What AI Retail Distribution Tells Us About the Future of Selling Services

Last updated: March 10, 2026

Every major distribution shift in the last 25 years followed the same pattern: retail first, services later.

E-commerce started with books and electronics, then moved to insurance comparison sites and online banking. Mobile commerce launched with retail apps, then reshaped how people apply for loans and manage policies. Marketplaces began with products (Amazon, eBay), then expanded to services (Uber, Upwork). Social commerce started with Instagram shopping tags, then financial influencers started driving insurance and investment decisions.

AI distribution is following the same trajectory. Retail is already 18-24 months ahead. The features rolling out for product shopping today are a preview of what’s coming for insurance, banking, lending, and B2B SaaS.

This article breaks down the specific AI retail features already live, what they do, and what the equivalent looks like for services.

The Pattern: Retail Leads, Services Follow

The gap between retail and services adoption has shortened with each wave. E-commerce took a decade to cross over. Mobile took five years. AI distribution is moving faster because the infrastructure (MCP, commerce protocols) is being built for both simultaneously.

AI Shopping Assistants: Conversational Product Discovery

What retail has now

The first wave of AI retail distribution is conversational shopping: AI assistants that help customers browse, compare, and choose products through natural conversation instead of search and filters.

Amazon Rufus is the most advanced example. Launched in early 2025, Rufus is an AI shopping assistant embedded directly in the Amazon app. It answers product questions, compares options, summarizes reviews, and makes personalized recommendations based on the customer’s needs and purchase history.

The scale is significant: Rufus handles hundreds of millions of shopping queries, and Amazon attributed approximately $12 billion in influenced sales to AI-assisted shopping in 2025. Rufus doesn’t just answer questions; it understands context. Ask “what do I need for a camping trip?” and it builds a personalized list based on where you’re going, how long you’ll be out, and what you already own.

ChatGPT Shopping Research takes a different approach. Instead of being embedded in a single retailer, ChatGPT acts as a cross-retailer shopping assistant. Users ask questions like “best noise-cancelling headphones under $300” and get structured product cards with images, prices, reviews, and direct links to buy.

OpenAI reports over 50 million shopping queries per week on ChatGPT, with no ads or sponsored placements. Products surface based on relevance, reviews, and price. This is discovery driven by conversation, not by ad spend.

What this means for services

The services equivalent of AI shopping assistants is an AI that can answer real questions about your product using live data, not training data from months ago.

For an insurance company, this means a customer asking Claude “what home insurance options exist for a two-bedroom apartment in Madrid?” gets an answer that includes your actual products, with real coverage details and pricing ranges, because your product data is connected to the AI through MCP.

For a B2B SaaS company, it means a product manager asking ChatGPT “which project management tools support Jira integration for teams of 50?” gets your product in the response with accurate pricing, current feature availability, and a path to start a trial.

The infrastructure for this already exists. WaniWani deployed the first insurance quoting app on ChatGPT in early 2026 for Tuio, a Spanish digital insurer. The app connects to Tuio’s pricing engine via MCP, answers product questions with live data, and generates personalized quotes inside the conversation. 15%+ of Tuio’s new business now comes from AI conversations.

The takeaway: Retail has conversational discovery at scale. Services can have the same thing today, but only if your product data is connected to AI platforms through MCP.

In-Chat Checkout: What Retail Tried, and What It Learned

What retail attempted

The second wave was supposed to be transactional: not just discovering products through AI, but completing the entire purchase without ever leaving the conversation.

ChatGPT Instant Checkout launched in September 2025 as a partnership between OpenAI and Stripe. When a user found a product through ChatGPT Shopping, they could buy it directly inside the conversation using Stripe’s checkout flow. No redirect to the retailer’s website. No separate cart. The transaction happened in the chat. Merchants on Shopify, Etsy, and other Stripe-connected platforms could enable Instant Checkout without building custom integrations.

By March 2026, reports emerged that OpenAI was pulling back. According to The Information, only about 12 of over one million eligible Shopify merchants had integrated. Sales tax collection had reportedly never been built. Inventory sync reportedly never worked at scale. Forrester confirmed that completing a purchase inside an AI platform was the least-adopted use case among regular users. An OpenAI spokesperson reportedly described Instant Checkout as “transitioning to apps, where purchases can occur more seamlessly.”

The market corrected toward merchant-owned apps. Instacart, DoorDash, Target, Expedia, and Booking.com now have their own apps inside ChatGPT, where the merchant owns the transaction experience, the customer relationship, and the data. This mirrors what happened with Google’s “Buy on Google” feature, which launched in 2015, attracted only about 8,000 sellers despite years of effort, and was eventually killed. Discovery platforms are good at discovery. The transaction belongs with the merchant.

Perplexity Buy with Pro still takes a direct checkout approach. Perplexity’s “Buy with Pro” feature lets subscribers purchase products directly from search results, powered by a PayPal integration. Whether this model proves more durable remains to be seen.

What this means for services

The lesson from retail’s checkout experiment is clear: the transaction layer should live where trust and expertise already exist, inside the merchant’s own experience.

Services never had a “buy now” button in the first place. A home insurance policy requires underwriting. A B2B SaaS contract requires negotiation. A mortgage requires verification. What looked like a limitation turns out to be an advantage: services companies were always going to need merchant-owned AI apps, not checkout buttons.

The services equivalent is a quote-to-lead flow that happens entirely inside the AI conversation. Instead of “Buy now,” it’s “Get a personalized quote.” Instead of a Stripe checkout, it’s a lead capture with full context about what the customer asked, what they were quoted, and what they need.

This is already happening. When a customer asks ChatGPT “how much would home insurance cost for my apartment in Barcelona?”, the Tuio AI app generates a real premium estimate based on the customer’s inputs, presents coverage options, and captures the customer’s email to send a formal quote. The customer never visits a website. The lead arrives in Tuio’s CRM with the full conversation context.

The conversion advantage is substantial. AI-sourced traffic converts 3-6x higher than traditional search traffic (Morningstar/PR Newswire, 2026). The reason is simple: by the time a customer gets a personalized quote inside an AI conversation, they’ve already been qualified, informed, and engaged with specific numbers. That’s a fundamentally different lead than someone clicking a Google ad.

The takeaway: Retail tried to close the sale natively in the chat and learned the hard way that the merchant needs to own the conversion. Services companies that build merchant-owned AI apps today are building on the model that retail is now converging toward.

Commerce Protocols: The Infrastructure Layer

What retail has now

Behind the visible features (shopping assistants, checkout flows) is an infrastructure layer that makes AI commerce work at scale. Two protocols are defining how AI agents interact with merchant systems.

Agentic Commerce Protocol (ACP) is the open standard (Apache 2.0) that OpenAI built with Stripe. Originally designed to power ChatGPT Instant Checkout, ACP defines how merchant apps interact with AI platforms: product discovery, catalog browsing, cart management, checkout, payment processing, order tracking, and returns. With native checkout reportedly being scaled back, ACP’s primary role is shifting toward powering merchant-owned apps like Instacart, DoorDash, and Target inside ChatGPT.

Universal Commerce Protocol (UCP) is Google’s response, developed with Shopify, Walmart, and Target. UCP is designed for Gemini and Google’s AI ecosystem, enabling AI agents to interact with merchant systems across discovery, comparison, purchase, and fulfillment. Where ACP is tied to Stripe’s payment rails, UCP is designed to be payment-agnostic and works across a broader set of retail partners.

These protocols matter because they standardize how AI talks to businesses. Before ACP and UCP, every AI-to-merchant connection was custom. Now, a merchant connects once and is accessible across every AI agent that speaks the protocol.

What this means for services

The services equivalent is MCP, the Model Context Protocol created by Anthropic. While ACP and UCP focus on product transactions (catalogs, carts, payments), MCP is designed for the broader case: enabling AI assistants to interact with any external service through structured tools, resources, and context.

For services companies, MCP is what makes AI distribution possible. It defines how an AI assistant can:

1. Discover what your service offers (product catalog, coverage options, pricing tiers)
2. Query your systems in real time (eligibility checks, personalized quotes, feature availability)
3. Capture customer information (lead details with full conversation context)
4. Maintain compliance (audit logging, disclosure requirements, regulatory controls)

The key difference: ACP and UCP were designed for retail product transactions with fixed prices and instant fulfillment. MCP handles the discovery-to-lead pipeline because services require personalization, underwriting, or human follow-up before a transaction closes. But with retail’s own shift toward merchant-owned apps, the models are converging. All three protocols now assume the merchant owns the experience; they differ in what that experience looks like.

The trajectory is clear. As AI commerce matures, services will eventually support full in-conversation transactions for simpler products (travel insurance, basic SaaS subscriptions, standard loan products), just as retail started with simple checkouts and expanded to complex configurations.

The takeaway: Retail has two competing commerce protocols (ACP, UCP) enabling AI-native transactions. Services have MCP enabling AI-native discovery and quoting. The infrastructure is being built now; the companies that connect early will have a structural advantage.

AI-Native Discovery: From Search to Conversation

What retail has now

The way customers find products is shifting from search engines to AI conversations. This is not a future prediction; it’s measurable today.

AI-referred traffic to retail sites grew 805% year-over-year through late 2025 (Adobe Analytics). During Black Friday 2025, AI influenced $14.2 billion in online spending. 74% of US consumers used AI for shopping-related activities in 2025.

Retailers that optimized for AI discovery early (structured product data, fast APIs, rich descriptions) captured disproportionate traffic. This mirrors the early days of SEO: the companies that understood Google’s algorithm first built positions that were expensive to displace later.

The shift is from keyword-based search (“best home insurance Spain”) to conversational queries (“I just bought a two-bedroom apartment in Madrid and I need to insure it, what are my options?”). AI can handle the second query and provide a useful answer. Traditional search cannot.

What this means for services

51% of US consumers already turn to AI for financial advice or information (ABA Banking Journal / FNBO Financial Wellbeing Study, 2025). The demand is there. The question is whether your service shows up in the answer.

For most services companies today, it doesn’t. AI assistants rely on training data, which may be months old, incomplete, or inaccurate. Your pricing changed last quarter? The AI doesn’t know. You launched a new product? The AI hasn’t seen it. A competitor has better-structured data? The AI recommends them instead.

Connecting your service to AI platforms through MCP solves this. The AI pulls live data from your systems instead of relying on stale training data. Your products appear with current pricing, accurate features, and real eligibility criteria.

Google Analytics UTM tracking captures only about 25% of actual AI-sourced traffic, meaning most companies underestimate this channel by 4-5x (WaniWani attribution analysis, 2026). Companies that measure AI traffic properly and optimize for it are seeing 15-20% of new business come from AI conversations.

The takeaway: Retail is already seeing massive traffic shifts from search to AI. Services will follow the same pattern. The companies building AI presence now will capture positions that are hard to displace once the channel matures.

Agentic Commerce: AI That Buys for You

What retail has now

The most advanced wave of AI retail distribution is fully autonomous purchasing: AI agents that don’t just recommend products but actually buy them on the customer’s behalf.

Amazon’s Buy for Me feature, launched in 2025, lets customers purchase products from third-party websites directly through Amazon’s AI. The AI navigates to the external site, adds the product to cart, enters shipping and payment details, and completes the purchase. The customer never leaves Amazon.

Google’s AI Mode Shopping in Search includes agentic capabilities where Gemini can compare products across retailers, apply available coupons, and initiate purchases through the Universal Commerce Protocol. The AI acts as a purchasing agent, not just an information assistant.

This is the endgame of AI commerce: the AI becomes the customer. It shops, compares, negotiates, and buys. The human sets the criteria and approves the transaction.

What this means for services

Autonomous AI purchasing for services is further out than retail, for good reason. You can’t have an AI autonomously buy a life insurance policy or sign a SaaS contract without human review. The stakes are higher, the products are more complex, and the regulatory requirements are stricter.

But the intermediate step is already visible: AI agents that autonomously gather quotes, compare options, check eligibility, and present a shortlist for human decision. Instead of a customer visiting five insurer websites, filling out five quote forms, and comparing PDFs manually, an AI agent queries five insurers’ MCP servers, generates five personalized quotes, compares them on the criteria the customer cares about, and presents a recommendation.

This is not theoretical. The infrastructure (MCP) already supports it. The limitation today is adoption: most services companies haven’t connected their pricing and product data to AI platforms yet.

The takeaway: Retail is moving toward AI agents that buy autonomously. Services will move toward AI agents that shop autonomously and present recommendations for human approval. The companies whose data is accessible through MCP will be the ones these agents can work with.

The Timeline: Where Services Stand Today

Based on the retail-to-services pattern, here’s where AI distribution stands for services:

Services are roughly 12-18 months behind retail in adoption, but the infrastructure gap is much smaller. MCP already supports everything needed for services distribution: product discovery, real-time quoting, eligibility checking, lead capture, and compliance logging.

The bottleneck is not technology. It’s adoption. Most services companies haven’t connected their systems to AI platforms yet. The ones that do it now will have the same structural advantage that early SEO adopters had in 2005: a position that compounds over time and becomes expensive to displace.

What to Do Now

If you sell a complex, quote-based service (insurance, banking, lending, B2B SaaS), the retail playbook tells you exactly what’s coming and what to build for:

1. Connect your pricing and product data to AI platforms via MCP. This is the foundation. Without it, AI assistants can’t recommend your service with accurate, current information.
2. Enable in-conversation quoting. When a customer asks an AI about your category, you want to be the service that returns a personalized answer, not a generic range from training data.
3. Capture leads with context. AI-sourced leads convert 3-6x higher because they arrive with full conversation context. Build the lead capture into your MCP integration, not as an afterthought.
4. Measure AI as a channel properly. UTM tracking underestimates AI traffic by 4-5x. Set up proper attribution before you decide the channel isn’t worth investing in.
5. Build compliance infrastructure now. Regulators classified comparison websites as insurance intermediaries. They will classify AI assistants the same way. The companies with audit trails, disclosure enforcement, and regulatory logging will have a structural advantage when clarity arrives.

Retail didn’t wait for AI distribution to be “proven.” Amazon launched Rufus. OpenAI launched native checkout and is reportedly pivoting to merchant apps. Google built UCP. The infrastructure is live, the model is settling, and it favors companies that own their AI presence.

Services companies that wait for the channel to mature before participating will find themselves in the same position as companies that ignored SEO in 2005 or mobile in 2010: playing catch-up in a game where the early movers already captured the best positions.

WaniWani provides the AI distribution infrastructure that gets services companies live on ChatGPT, Claude, and every AI platform in as few as two weeks. Request a demo to see how it works for your business.

FAQ

Is AI distribution only relevant for insurance companies?

No. AI distribution applies to any company with complex, quote-based products that require personalization. Insurance is the first vertical with live deployments (Tuio on ChatGPT), but banking, lending, B2B SaaS, and professional services all follow the same pattern. Any service where customers ask “how much would this cost me?” is a candidate for AI distribution.

How is this different from optimizing for AI visibility (GEO)?

GEO (Generative Engine Optimization) focuses on monitoring and improving how AI mentions your brand. That’s the top of the funnel. AI distribution goes further: it connects your actual pricing engine and product data to AI platforms, enabling personalized quotes, eligibility checks, and lead capture inside AI conversations. GEO tells you if AI recommends you. AI distribution makes AI able to sell for you.

Do I need to rebuild my tech stack to support MCP?

No. MCP connects to your existing APIs. If you already have a pricing API that powers your website’s quote flow, the MCP server calls the same endpoints. If your pricing logic is embedded in a monolithic application, you need to extract it into a callable API first, but the MCP layer sits on top of your existing systems, not instead of them.

Which AI platforms support MCP today?

As of March 2026, MCP is supported or emerging across ChatGPT (via OpenAI’s integration program), Claude (by Anthropic, who created the protocol), Gemini (Google is developing WebMCP), and Perplexity. You build one MCP server and deploy across every platform that supports the protocol.

How long before AI agents start autonomously shopping for services?

The infrastructure is ready today. MCP already supports the queries an AI agent would need to compare services: product information, eligibility checks, personalized quotes. The limitation is adoption (most services aren’t connected yet) and regulation (autonomous purchasing of financial products requires regulatory frameworks that don’t exist yet). Expect AI-assisted comparison (agent gathers quotes, human decides) within 12 months and more autonomous flows within 24-36 months for simpler products.

What about compliance and regulation?

Regulators will classify AI-mediated service distribution the same way they classified comparison websites: as distribution activity subject to licensing and conduct rules. EIOPA is publishing AI-specific guidance. The NAIC is drafting model laws. EU AI Act enforcement begins in 2026. Companies building compliance infrastructure now (audit logging, disclosure enforcement, synthetic testing) will have a structural advantage. Read our deep dive on compliance in AI distribution.

How do I measure if AI distribution is working?

Standard UTM tracking underestimates AI-sourced traffic by 4-5x. Proper measurement requires combining UTM data with declarative attribution (“How did you hear about us?”), MCP server analytics (how many queries and quotes your AI app handles), and lead source tracking. WaniWani’s platform provides full-funnel analytics across all AI platforms from a single dashboard.

What does it cost to get started?

It depends on your starting point. Companies with existing pricing APIs can be live on ChatGPT in as few as two weeks. Companies that need to build the API layer first have a longer path. WaniWani charges a one-time setup fee for AI app development plus a recurring platform fee for infrastructure access (analytics, compliance, multi-platform deployment). Request a demo for pricing specific to your setup.