The Infrastructure of AI Commerce Is Being Built Right Now
And most businesses aren't paying attention to the right things.
This week produced three stories that, on the surface, look unrelated. Visa launched a new payments platform. OpenAI hit a revenue milestone on its advertising business. And a major consulting firm dropped a study about which companies are actually making money from AI.
Read them separately and they’re interesting. Read them together and they’re a roadmap.
Here’s what happened and what it means for your business.1. Visa Built the Payment Rails for AI Agents
On April 8, Visa launched a platform called Intelligent Commerce Connect. The announcement didn’t get nearly the attention it deserved.
Here’s what it does: it lets AI agents browse merchant catalogs, select products or services, and complete purchases on your behalf, without a traditional checkout flow. You set the rules. The agent executes. Visa handles the payment infrastructure underneath.
The platform is in pilot now with a set of select partners and is expected to be broadly available by June. It works across card networks, not just Visa cards. And it’s already compatible with four competing AI commerce protocols being built by OpenAI, Google, Stripe, and Visa itself. By supporting all four, Visa is positioning itself as the neutral payment layer no matter which protocol wins out.
This isn’t a distant future story. McKinsey forecasts AI agent-driven sales will surpass $5 trillion by 2030.
What this means for service business owners:
If you run a business that sells anything: products, bookings, subscriptions, services, the question you need to start asking is: is my business discoverable to an AI agent?
When a person uses an AI agent to find and book a photographer, or hire a copywriter, or order supplies for their studio, the agent is going to pull from merchant catalogs and structured data. If your business isn’t showing up in those environments, you won’t be in consideration. Full stop.
This is not something you need to implement today. But it is something you need to understand now so you’re not scrambling in 18 months. The businesses that will win in agentic commerce are the ones who got their information organized and accessible early.
2. ChatGPT Is Now an Advertising Platform, and It’s Opening to Anyone
OpenAI’s advertising business, which launched as a limited pilot in February, just crossed $100 million in annualized revenue. This month, the company is expected to open self-serve access, meaning any advertiser will be able to buy ads inside ChatGPT without going through a managed sales process.
To put that milestone in context: $100 million in annualized revenue in roughly six weeks is a faster ramp than most new ad products have ever seen. For comparison, the 600 brands in the early managed pilot were spending between $50,000 and $100,000 per campaign just to participate.
Self-serve access will change that. It should bring the barrier down significantly and open the platform to mid-market advertisers, not just Fortune 500 brands.
OpenAI is projecting $2.5 billion in advertising revenue for 2026. Their internal roadmap targets $100 billion annually by 2030, which would make it competitive with the largest advertising platforms in the world. Those projections assume the company reaches around 2.75 billion weekly users, ambitious, but worth noting that it already has 900 million today.
The ads are currently clearly labeled, appear separately from responses, and are not supposed to influence what ChatGPT says. Early advertisers are getting aggregate performance data: views and clicks. User conversation data is not shared with advertisers.
What this means for service business owners:
Two things.
First, if you have a marketing budget and you’re not thinking about ChatGPT ads yet, this month is the moment it becomes accessible to you. Early positioning on new platforms almost always rewards the people who move first, before costs go up and competition increases. The same dynamic played out in the early days of Google Ads and Facebook advertising. You don’t need a massive budget to test and learn right now.
Second, and more importantly: your potential customers are increasingly using ChatGPT to make decisions. They’re asking it to recommend service providers, compare options, and help them figure out who to hire. Your visibility in that environment, whether through ads or through how your business shows up in AI-generated recommendations, matters more than it did a year ago.
3. Most Businesses Are Running AI Pilots That Go Nowhere, and the Gap Is Widening
PwC released its 2026 AI Performance Study this week, based on interviews with over 1,200 senior executives across 25 industries. The headline finding is blunt:
74% of AI’s economic value is being captured by just 20% of organizations.
The majority of companies are stuck. They’re running pilots. They’re testing tools. They’re talking about AI at the leadership level. And they’re not seeing meaningful financial results.
The companies pulling ahead are doing something different. They’re not deploying more tools. They’re using AI to pursue entirely new growth opportunities and rebuilding their workflows around AI rather than layering AI tools onto existing processes.
PwC found that the leading companies are twice as likely to redesign workflows around AI. They’re nearly twice as likely to let AI operate with real autonomy inside defined guardrails. And they’re 2.8 times more likely to have increased the number of decisions made without human review for routine work.
The laggards, by contrast, are using AI for isolated, incremental tasks. They’re getting marginal efficiency gains that feel like progress but don’t add up to transformation.
What this means for service business owners:
This is the finding I think about most as someone who builds AI systems for service businesses at BAMPT.
The question I hear most often is: “Which AI tool should I be using?” That’s the wrong question. The right question is: “Which process in my business should I rebuild?”
Adding an AI writing tool to the way you already produce content is not the same thing as rebuilding your content workflow around AI. Using a chatbot to answer client questions is not the same as redesigning your client onboarding process with AI at the center. The first approach gives you marginal time savings. The second approach changes what’s possible for your business.
The PwC data backs this up at scale. And the gap between companies doing the first thing and companies doing the second thing is getting wider, not narrower.
What This Means: The Infrastructure Is Being Laid Right Now
Visa is building the payment rails. OpenAI is building the advertising and discovery layer. And the companies that will benefit most are the ones redesigning how they work, not just adding tools.
Here’s the pattern I’m watching: the foundational infrastructure of AI commerce is being built in real time, and it’s being built by the largest companies in the world. By the time most businesses are aware of it, the standards will be set and the early movers will have the advantage.
This doesn’t mean you need to panic or overhaul everything at once. But it does mean the decisions you make in the next 12 to 18 months about how your business uses AI, not whether but how, will have a compounding effect on your competitive position.
What Business Owners Should Actually Do
Audit your discoverability. Not for Google, for AI. Does your business have clear, structured information about what you offer, who you serve, and how to work with you? That’s what AI agents will pull from when your potential customers ask them for recommendations.
Get ready to experiment with ChatGPT ads. Self-serve access is expected to open this month. When it does, the costs will still be relatively low, the audience is massive, and early movers on new ad platforms historically see better returns before competition drives prices up. Start thinking now about what you’d test so you’re ready to move when access opens.
Ask the process question, not the tool question. Pick one high-friction process in your business: client intake, proposal creation, scheduling, follow-up, reporting, and ask what it would look like if you rebuilt it with AI at the center. Not bolted on. Built around.
Stop counting pilots as progress. If you’ve been testing AI tools for more than six months without seeing measurable impact on your revenue or time, that’s a signal you’re in the wrong frame. The goal isn’t to be experimenting with AI. The goal is to have it running inside your business.
The window to move from early adopter to leader is still open. But PwC’s data makes clear it won’t be open indefinitely. The companies pulling ahead are compounding their advantage every quarter.
The good news: for a service business, you don’t need enterprise-scale infrastructure to get there. You need the right processes and the willingness to actually change how you work.
That’s the work we do every day at BAMPT. If you want to go deeper on any of these stories or talk about what they mean for your specific business, reply to this email or find me on LinkedIn.
Chantal Emmanuel is the founder of BAMPT, where she helps service businesses implement AI-powered operations using the A.G.E.N.T. Framework. She’s also CTO of LimeLoop and writes about automation, systems thinking, and building businesses that scale.