This autumn Adobe put a number on something that had been anecdote until recently. Across the 2026 holiday season it expects traffic arriving at US retail sites from AI assistants to grow 130% year over year, after a 693.4% jump the season before. Visits that arrive from an AI source generate 83% higher revenue per visit than visits from anywhere else, and 77% of people who have used AI to shop told Adobe they felt more confident about what they bought.
Those are retail figures, and most people reading this do not work in retail. But they measure something that is not really about retail: a growing share of the arrivals at a digital product did not begin with a person typing a query. Something read the page first, formed a view, and passed it on.
We did not ask anybody about that. Over the past year we put one open question to the people speaking at DDX — what will shape the next wave of digital innovation? Fifty-seven answered, from Google, Adobe, Microsoft, Amazon, Google DeepMind, Bloomberg, Target, Walmart Global Tech, PlayStation Studios, Universal, American Express, Lloyds Banking Group and the Obama White House, among others. No topic, no prompt, no list to choose from.
Eight of them, independently, described the same shift: the user is no longer only a human.
The buyer
The sharpest version came from someone who sells to enterprises for a living.
Your next best customer might be an AI agent buying on someone's behalf. That changes everything about how you design a digital experience, because now it has to work for the human and the machine simultaneously. The companies already designing for that are going to set the terms for everyone else.
Read that next to Adobe's numbers and it stops being a prediction. The agent is already in the funnel; what is not yet settled is whether it finds a product it can understand.
The reader
The second group is further upstream, and their concern is the page itself.
I’m particularly interested in how the role of the website will evolve, and how our disciplines will need to adapt. We are starting to think about designing webpages not just for users, but for LLMs, and keeping a pulse on how user behavior and attitudes are changing will continue to be integral to ensure we are building digital experiences which align with what matters to our users.
This is the quiet part of the shift. A machine reading your page cannot see your typography, your spacing, your hover states or the photograph that carries the argument. It reads text and structure. Everything you expressed through craft alone is, to that reader, absent. The question what does this page say when it is read as plain text has become a design question, and almost nobody has it on a checklist.
The same logic reaches inside the organisation. One of the most quotable lines we received this year is about a design system:
As AI makes digital creation dramatically faster, the role of designers won’t become less important; it will move upstream. Our opportunity is to design the systems, principles, and constraints that allow both humans and machines to create with intention, consistency, and purpose.
The best design system in the AI era isn’t one people remember to use. It’s one AI can’t forget to use.
A design system has always had a compliance problem: people forget it exists. A machine does not forget, but it also does not infer intent. That makes the system's hardest user a very fast reader with no judgement — which is an argument for writing it down far more precisely than a human-only team ever needed to.
The delegate
The third group is looking at what happens after the machine arrives.
We’re moving from designing interfaces to designing relationships. The next generation of digital products are intelligent systems people delegate to, collaborate with and trust to act on their behalf. That makes designing for trust, transparency and human agency one of the most interesting challenges ahead.
Brands and trust are built over years, often decades, through consistency, quality, and countless small interactions. As AI makes it possible to generate interfaces, content, and experiences faster than ever, that trust can also be put at risk faster than ever. The future of digital innovation will depend on how well we design the guardrails, systems, and judgment that keep generated experiences coherent, useful, and worthy of the brands behind them.
And a practical note from someone building these systems, which cuts against the instinct to chase tools:
People are chasing the next model or the next tool, but the winners of this era will be those who invest in curating and structuring their context and their knowledge, because tools change every week, but context never gets old.
What the room disagreed about
A set of answers that all point the same way is usually a set of answers that was steered. This one was not, and the proof is that the same cohort contains the strongest objection.
While AI can read a deck, a doc, a website, your emails…it can’t read the room. AI is a tool, just like the internet was a tool. It may augment us, but it can never replace us.
As AI becomes more capable and generates more machine output, human originality becomes a scarce resource. We need to protect and foster the conditions where humans can continue to create original ideas.
No matter how much technology evolves and products change with AI, you cannot take human judgment out of the process. While AI can generate hundreds of ideas, it still takes humans to decide what works and what doesn’t. The fundamentals and design principles we’ve applied over decades still matter, and will continue to matter, as long as we’re designing for humans.
The two camps are less opposed than they look. Nobody in the first group argued that the machine is the point. They argued that the machine is now a reader, a buyer and a colleague — and that a product which only works when a person is looking at it will quietly stop working. Nobody in the second group argued that the machine can be ignored. They argued that the decision about what is worth building stays human, and gets harder as production gets cheaper.
The bridge between them was put best by someone who has to make both cases inside a large retailer:
The next era of innovation will not belong to the companies that automate the fastest, but to those that intentionally design systems in which humans and intelligent technology can learn, adapt, and make better decisions together.
Three things this changes on Monday
Your page has an audience that cannot see it. Open your most important page, strip the CSS, and read what is left. If the argument disappears, so does it for every assistant that summarises you to a prospective customer.
Your design system needs to be legible to something with no judgement. Principles written to be interpreted by a thoughtful colleague will be followed literally. Say what you mean, including the constraints you have always left implicit.
Trust has moved from the interface to the instruction layer. When a system can act for your brand, the brand lives in what you told it to stand for, not in what the screen looks like. That is a strategy artefact, and almost nobody owns it yet.
How this piece was made
Every quote above was written by the person named, for DDX, in answer to a question with no topic attached. We publish them because a conference's most valuable asset is not its stage — it is what several hundred senior practitioners will tell you when you ask them the same question and do not steer the answer.
The next round of that question is being asked in London on 20 November, and in Tokyo in 2027.
Sources for the figures in the opening paragraph: Adobe Digital Insights, 2026 US holiday forecast (28 September 2026) and Adobe Analytics retail AI-traffic analysis (October 2026).
