Your Next Customer Won’t Leave A Voicemail

Oct 7

Blame it on OpenClaw. One developer's hobby project taught chatbots to become agents, assistants that do things instead of just saying things, and the big labs followed. Now those agents are showing up all over the internet to shop, research, and book on people's behalf. The problem is most small business sites were built for a different kind of visitor.

Picture someone asking their phone, "Find me a plumber who can come Thursday, with a call-out fee under $150, and book the best one." The assistant checks a handful of plumbers. One site lists its hours in a photo of a handwritten sign and says "call for pricing." Another has a contact form that emails the owner, who promises answers next day. A third lists its services and prices in plain text, shows their calendar with open Thursday slots, and confirms a booking on the spot. The phone books the third plumber. The first two never even know they were considered.

Two Front Doors

With the release of Meta's Muse, their personal agent app, that scene is already starting to play out in small pieces. People are using Muse to book reservations and appointments, track price drops, and cancel forgotten subscriptions. One user says it negotiated a lower internet bill for them. As people come to use and trust it more, we'll only see it, and similar agent apps, playing a larger part in their online lives.

On the web side, Google's Chrome team announced a browser standard called WebMCP in February that lets a website declare actions an agent can take, such as booking or comparing products, instead of leaving the agent to guess its way around the page. In March, Shopify turned on agent-ready storefronts by default for eligible US merchants. The pattern is the same, sites growing second access points, one for people and one for the AI working on their behalf. Some services will skip the human interface entirely, building only for agents to query. For many businesses, the practical version is one site with both doors.

That's not to say it's necessarily time to run out and revamp your entire website. WebMCP is still in browser trials, far from universal, and agents already read ordinary pages. They just do better with pages that were made for them.

In the future though, it's easy to envision another internet, an upside-down version for you sci-fi fans, that mirrors the one we see, but is created solely for the benefit of AI agents. This version will be filled with no-frills, completely utilitarian sites that allow AI to easily parse and ingest their content. Instead of splash pages and banners, the sites will emphasize their available tools and points of access for their data. These sites will do what service entrances do for the maids and butlers in the dramatic period shows, allow the ones doing the work to perform a task and get on with it without disturbing the rest of us.

I think in the coming years this will be as normal and ubiquitous as an online store with a shopping cart.

Now, I've been very vocal that AI Agents have issues that need to be addressed before I consider them trustworthy to take on vital tasks in a business. And with the creation of consumer facing apps like Muse, we're seeing those shortcomings play out in very public ways. They weren't ready, but they're here. So how do we account for their mistakes and take best advantage of their capabilities?

Here are some suggestions...

The Unglamorous Checklist Of Agentic Access

The fixes aren't flashy, which is why they pay off:

Put the basics in text. Hours, services, service area, and price ranges belong in plain text on the page, where software can read them. A photo of a menu or a scanned price sheet looks fine to a person and is harder for an agent to rely on.
Make the request finish itself. A booking or quote form should confirm automatically, put the appointment on a real calendar, and show real availability. An agent comparing five businesses won't wait for the one that needs a callback.
Keep every listing in agreement. Your website, Google Business Profile, and directory listings should say the same thing about hours and prices. Conflicting details make any agent, or any person, unsure who to trust.
Answer the real questions in plain language. A short FAQ built from what customers actually ask gives an agent clean answers to work from.
Decide what an agent may do alone. Reading a price list is harmless. Accepting a booking, issuing a quote, or taking a payment deserves a rule about when human approval is needed.
Decide what only AI sees. Agents can be steered by hidden text on a page, so the same caution applies to the tools you use to create the page as to the ones visiting you.

Behind the Door

Access only matters if everything behind it works together. When a booking lands, the calendar entry, the confirmation text, the reminder, the job record, and the invoice should all happen without anyone retyping the details. That's the same grunt work already eating hours, and an agent customer raises the stakes.

This is where a business that has their workflows and data already prepared can really shine. Automation and integration are what this new customer is looking for first. Without them the product or service isn't even considered. Requests arrive at any time, day or night. Answers are expected in seconds. A business that depends on someone checking an inbox falls behind. The work was worth automating for time savings alone. Now there's a customer who directly rewards it.

This Sounds Like Science Fiction

I'm the first to admit that all of this is still speculation, and no one can say exactly what shape it will take. But it is a fact that the agents have arrived, and the framework is already going up with WebMCP in the browser and agent-ready storefronts on Shopify. Also, we've seen a shift like this before. When the first shops appeared on the internet in the 1990s, plenty of people called them a novelty, and plenty of customers hesitated to type a card number into a website. The owners who put up a storefront early had no way of knowing how big the web would get. But the ones who took it seriously, and created working sites that delivered their products in a way that customers understood, flourished.

It's very possible that the first AI Amazon is being built right now. It won't be perfect at first, but the businesses that are willing to try, and to iterate on the changing standards, will be the ones best positioned to take advantage of the changing times.

Agents aren't perfect, but they're here. Those organizations that are willing to work sensibly and securely with them could be tapping into a vast well of new customers. Just make sure you're ready when they come.

Sam Hardin is the founder of Bright Pivot LLC, an AI automation and operations consultancy based in Covington, Louisiana. Bright Pivot helps small and medium businesses on the North Shore automate the routine and augment the rest.

Have questions about what this means for your business? I'd love to talk it through. Book a time on my Calendly, or reach me directly at sam@bright-pivot.com.