← Digital Impacts

Agentic AI: From Showing to Doing

The user expects the system to act, not just to show. Delegation, trust calibration, and making a traditional product usable by an agent.

This section does not describe the current state of generalist use, but the direction of expectations: where the phenomenon is going and why it is worth addressing now.

2026 captures the current phase: infrastructure accelerates, trust lags behind. Google launched background search agents and agentic checkout inside AI Mode, and presented at the NRF in January 2026, together with Shopify, an open standard for agent commerce — Universal Commerce Protocol, with Etsy, Wayfair, Target, Walmart among the co-developers and over twenty supporting entities including Visa, Mastercard, Stripe, and American Express; in May it was extended with universal cart, deferred payment, and new markets. On the opposite front, OpenAI scaled back native checkout in ChatGPT after five-six months (see Cross-cutting themes). The direction is traced; adoption will depend more on the design of trust than on technology — and, in Europe, on the regulatory framework (see The regulatory perimeter).

The User Expects the System to Act, Not Just to Show

The user describes the desired result and delegates the orchestration: they define the output, they do not manipulate elements one by one.

Making a Traditional Product Usable by an Agent

Building an agent is a project; making your service agent-ready is maintenance work that concerns the products we already have in production.