Picture a group of friends planning a surf trip to Bali. Instead of anyone opening a dozen browser tabs, one traveler’s personal AI assistant is quietly tracking wave forecasts in the background. The moment conditions look right, it searches flights, filters accommodation, bundles everything into a single proposal, and drops it into the group chat. Once everyone taps “yes,” that same assistant finishes the booking on its own.
That scenario opened a recent industry talk called “Owning the Experience: CX for the Agentic Era,” delivered by Tuomas Peltoniemi, Managing Director and Design & Digital Products Lead for Accenture Song in Southeast Asia — the customer-experience consulting arm of Accenture, one of the world’s largest professional-services firms. Peltoniemi wasn’t really talking about vacations. He was describing a shift that’s already reshaping who earns money, and from whom, across the freelance and small-business economy: the customer increasingly isn’t deciding alone. They’re handing pieces of the decision to an AI agent — software that can research, compare, negotiate, and sometimes even pay on a person’s behalf. If that agent can’t find you, can’t understand what you sell, or doesn’t trust what you claim about yourself, you’re effectively closed for business — no matter how good your actual work is.
This is not an abstract enterprise problem. It’s directly relevant to anyone earning through freelance platforms like Fastwork (Thailand’s homegrown freelance marketplace, functionally similar to Upwork or Fiverr but built for the local market), through Upwork, Fiverr, or Shopee and Lazada storefronts, or through a personal website. The data behind Peltoniemi’s talk — Accenture’s Consumer Pulse survey of 25,590 people across 16 countries, fielded between January 7 and 22, 2026 — maps out exactly how far this shift has already gone, and what solo operators need to do about it.
Three Eras of “Who Owns the Sale” — and We’re Now in the Third
Every 20 to 30 years, the commerce world resets the answer to a basic question: who actually controls the customer relationship? In the industrial era, companies decided what to sell and how; buyers chose from what was in front of them. When the web arrived, buyers took over — researching, comparing tabs, assembling their own decision from scratch. That took real effort, and effort is exactly what’s disappearing now. Peltoniemi’s research found that 74% of respondents said they trust a personal AI agent more than their closest friend when it comes to making a purchase decision for them — a number he admitted, on stage, is startling even to people who work in this field every day. Executives see the same trend from the other side: 70% expect AI agents to be making marketing and sales decisions directly, without a human in the loop, within three years.
This isn’t a future hypothetical. By February 2026, OpenAI reported ChatGPT had reached 900 million weekly users, with a meaningful share of that traffic already going toward product research and shopping. Analysts at Salesforce estimated AI agents drove roughly 20% of global online orders during the 2025 holiday season, worth about $262 billion, and McKinsey has projected that “agentic commerce” — AI doing the shopping — could handle $3 to $5 trillion in transactions globally by 2030. The technical machinery behind this moved fast, too: OpenAI launched in-chat “Instant Checkout” in September 2025, then pulled back from operating checkout itself by March 2026, settling into a model where AI agents handle discovery and comparison while the actual payment happens on the merchant’s own site. Google, Microsoft, and Perplexity built comparable systems the same year. The exact plumbing keeps changing. The direction doesn’t: an AI is increasingly the first thing that “meets” your business, before any human does.
The “Delegation Dial”: What Clients Will Hand Over — and What They Won’t
Here’s the part that should reassure freelancers who fear being replaced by robots entirely: people don’t delegate everything equally. Accenture built a concept it calls the Delegation Dial to map this. At one end sit purchases loaded with emotion and personal stakes — skincare, beauty, travel, anything creative or identity-linked. People are happy to have AI do the legwork here, but they still want the final call. At the other end sit recurring, low-stakes, repeat purchases — subscriptions, weekly household staples — where many people have already let an AI agent run the whole loop unsupervised.
One Hong Kong millennial in the research drew the line memorably: letting an AI agent handle flight logistics was fine, but choosing a hotel room was not, because small details — the view, the exact location, the feel of the room — mattered too much to hand over. The same pattern showed up with first-time homebuyers: choosing a neighborhood, or making the actual offer on a house, still felt too personal to delegate, while calculating affordability or arranging mortgage insurance felt like exactly the kind of task people wanted off their plate.
Translate that to freelancing and the implication is sharp. A wedding photographer, a brand designer, or a bespoke copywriter is selling something on the emotional end of the dial — clients will use AI to shortlist candidates, but they still want to click through portfolios and feel a connection before hiring. A bookkeeper, a data-entry virtual assistant, or a recurring content-writing retainer sits closer to the commodity end — and is far more likely to be selected, renewed, or replaced entirely by an AI agent comparing rates and reviews with no human glancing at the decision at all. Knowing which end of the dial your service sits on should shape how you market it.
Stop Managing Channels. Start Managing the Whole Journey
Companies — and by extension, solo operators — often build their online presence one channel at a time: a website here, a Facebook page there, a Fastwork or Upwork profile somewhere else, each maintained separately, sometimes with slightly different prices or claims. Peltoniemi argues this is now a serious liability, because an AI agent doesn’t respect those boundaries. It pulls whatever it can find, wherever it can find it, and if your Instagram bio says something different from your Upwork profile, the agent either gets confused or simply distrusts you.
Accenture’s proposed fix has two parts: orchestrated intelligence — effectively one consistent brain behind everything a customer or agent might touch, so a question gets the same accurate answer no matter where it’s asked — and composable capabilities, meaning you build a piece of information or service once (your pricing, your portfolio, your availability) and let every channel pull from that single source, rather than re-entering it five times inconsistently. In Thailand this problem shows up constantly with LINE OA (LINE Official Account, a business storefront built inside LINE, the messaging app almost everyone in the country uses) running as its own island, disconnected from a seller’s website or marketplace listing.
A bank example from the research illustrates the upside well: if a bank’s backend systems can see a customer’s full financial picture, it can proactively message that customer — before they’ve even applied — that they’re ready to buy their first home, and bundle the mortgage, insurance, and transfer steps into one flow. Marriott has built something similar by letting a guest’s personal AI agent connect directly and book rooms according to that guest’s standing preferences, with no manual searching at all. The freelance equivalent: a solo consultant whose calendar, pricing, and case studies are all accessible through one consistent, structured source is far easier for an AI agent to book confidently than one scattered across five inconsistent profiles.
แนะนำ
Can the Machine Even Read You? Computational Legibility
This is the single most actionable — and most overlooked — idea in the whole research. Peltoniemi calls it computational legibility: whether an AI agent can actually parse and understand what you’re offering. As a test case, Oreo — a cookie brand nearly everyone on earth recognizes — audited its own website and found that only 10% of AI agents crawling it could correctly read and understand its own product information. If a company with that level of global recognition struggles, a freelancer’s personal website or under-optimized marketplace profile is almost certainly worse off.
The practical fallout is blunt: no legibility means no recommendation. Agents favor businesses with structured data — machine-readable tags describing your prices, services, and availability in a standard format, rather than just pretty paragraphs a human can read but a machine can’t parse cleanly. Pages built with structured data are already being cited roughly three times more often in AI-generated search summaries than pages without it. Price transparency matters just as much: vague “contact for a quote” listings are far less agent-friendly than a clear, published rate. And agents increasingly cross-check your claims against outside evidence — OpenAI, for instance, has connected directly to Reddit so it can verify near-instantly whether a business’s public reviews actually match what it claims about itself.
There’s a cautionary tale here worth knowing: Amazon has reportedly blocked some AI crawlers, like ChatGPT’s shopping bot, from indexing its product listings — meaning Amazon-hosted products are largely invisible inside ChatGPT’s shopping results, an accidental gift to competitors willing to stay open. The lesson for a small operator: check whether your own website’s crawler settings are quietly blocking the very AI systems your future clients are using to find you. If you don’t know the answer, that’s the first thing to fix.
Freelance marketplaces are racing to build their own version of this. Upwork’s AI system, “Uma,” now scores interviews and drafts contracts, not just matching proposals. Fiverr has built its own AI assistant, “Fiverr Go,” and reported AI-related order growth of over 100% year over year in early 2026. Upwork separately reported that demand for AI-related freelance skills grew 109% year over year, with AI video editing alone surging 329% to become the platform’s fastest-growing category. One consistent finding across several 2026 trust studies is worth remembering here too: buyers overwhelmingly want AI involvement disclosed upfront, and discovering hidden AI use later damages trust far more than honest disclosure ever does.
Loyalty Now Comes With Conditions
Peltoniemi shared a personal example: he’s bought the same sneaker brand for 20 years out of habit, but admitted that if his own AI agent recommended a better-fitting alternative tomorrow, he’d probably take the suggestion. The data backs this up — 37% of self-described “loyal” customers said they’d switch brands the moment an AI agent found something that better matched their needs. Loyalty hasn’t disappeared; it’s just become conditional, and the entity setting the conditions is increasingly a machine comparing your actual track record against your competitors’ in real time.
For a freelancer, this means a client relationship you thought was secure can quietly erode if a competitor’s reviews, delivery times, or pricing become more machine-legible than yours — even if your actual work quality hasn’t changed at all.
Beyond Thailand: The Same Story in Vietnam, Indonesia, and Globally
This shift isn’t confined to any one country. Indonesia’s e-commerce market, the largest in Southeast Asia at roughly $67 billion in 2025, was reshaped by the 2024 merger of Tokopedia (Indonesia’s largest homegrown marketplace) and TikTok Shop, a combination that now commands close to 28% of regional platform sales, driven heavily by AI-assisted product discovery and livestream selling. Shopee, the region’s dominant marketplace, has rolled out AI ad tools branded “GMV Max” and “Brand Max” that automate targeting for sellers across Thailand, Vietnam, Indonesia, Malaysia, and the Philippines. Broader estimates suggest AI could add as much as $1 trillion to Southeast Asia’s economy by 2030. None of this is unique to any single market — it’s the same agentic shift playing out everywhere at once, just with different local platforms wearing the front end.
The Takeaway: Treat AI Legibility Like the New SEO
If you earn money online — through freelancing, a small storefront, or personal-brand consulting — the practical to-do list here is short and doable without a technical team. Make your pricing, availability, and services legible in a consistent, structured way everywhere you appear online, not just readable to humans. Check whether your website is accidentally blocking AI crawlers. Keep your reviews honest and your claims verifiable, since agents now check both. And be deliberate about where your service sits on the Delegation Dial: commodity work should be priced and packaged for machine comparison, while emotionally driven, high-trust work should lean harder into the human story an algorithm still can’t fully replicate. The freelancers and small sellers who treat “is the AI reading me correctly” as seriously as they once treated Google search ranking are the ones who’ll keep getting chosen — by humans and their AI agents alike.
แนะนำ
Key Takeaways
- AI agents are increasingly the first “customer” your business meets — 74% of people say they trust a personal AI more than their best friend for purchase decisions.
- People delegate selectively: emotional, high-stakes purchases stay human-controlled, while recurring, commodity-style services get handed fully to AI — know where your service sits.
- Keep pricing, portfolios, and claims consistent across every platform; agents pull from all of them and punish inconsistency.
- “Computational legibility” — whether AI can actually read your site or profile — is the new SEO; structured data and transparent pricing directly affect whether you get recommended.
- Client loyalty is now conditional: 37% of “loyal” buyers say they’d switch if an AI agent found a better-fitting alternative, so verifiable reviews and delivery history matter more than ever.
Frequently Asked Questions
Q: Can AI agents actually make purchases or hire freelancers on someone’s behalf in 2026?
A: Yes, to a growing degree — platforms like ChatGPT, Google, and Microsoft Copilot now support AI-assisted discovery and, in many cases, completed checkout, though most transactions still finish on the merchant’s own site rather than fully inside the chat.
Q: What is “computational legibility” and why does it matter for a small business?
A: It refers to whether an AI system can correctly read and understand your website, pricing, and offerings; even major brands like Oreo found only 10% of AI agents could parse their site correctly, meaning smaller, less-optimized sites are likely worse off.
Q: Will AI agents replace freelancers entirely?
A: Not entirely — research shows people still want to personally choose freelancers for emotionally significant, high-stakes work, but they’re increasingly comfortable letting AI fully select and manage commodity-style, recurring services.
Q: How can a freelancer make their profile more visible to AI agents like ChatGPT or Claude?
A: Use structured, machine-readable data for pricing and services wherever possible, keep information consistent across every platform you’re listed on, and check that your website isn’t accidentally blocking AI crawlers.
Q: What is the Delegation Dial mentioned in the Accenture research?
A: It’s a framework describing how much decision-making power people are willing to give an AI agent, ranging from full control over routine, recurring purchases to only research help on emotional, high-value decisions.
Q: Are freelance marketplaces like Upwork and Fiverr already using AI to match clients and freelancers?
A: Yes — Upwork’s AI assistant “Uma” now scores interviews and drafts contracts, and Fiverr’s “Fiverr Go” assistant has coincided with reported AI-related order growth of over 100% year over year in early 2026.
Q: Does disclosing AI use in freelance work hurt a freelancer’s chances of getting hired?
A: Available 2026 research suggests the opposite — buyers consistently say they want AI use disclosed, and finding out later that AI was used without disclosure damages trust more than honest disclosure does.
Q: How big is the AI-driven shopping trend globally?
A: Analysts estimated AI agents drove around 20% of global online orders during the 2025 holiday season, worth roughly $262 billion, with projections suggesting agentic commerce could reach $3 to $5 trillion globally by 2030.
Q: Is this AI-agent shopping trend happening in Southeast Asia too, or mainly in the US?
A: It’s happening region-wide — Indonesia’s Tokopedia-TikTok Shop merger, Vietnam’s fast-growing e-commerce sector, and Shopee’s AI ad tools across the region show the same shift toward AI-assisted discovery and selling.
Q: What is Fastwork, and how does it compare to Upwork or Fiverr?
A: Fastwork is Thailand’s homegrown freelance marketplace, connecting local freelancers with clients much like Upwork or Fiverr do internationally, though built specifically around the Thai market and payment habits.
Q: Why would a loyal customer suddenly switch to a different freelancer or brand because of AI?
A: Because AI agents evaluate options based on verifiable evidence — reviews, delivery history, pricing — rather than emotional habit, so a competitor with stronger documented performance can outrank a long-standing favorite in an agent’s recommendation.
Q: What’s the single most useful first step for a freelancer worried about AI agent visibility?
A: Start with an “AI discoverability audit” — check whether AI crawlers can actually access and correctly read your website, portfolio, and pricing, since many small business sites unintentionally block the very systems now doing the recommending.
