It reflects a deeper change in how we think about software, about autonomy — and ultimately about User Experience.
Agentic AI plans, re
asons, and coordinates multiple actions to achieve a complex goal.
I think of an AI agent as a skilled worker a great co-worker. Agentic AI is more like a manager — one that decides what needs doing, delegates to the right tools, and adapts when circumstances change.
It is a simple definition. But simple definitions only tell us what something is. They say very little about why it matters.
For decades, we built software around tasks. A user wanted to accomplish something, the system responded. Book a flight - and I work for so many years on such services for Delta Air Lines and Lufthansa. Approve an invoice. Complete a checkout - might it be in real life or the digital world. The interaction always followed the same basic principle: a person initiates, the system executes.
Even many of today's AI agents still fit that model. They may work faster, understand natural language, or combine several capabilities — but in the end, they are responding to a request. Remarkably capable assistants. Still assistants.
Agentic AI feels different.
Instead of waiting for the next instruction, it starts thinking about objectives. It does not ask "What should I do next?" It considers "What is the best way to achieve this goal?" That subtle shift changes almost everything. Planning replaces execution. Reasoning replaces reaction. The system decides which tools to use, which information is still missing, whether another step is needed — or whether it should pause and ask the user first.
The interaction becomes less like using a tool and more like working with a colleague.
Consider a self-checkout experience. Traditionally, we designed a sequence: scan items, choose payment, confirm. Every step was a deliberate interaction — the user in control, the system responding.
An AI agent can improve that experience. It might detect a scanning error before the user notices, or suggest a faster payment method based on past behaviour. Still responsive. Still task-driven.
Agentic AI changes the dynamic entirely. The system might notice that an item was recalled last week, surface a loyalty discount the user did not know existed, and quietly resolve a pricing inconsistency — all before the user reaches the payment screen. It is not waiting to be asked. It is already working toward the outcome.
As a UX designer, the question is no longer:
How do I make each step clear and easy?
It becomes:
How much should the system decide on its own?
When should it explain itself?
And when does helpful initiative start to feel like loss of control?
For many years, UX designers have focused on interactions. We crafted screens, defined navigation paths, refined error messages, and tried to make every single step as clear as possible. We were, in hindsight, designing conversations between a human and a machine — one interaction at a time.
Agentic AI asks us to think differently.
Our task may or is no longer be primarily to design interactions. We may be starting to design behavior and really the 'feel' - as we often talked about 'look' and 'feel' — how much initiative a system may take, when it should explain itself, when it should stay silent, and when it should deliberately hand control back to the user.
I believe this is one of the biggest shifts our profession has faced since graphical interfaces replaced the command line. We are moving from designing what a system looks like to designing how it acts.
The real distinction between AI agents and Agentic AI is not about technical capability. It is about philosophy. One is centred on completing tasks. The other is centred on pursuing goals.
For engineers and developers, that is primarily an architectural question.
For UX designers, it is something more fundamental: a renegotiation of who is in control, and what it means for a person to trust a system.
We have spent decades making software easier to use.
Agentic AI asks us to make it worthy of being trusted.
That is a different problem and task — and I think it is the most interesting one our profession has faced in a long time.
My two cents 😉 as grumpy old designer
Comments
Post a Comment