes less than two hours to read at a basic level. Understanding the difference between a product description that humans read and one that machines can process is foundational knowledge for any career in marketing, sales, product, or entrepreneurship in the next decade.
3. Set up an alert for your own digital footprint in AI. Use a free tool — there are several listed in the bibliography — to track whether your name, your brand, or your employer appears in AI-generated answers. You may be surprised by what you find, or by what is absent.
4. Have the "parameters" conversation. Next time you use an AI assistant for anything — shopping, research, booking — consciously think about what instructions you are giving it. What are you actually delegating? What controls are you setting? This is not just a consumer behaviour exercise. It is a philosophical one. The clarity with which you define your own goals is the clarity with which an agent can serve them.
5. Read one primary source. Pick one report from the bibliography below — the McKinsey piece, the Deloitte guide, the Checkout.com consumer trust research. Read the whole thing. The ability to go to primary sources rather than accepting summaries is a skill that distinguishes the person who understands a trend from the person who has heard of it.
What to Avoid — Five Things Not to Do
1. Do not assume this is only a technology story. Every technology story is an economics story, a politics story, and a power story. The question of who controls the AI agents that make purchasing decisions is as important as the question of what those agents can do.
2. Do not build a brand strategy that ignores the machine audience. Every piece of content you create, every product description you write, every price you set — ask yourself not only "will a person understand this?" but "will a machine be able to act on this?"
3. Do not confuse accessibility with trust. An AI agent can find your brand easily. That does not mean it trusts your brand. Trust signals in the machine world are specific: verified reviews in schema format, accurate and consistent data across platforms, reliable API performance. Appearing in an AI's answer is not the same as being chosen.
4. Do not assume Gen Z fully trusts agentic purchasing. The data is more nuanced than the headlines suggest. Twenty-four per cent of consumers say they will never delegate purchases to AI. Among those who do delegate, the average spend limit is £177 without additional approval. Scepticism is significant and rational.
5. Do not wait for "full autonomy" to adapt. The transition is not binary. It is not "humans shop" then "AI shops." It is a gradual shift, already underway, that rewards early preparation. Every month spent optimising for machine readability before your competitors do is a month of compounding advantage.
Ten Questions Worth Remembering — With Answers
Q1: What is the fundamental difference between AEO/GEO and AAIO? AEO and GEO optimise for being cited in AI-generated answers. AAIO optimises for being selected by an AI agent that is completing a task with real financial consequences — a purchase, a booking, a contract.
Q2: What percentage of Gen Z used AI tools for a purchase in the past year, according to PayPal's research? Sixty-one per cent — a figure that indicates mainstream adoption rather than experimental early use.
Q3: How much does Gartner project AI agents will command in B2B purchases by 2028? Fifteen trillion dollars, with ninety per cent of all B2B buying intermediated by AI by the same date.
Q4: What did Merkle's testing reveal about AI agents evaluating product categories? When AI agents were sent to find and recommend specific products, the brand websites were never visited. All research and pricing comparison happened inside the language model, using structured data and databases — not marketing copy.
Q5: What is "prompt injection" in the context of agentic commerce? A security threat in which malicious sellers embed hidden instructions in their product data or web pages, attempting to manipulate an AI agent into recommending them regardless of genuine merit.
Q6: Why might small brands actually benefit from the shift to agentic commerce? Because the criterion for agent-selection is data quality and structure, not budget or brand recognition. A small brand with impeccable, machine-readable product data can out-compete a large brand with chaotic, inconsistent information.
Q7: What did Checkout.com's June 2026 research find about consumer spending limits for AI agents? Consumers said they would allow an AI shopping agent to spend an average of £177 per purchase without requiring additional human approval — indicating significant, but not unconditional, delegation.
Q8: According to Gartner's March 2026 survey, what percentage of B2B buyers prefer a rep-free experience? Sixty-seven per cent, with forty-five per cent reporting they used AI during their most recent significant purchase.
Q9: Why does McKinsey say trust in agentic contexts is "built through explainability and reversibility"? Because consumers delegating decisions to AI are not primarily worried about whether the agent will find the right product — they are worried about losing control. Systems that can explain their reasoning and be overridden or reversed build more sustainable trust than systems that simply act.
Q10: What is the "dual audience" challenge now facing brands? Every brand now effectively has two audiences simultaneously: human beings who experience emotion, respond to story, and form attachments; and AI agents that parse structured data, evaluate trust signals, and make decisions programmatically. Strategies that serve only one of these audiences are incomplete.
Bibliography
Primary Research and Reports
-
Gartner (November 2025). AI Agents Will Command $15 Trillion in B2B Purchases by 2028. Gartner IT Symposium/Xpo 2025. digitalcommerce360.com
-
Gartner (March 2026). 67% of B2B Buyers Prefer a Rep-Free Experience. gartner.com
-
Gartner (May 2026). 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights. gartner.com
-
McKinsey & Company (January 2026). The Automation Curve in Agentic Commerce. mckinsey.com
-
Deloitte US (February 2026). Agentic Commerce: AI Shopping Agents Guide 2025. deloitte.com
-
Checkout.com (June 2026). Agentic Commerce 2026: The State of Consumer Demand and Merchant Readiness. checkout.com
-
Bain & Company (July 2025). Agentic AI Commerce Hinges on Consumer Trust. bain.com
-
EY (2026). AI Sentiment Report 2026. Referenced in CEOWORLD Magazine, May 2026. ceoworld.biz
-
Capgemini Research Institute (January 2025). What Matters to Today's Consumer — 4th Edition. capgemini.com
-
PayPal (2025). Holiday Shopping Survey. Referenced in Envive.ai: envive.ai
Sector Analyses and Platforms
-
Commercetools (April 2026). 7 AI Trends Shaping Agentic Commerce in 2026. commercetools.com
-
Commercetools (May 2026). The Agentic Commerce Radar. commercetools.com
-
OroCommerce (May 2026). Agentic AI in Commerce: The 2026 Guide for B2Bs. oroinc.com
-
SAP News (January 2026). Agentic AI Is Reshaping Commerce: The Next Frontier of Discovery, Payments, and Trust. news.sap.com
-
Mirakl (2025). Retailer AI Readiness Gap: 2026 Commerce Predictions. mirakl.com
-
Techasoft (2026). Agentic Commerce in 2026: How AI Agents Are Reshaping Buying Decisions for Retailers. techasoft.com
-
Airia (January 2026). 2026: The State of Agentic AI in Retail. airia.com
-
AI Magicx (March 2026). AAIO: How to Optimize Your Content for AI Agents in 2026. aimagicx.com
-
AI Magicx (March 2026). Agentic Commerce Explained. aimagicx.com
-
Complete AI Training (February 2026). AI Agents Will Do the Buying in 2026 — Can Your Brand Get Chosen? completeaitraining.com
Gen Z and Consumer Behaviour
-
Inriver (January 2026). Consumers in 2026: When Wellness, AI, and Gen Z Redirect Your Bottom Line. inriver.com
-
PwC (2025). Gen Z Consumer Trends — Holiday Outlook Survey. pwc.com
-
Envive.ai (2025–2026). 30 Gen Z AI Shopping Statistics for Ecommerce. envive.ai
-
Creatuity (April 2026). AI in B2B Commerce Statistics 2026: 55 Data Points from McKinsey, Gartner, Forrester. creatuity.com
Security, Trust, and Ethics
-
Consumer Reports Innovation (September 2025). Building Trusted AI Agents for Consumers. innovation.consumerreports.org
-
arXiv (2025). Machine-Readable Ads: Accessibility and Trust Patterns for AI Web Agents Interacting with Online Advertisements. arxiv.org
-
Checkout.com (2026). How to Build Consumer Trust in AI Agents. checkout.com