The Experience Has Left the Building
For the better part of a decade, “improving the customer experience” meant hiring a UX consultant, redrawing a journey map, or A/B testing a button color. The assumption was always the same: a human being was on the other end, clicking through your website, filling out your form, waiting on hold. Design it better, and they’d feel it.
That assumption is breaking down. Fast.
Shopify merchants can now set up their product catalog once and sell inside a ChatGPT conversation — no separate integration, no custom app, no additional storefront. The buyer didn’t navigate to a website. An AI agent surfaced the product, answered their questions, and completed the transaction inside the conversation they were already having. Shopify and Google co-developed an open standard — the Universal Commerce Protocol — so that agents across platforms can natively complete checkout on a customer’s behalf. Hundreds of millions of ChatGPT users can now shop from Shopify merchants without ever leaving the chat window. That’s not a better checkout experience. That’s a different model entirely. The experience didn’t get redesigned. It got executed by something that never sleeps, never gets impatient, and never forgets what you told it last time.
The B2C proof of concept is already live. B2B is next.
What Shopify demonstrated in retail, agents are now building toward in procurement. Forrester predicts that buyers’ procurement teams will deploy agents capable of scaling negotiation across hundreds of suppliers simultaneously — turning static pricing pages into dynamic, real-time negotiation interfaces. A mid-market manufacturer’s purchasing manager won’t manually compare three vendor quotes and follow up by email. Their agent will evaluate all three, score them against pre-defined criteria, and initiate the next step — potentially before your sales rep has even opened their laptop.
This is where the conversation gets real for revenue leaders. If an AI agent is evaluating your quote response on behalf of the buyer, your 48-hour turnaround and a PDF attachment aren’t a minor inconvenience. They’re a disqualification. AI agents don’t call to clarify missing information or give you the benefit of the doubt on a vague line item. They move to the next supplier with complete, structured, machine-readable data. The experience your buyer has isn’t just shaped by your sales team anymore — it’s shaped by whether your systems can respond to agents at the speed and structure they require.
The scaffolding is being built in real time.
Three things happened in the last few months that, taken together, signal how fast this infrastructure is moving.
Zapier’s CEO announced the open beta of the Zapier SDK on April 7th, calling it the most powerful thing the company has shipped in years. Point your coding agent at it and it gains programmatic access to 8,000+ app integrations — Salesforce, Slack, NetSuite, you name it — without a single OAuth setup. Agents can now do things across your entire software stack through a single, standardized interface. The integration complexity that used to require months of development work is becoming a configuration problem.
Anthropic launched Claude Cowork in January, extending the same agentic architecture that powers Claude Code to everyday knowledge workers. It runs on desktop, works inside your local files and applications, synthesizes information across sources, and completes multi-step tasks without the user managing each step. In March, they added full computer use — Claude can now open apps, fill spreadsheets, export files, and complete workflows autonomously while you’re in a meeting. One demo showed a user running late, messaging Claude from their phone, and Claude exporting a pitch deck as a PDF and attaching it to a meeting invite before they walked in the door. That’s not a productivity feature. That’s a new category of coworker.
What connects all of this — Shopify’s agentic storefronts, Zapier’s SDK, Claude Cowork — is that the tooling for agent-mediated experience is now available to anyone. The barrier isn’t technology anymore. It’s whether your underlying systems are ready to participate.
The infrastructure problem hiding in plain sight.
Here’s the part most companies aren’t talking about yet: the companies that will win in an agent-mediated world aren’t the ones with the best-designed experiences. They’re the ones whose data is clean, their pricing rules are structured, and their systems can respond to an automated request in real time.
For mid-market manufacturers and distributors, this is where the stakes are highest. Your product catalog, your CPQ configuration, your quote-to-cash process — these aren’t back-office concerns. When an AI agent is evaluating suppliers on behalf of a procurement team, your ability to surface accurate pricing, available inventory, and complete product specs automatically is your experience. A CPQ that runs on batch processing and exports to Excel isn’t just slow. It’s architecturally incompatible with how buyers will increasingly be buying.
The companies investing in clean revenue infrastructure today — structured product data, real-time pricing logic, connected quote-to-cash systems — are building the foundation that agents run on. Everyone else is hoping the old way holds long enough to figure it out later.
Design doesn’t disappear. It bifurcates.
None of this means great UX stops mattering. It means the canvas has split — and the highest-leverage design work looks different than it did three years ago.
There are now two distinct surfaces that need to be designed well, and they require completely different thinking. The first is agent-facing: structured data, clean product schemas, machine-readable pricing, brand voice captured in a knowledge base an agent can actually read and act on. Shopify’s agentic storefronts let merchants define exactly how their products, policies, and brand identity appear inside AI conversations. That’s a design problem. It just doesn’t look like a Figma file.
The second surface is where human-facing design gets more important, not less. When routine transactions get handled by agents, what’s left are the high-stakes, high-judgment moments — a complex configuration decision, a contract negotiation, an onboarding experience, a renewal conversation that could go either way. Those touchpoints need to be exceptional precisely because they’re now chosen interactions, not forced ones. The buyer who shows up in person, so to speak, is already past the commodity layer. They’re there because the decision matters. Mediocre UX at that moment is no longer forgivable because “the process is just clunky.” The clunky parts got automated. What remains is a direct reflection of your brand.
The companies that will define this era aren’t choosing between agent-readiness and great design. They’re doing both — and they understand that the two are now inseparable.
What this means right now.
The experience era isn’t ending. It’s being handed off — and bifurcated. Routine interactions are moving to agents. Human-facing moments are becoming rarer and more consequential. And your brand’s ability to show up in an agent-mediated conversation depends entirely on whether the underlying architecture is ready.
Better UX alone won’t save you if your data is a mess. But clean infrastructure without a point of view on the human moments you still own is just plumbing. The companies that win will have both — and they’ll have started building before most of their competitors realized the canvas had changed.