The next wave of AI isn’t just changing how your team works. It’s changing how your customers shop — and most businesses won’t see it coming until it’s too late.
We’re entering the era of agentic commerce: AI agents that shop on behalf of humans. Not search engines. Not recommendation engines. Autonomous AI agents that compare, evaluate, negotiate, and purchase — all without a human clicking through your website.
The question every business owner needs to ask is no longer “how do we rank on Google?” The new question: will an AI agent find us, trust our data, check our stock, and complete an order — all automatically?
If your answer involves spreadsheets, manual data entry, or a catalogue that hasn’t been updated since last quarter, you’re about to become invisible.
What Agentic Commerce Actually Means
Agentic AI goes beyond chatbots and copilots. Specifically, where a copilot assists a human making a decision, an AI agent acts independently to complete a goal. You’ve already seen early versions: AI that books your restaurant reservation, fills in your expense report, schedules your next meeting.
Now extend that into commerce.
A procurement manager at a manufacturing company says to their AI agent: “Reorder our standard 3mm tempered glass panels, check three suppliers for current pricing and lead times, and place the order with whoever has the best combination of price and availability.”
The Agent in Action
The agent goes to work. Rather than browsing websites the way a human does, it queries APIs and reads structured product feeds directly. The agent evaluates schemas, checks real-time inventory, and processes pricing rules — then either completes the order or comes back with a shortlist of options for final approval.
This isn’t science fiction. OpenAI’s GPT-4 with function calling, Anthropic’s Claude with tool use, Google’s Gemini with agent frameworks — technology vendors are deploying the infrastructure for agentic commerce right now. Gartner predicts that by 2028, AI agents will make 15% of day-to-day business decisions autonomously. In B2B commerce, where procurement is already structured and rule-driven, that number will be far higher, far sooner.
The Two Types of Businesses That Will Emerge
Here’s what becomes clear when you think about agentic commerce: not all businesses are equally findable by AI agents.
AI agents don’t browse the way humans do. They don’t respond to beautiful web design, persuasive copy, or a well-crafted hero image. Instead, they query structured data and read APIs — processing machine-readable product information: SKUs, specifications, pricing, inventory counts, lead times, weight, dimensions, categorisation.
This creates a binary divide:
Agent-ready businesses have their product data structured, their systems integrated, and their APIs open and responsive. When an AI agent comes looking, it finds exactly what it needs, in a format it can process, and moves on to completing the transaction. Millions of AI-assisted procurement decisions will discover these businesses automatically — transactions they never even know are happening.
The Agent-Invisible Reality
Agent-invisible businesses have their product data locked in spreadsheets. Their inventory lives in one system, their pricing in another, their order management in a third — and none of them talk to each other in real time. When an AI agent queries them, it gets either nothing or unreliable information. The agent moves on. These businesses lose transactions they never knew were available.
I’ve been in systems integration for over 25 years. In that time, I’ve watched businesses lose competitive advantage to better-integrated competitors again and again — in ways they didn’t fully understand until the revenue was already gone. Agentic commerce is about to do the same thing, at a scale and speed we haven’t seen before.
Why Your Integration Architecture Is Now a Commercial Asset
Most business owners think of their systems integration as back-office infrastructure. It’s the plumbing. Necessary, but not strategic.
However, agentic commerce reframes that entirely.
Your integration architecture is now a commercial asset — one that either opens you up to a new wave of AI-driven buyers or closes you off from it completely.
Here’s what “agent-ready” actually requires:
The Four Requirements for Agent-Readiness
1. Real-time data, not batch exports — If your inventory updates once a night via CSV, an AI agent querying your stock levels at 2pm is working from stale data. It might complete an order for items you don’t have. Or — more likely — it simply won’t trust your data and will move to a competitor who can provide a reliable real-time feed.
2. Structured, consistent product data — AI agents are pattern matchers. They rely on consistent schema: standard product identifiers (GTIN, SKU), consistent attribute naming, complete specifications. A product catalogue where “colour” is sometimes spelled differently, or where attributes are missing entirely, is not parseable by an agent. Speed and reliability are non-negotiable in agentic commerce.
Technical Integration Requirements
3. API-first architecture — An AI agent needs a machine-readable interface to your catalogue, inventory, pricing, and ordering capability. This means either a proper API or a standardised data feed (XML, JSON) that exposes your product and availability data in real time. A website, no matter how well designed, is not an API.
4. Bidirectional integration — Agent-ready isn’t just about being findable — it’s about being actionable. The agent needs to be able to place an order, trigger fulfilment, and receive an order confirmation, all programmatically. That requires your eCommerce website, your ERP or inventory system, and your order management process to be connected and able to respond to machine-initiated transactions.
How CODI Delivers Agent-Readiness
This is exactly what CODI — our Convergence Optimised Data Integration platform — was designed to do. For more than a decade, CODI has been connecting eCommerce websites with backend ERP, CRM, accounting, and inventory management systems for growing businesses. Hub-and-spoke architecture, real-time bidirectional sync, translation and validation between incompatible systems.
Businesses running CODI are, structurally, already closer to agent-ready than most. The integration foundation is there. The question is whether the final layer — structured, queryable, API-accessible product data — sits on top of that foundation.
The Link to Private AI: Two Sides of the Same Coin
There’s a dimension to agentic commerce that most commentary misses, and it’s the one I’m most excited about.
So far, we’ve talked about outbound agent-readiness: making sure AI agents from outside your business can find and buy from you. However, there’s an equally powerful opportunity in the other direction.
What if your business ran its own AI agents, querying your own integrated data?
This is exactly what we’re doing with our Integration for Private AI offering. And it turns out that the same integration work that makes you agent-ready for external commerce also makes you capable of running powerful internal AI agents.
How Integration Unlocks Internal AI Intelligence
Here’s why: AI agents are only as good as the data they can access. An internal AI agent — one that you deploy on your own dedicated infrastructure, with no data leaving your building — can only answer useful questions if it has clean, structured, integrated access to your business data.
The same real-time bidirectional sync that powers your external agent-readiness also drives your internal AI capability. Your internal AI agent can answer questions like:
- “What are our five best-margin product lines this quarter, and do we have sufficient stock to run a campaign on them?”
- “Which of our B2B accounts have reduced their order frequency in the last 60 days?”
- “What’s our landed cost for this supplier’s products given current freight rates, and what margin does that leave us at our current retail price?”
These are questions that today require three separate reports from three separate systems, a spreadsheet, and an analyst. With proper integration, a private AI model on your own infrastructure — tools like Ollama running Meta’s Llama or Mistral — turns these into natural language queries. Anyone in your business can run them in seconds, without IT involvement.
The integration work is the same work. Connecting your ERP to your eCommerce website to your inventory and your CRM — that’s what CODI does. Once that integration exists, you can point both an outbound agent-readiness layer and an inbound private AI intelligence layer at the same connected data. One investment, two transformational capabilities.
The Privacy Dimension of Agentic Commerce
Here’s something worth flagging as AI agents become more capable: the data flows are going to get murky very quickly.
If a customer’s AI agent is querying your product catalogue and pricing, you’re sharing commercial data with a third-party AI system. If a major platform runs that agent — think Amazon, Shopify, or a large procurement platform — it may process, store, and learn from the data you share. You cannot control how.
For businesses where pricing strategy, supplier relationships, and product roadmaps are competitively sensitive, this matters.
The NZ Privacy Act 2020 is already creating obligations around how data is shared with third-party systems. This is reinforced by Information Privacy Principle 3A (IPP 3A), which came into force 1 May 2026. As agentic commerce scales, the intersection of commerce automation and privacy compliance is going to require careful architecture.
Structuring Your Response
The answer isn’t to opt out of agentic commerce — that’s not commercially viable. Instead, control the layer at which data is shared: expose structured product and availability data to external agents via a purpose-built API.
Meanwhile, keep your commercial intelligence — your margins, your customer data, your supplier terms — on your own private AI infrastructure where no public AI models are in play.
External agent-readiness through properly structured and exposed data. Internal intelligence through private AI that never leaves your dedicated infrastructure. Clean separation, clear control.
What to Do Now
Agentic commerce is not a 2030 problem. In fact, technology vendors are deploying the infrastructure right now.
Early movers will build agent-readable systems and capture AI-driven traffic before their competitors even understand what’s happening.
As a result, here’s a practical starting point:
- Audit your data quality. Can you produce a complete, structured, machine-readable product catalogue right now — with accurate inventory, real-time pricing, and consistent attribute naming across every SKU? If not, that’s where to start.
- Review your integration architecture. Are your ERP, inventory, and eCommerce systems connected in real time? Or are you running on batch exports and manual intervention? Every manual step in your data flow is a point of failure for agent-readiness.
- Think about your API layer. Do you have a mechanism for a machine — not a human — to query your product data and place an order? If not, you’re relying on human-friendly interfaces that AI agents can’t use reliably.
- Consider your private AI opportunity alongside this. The same integration work that makes you agent-ready externally can fuel private AI intelligence internally. It’s not two separate projects — it’s one foundation with two valuable applications.
If you’d like to understand what this looks like for your specific systems and applications, that’s exactly the kind of assessment we do — no obligation, even if that is not to proceed with us. We’ve been connecting eCommerce and backend systems since 2009. The shift to agent-readiness is the most significant evolution we’ve seen in that time, and we’re ready to help growing businesses navigate it.
The Bottom Line
The AI shopping agent doesn’t care about your brand story, your checkout experience, or your loyalty programme. It cares about one thing: can it reliably get the structured data it needs to complete a transaction on behalf of its user?
Retailers and B2B suppliers whose backend systems are properly integrated will be agent-ready — AI systems will find, evaluate, and order from them, without those sales teams ever knowing.
Those without clean data pipelines, real-time integration, and machine-readable product data will be invisible to AI shoppers — and invisible in agentic commerce means irrelevant in commerce, full stop.
Therefore, if you’ve been thinking about getting your systems properly integrated, you’ve never had a better reason.
About the Author
Mark Presnell is Managing Director of Convergence Ltd, systems integration specialists since 1997. Convergence builds and manages integrations between eCommerce websites and backend business software through CODI — Convergence Optimised Data Integration — and offers Integration for Private AI for businesses who want to run their own private AI models on their own dedicated infrastructure. Book a free requirements assessment at convergence.co.nz.