• Shopware Gold Partner & Shopify Partner

Chapter 14

Roadmap for the next 24 months

Table of contents

At first glance, market forecasts for agentic commerce look contradictory. The difference lies not only in the forecast, but already in the definition of what is being counted.

Why do the market forecasts contradict each other?

  • McKinsey (October 2025): 3 to 5 trillion US dollars worldwide by 2030 – measured as goods revenue "orchestrated" by AI agents, of which up to 1 trillion in US consumer retail.

  • Morgan Stanley: 190 to 385 billion US dollars – 10 to 20 percent of US e-commerce.

  • Bain & Company: 300 to 500 billion US dollars – 15 to 25 percent of US e-commerce, counted as purchases that agents trigger, influence or complete.

What exactly counts as agentic commerce, which market is considered, and how far does an agent have to be involved in the purchase? Choose the narrowest definition and you only count purchases completed entirely within an AI interface. Choose the broadest and you count every sale an AI system influenced anywhere along the chain.

For a roadmap, this means: the figures are good for determining direction, not for planning. That is why this chapter does not sort the work by dates but by certainty:

  • Non-negotiable foundation – pays off in every scenario: accessibility for crawlers, complete product identifiers, a clean data feed, correct schema, a checkout without unnecessary friction, a reliable measurement basis.

  • Sensible medium-term investment – retains its value even if platforms shift: reworking product and guide texts, active review management, genuine presence on the trust hubs, regular updating of existing content.

  • Speculative bet – open and therefore last: which protocol prevails, whether several coexist, how quickly consumers entrust a complete transaction to an agent. Work on a test system first here.

What is mandatory today and not a bet?

The basics from this guide are not a bet on the future, but the ongoing present: accessibility for crawlers, complete product identifiers, a clean data feed, correct schema, a checkout without unnecessary friction, a reliable measurement basis.

This work pays off regardless of which forecast turns out to be right, which protocol prevails and how fast adoption proceeds. If you have not yet done these basics, you do not need a 24-month roadmap. You need a priority list – the one from Chapter 11.

What is a sensible bet?

In the medium term, it pays to invest in areas that retain their value even if individual platforms develop differently than expected: reworking product and guide texts according to the principle from Chapter 7, active review management, genuine presence on the trust hubs from Chapter 4, and regular updating of existing content.

These measures share one property: they improve classic SEO, AI visibility and the human purchase decision at the same time. They are broad enough to serve several possible futures.

What remains speculation?

Honesty belongs in the guide here. Which protocol prevails, whether several coexist, how quickly consumers entrust a complete transaction to an agent – that is open.

One finding from market research is relevant here: according to Bain, a considerable share of US consumers already use AI for product research and comparison. Far fewer are prepared to hand over a transaction completely.

Simple, specification-driven purchases are particularly suitable for early automation. But that does not mean product ranges that require a lot of advice will be spared. On the contrary: that is exactly where AI can take over a large part of the previous research and comparison work. The difference lies less in whether AI becomes relevant than in which task it takes on in the purchase process.

Which attitude runs through this guide?

Looking back at the fourteen chapters, you can see a pattern. Every time a topic was heavily promoted – llms.txt, a single protocol integration, perfecting values beyond what is necessary – restraint paid off. And every time a topic was unspectacular but fundamental – complete product data, honest return information, a working mobile checkout – early investment was right.

This roadmap carries exactly this pattern into the future: robust, platform-independent foundations first, speculative individual bets last.

Our assessment

The most common mistake with a roadmap in this field is to set it up once and then not touch it for a year. Too much changes too fast for that – this guide itself had to be corrected several times while we were working on it because facts had shifted.

The second most common mistake is the opposite: overturning priorities with every new announcement. The solution lies in between: a fixed date, for example quarterly, on which the assessment is reviewed. Not more often, but not less often either.

AI readiness does not mean predicting the future correctly. It means setting up the store so that it is ready for several possible futures.

The next step: what you can do today

  1. Assign your list from Chapter 11 to the three categories of this chapter: non-negotiable foundation, sensible medium-term investment, speculative bet.

  2. For each point in the third category, check how much effort has already been invested – and how much of it would be lost if the platform changed course.

  3. Set the date for the next review now, not when the time comes.

Frequently asked questions

The forecasts differ widely. McKinsey names 3 to 5 trillion US dollars worldwide by 2030, Morgan Stanley and Bain 190 to 500 billion US dollars for US e-commerce. The difference lies mainly in the definition of what is counted.

About quarterly, on a fixed date. Less often and the assessment becomes outdated; more often and you end up overturning priorities with every new announcement.