Agentic Commerce: Why the next 18 months will decide the next ten years

Agentic AI in retail: What has slowed down many retailers so far

The conversations I have been having with retail decision-makers for months are similar: they are about agentic commerce, AI agents and product data strategies. The topics are understood. So is the urgency. And yet too little is happening.

This is not due to a lack of interest. It is due to a structural weakness that I observe time and again in many companies: the inability to make a clear strategic decision when the answer is not yet clearly on the table.

What is currently paralyzing many companies is not a lack of knowledge. It is a lack of clarity about responsibilities.

Instead of taking a strategic position, many small use cases are being tried out. A pilot here, a vendor discussion there, a workshop in the next quarter. Each of these steps makes sense on its own. Taken together, however, the result is not a strategy, but activity. And activity without direction is a waste of budget and time.

A global look at Agentic Commerce: What happens when we weigh things up

McKinsey estimates that AI agents will influence 3 to 5 trillion dollars in retail sales worldwide by 2030. The shift is not a distant forecast – it is already underway. It’s worth stepping back from the German perspective for a moment. Four markets (and one special case) illustrate how agentic commerce is gaining momentum and speed globally:

What I see with retailers from DACH

What the markets described have in common: They have stopped asking whether they should trade. They only ask how.

In conversations with decision-makers from German-speaking retail, on the other hand, I see four patterns again and again:

1. Lots of small use cases instead of one strategy: the assumption behind this is that you will find the right model through experimentation. This logic works for tactical issues. It does not work for structural shifts. If you want to reorganize the interface between brand and customer, you need a strategic hypothesis – not twenty pilots pointing in different directions.
2. A blind spot in their own position in the LLM landscape: Most companies cannot answer how often their products appear in ChatGPT, Gemini or Perplexity in relevant search queries. They know their SEO position. Not their AI visibility. What is not measured is not controlled.
3. Assumptions about customer needs that are no longer true: The idea that German customers are more conservative than American customers and will therefore stick to traditional purchase paths for longer is an illusion. 35% of US shoppers now use AI assistants to make purchases – a year ago it was 12%. 51% are prepared to leave the entire purchasing process to AI. These figures also apply in Germany. The only question is whether in 12 or 24 months’ time.
4. The data infrastructure is being tackled too late and too timidly: Data quality is treated as an IT hygiene issue, not as a strategic foundation. As long as product data is not machine-readable, complete and API-accessible, all other discussions about AI agents are theory. The order is clear: first the database, then the platform integration, then your own agent. Anyone who reverses this will burn budget and still lose visibility.

Personal assessment: How quickly the market is shifting

We at valantic assume that the share of agent-driven purchases in the German grocery and CPG market will develop in waves over the next ten years. The following estimate is based on historical adoption curves (e-commerce, mobile payment, voice commerce) and the current speed of protocol development.The figures should be seen as a guide, not as a validated forecast. The decisive factor is the direction:

Even in the most conservative scenario, a significant proportion of standard purchasing behavior will shift to agent-driven channels over the next few years. Those who make the infrastructure decisions now can help shape this shift. Those who wait will see it from the outside at some point.

What it takes to be able to act for Agentic Commerce

The most important point is not technical. In most German companies, nobody knows exactly who is responsible for agentic commerce. The topic lies somewhere between marketing, e-commerce, IT and product management – and therefore nowhere.

As long as Agentic Commerce hovers between the departments, no decisions are made. Only workshops are created.

What is needed is a clear assignment to a person high enough up to mobilize resources. With decision-making scope, budget and explicit permission to explore. Without this structure, no strategy is created.

The companies that are setting the pace today do not have the best understanding of tech. They have the clearest organization.

Why the next 18 months are crucial

The timing is not arbitrary. It results from three coinciding developments:

• The protocol infrastructure is live. ACP, UCP and MCP (Model Context Protocol) are no longer a roadmap, but productive. Anyone who has not yet been integrated in 18 months has a measurable reach backlog.
• Data quality cannot be achieved in six weeks. If you start today, in 12 to 18 months you will be at the same level as your international competitors. If you start in 12 months, you will have a two-year head start in 24 months.
• Asia-Pacific will generate around two thirds of global new retail sales in the next five years. International visibility without agentic commerce will become structurally more difficult from 2027.

Conclusion: What steps retail leaders need to take now

From my observations and conversations, I can conclude three steps that retail decision-makers should tackle and implement now:

1. Appoint a person responsible for Agentic Commerce: Not a working group, but a person high enough up; equipped with budget, decision-making authority and explicit permission to try things out.
2. Take stock: Where does the company stand in terms of product data maturity, platform integrations and checkout architecture? Not every step needs to be implemented immediately. But every step must be evaluated.
3. Take a position: There are several legitimate strategies – from not using third-party platforms at all to full integration. What doesn’t exist is a legitimate option to wait and see.

The next 18 months will not decide everything. But they will decide which starting position companies take into the next phase. If you move now, you build up options. Those who stand still lose them. It’s as simple – and as uncomfortable – as that.