Dotdigital Summit 2026: Why Smarter Marketing Still Needs a Human Hand

An AI agent can build your segments these days. It can draft the campaign copy too, and sketch out a customer journey before you’ve finished your coffee. Yet the message that ran right through Dotdigital Summit 2026 was not “hand it all to the machines”. It was closer to the opposite: the technology earns its place by buying people time to think, to judge and to properly understand their customers.

Old Billingsgate filled up on 2 July 2026. More than 800 marketers turned out for the annual summit, which spread itself across five tracks and took in behavioural science, automation, data, personalisation, loyalty and a fair amount of leadership talk besides. Sit in any room and the same thought surfaced. Efficiency is fine. Efficient but forgettable is a waste of the budget.

Marketing made human, in practice

The rebrand had landed only the day before, with Dotdigital now trading under the line Marketing made human. An easy target, on the face of it. Here was a humanity-first slogan fronting a programme heavy with AI agents, natural-language segmentation and automation at every turn. Spend the day in the sessions, though, and the slogan looked less like spin and more like a working brief.

CMO Juliette Aiken put a number on the tension when she announced the rebrand. Ninety-nine per cent of marketers are using AI in some form, she said. Only 36% enjoy it. Her reading of that gap was blunt: the future is human plus AI rather than human versus AI, with the technology soaking up the repetitive work so that people can get back to the customers.

Katie King, CEO of the consultancy AI in Business, explored what that means for how teams are run when they comprise both people and AI systems. Her argument was less about making machines imitate human capability and more about knowing where each belongs. AI can process data, identify patterns and handle repetitive production tasks at speed. It remains far less dependable when a campaign requires cultural sensitivity, original thinking, empathy or an informed challenge to the brief.

Dotdigital’s official summit recap accordingly dwells as much on authenticity, trust and leadership as it does on software, and that balance matters. Wire AI into a broken process and the process stays broken; it just breaks quicker. What pays is rethinking the work first. Give the machine the drudgery, and keep hold of any decision where context matters and somebody has to carry the can.

Segmentation is becoming conversational

Some of the most practical announcements concerned the work that happens before a campaign reaches the creative stage. Dotdigital’s Segment Agent lets a marketer describe an audience in everyday language rather than constructing a long series of database rules by hand. A request might combine purchase history, location, interests and churn risk, and the agent turns those requirements into a usable segment.

The efficiency case is obvious, particularly for smaller teams without a dedicated data specialist. Conversational tools may also open sophisticated targeting to people who understand their customers deeply but are less at home in complex segmentation interfaces.

What they do not remove is the need for sound data. If customer records are incomplete, duplicated or badly structured, an elegantly phrased request will still produce a questionable audience. We would put it more strongly: the value of the whole new product suite rests on clean contact records, behavioural data worth reading, and segments that were designed rather than merely accumulated over the years.

There is a governance question too. Alongside the Segment Agent sits a new Content Agent, which brainstorms, drafts, edits and styles copy inside the email editor, and can generate images to go with it. Both agents are in private preview with a customer waitlist, which was still the position in early August 2026, and Dotdigital’s July product update is explicit that every suggestion requires human approval before anything goes live; in the company’s own words, nothing goes out without a human touch. That is a sensible distinction. Assistance can make campaign development quicker without allowing unexamined content or targeting decisions out of the door.

Personalisation must be relevant, not merely possible

Most platforms will happily drop a customer’s first name into an email these days. That does not make the message personal. The customer stories at the summit pointed towards a more useful definition of personalisation: responding to real signals about what somebody needs, values or is likely to do next.

Goodwood is a useful example. In a case study published by Dotdigital, the estate describes using generative AI to write and test SMS variations for different segments ahead of an August Bank Holiday race meeting. Ticket sales doubled on the day the message went out compared with the week before, and a follow-up feedback request managed a 25% completion rate against 10% the previous year. None of that came from sending more. It came from sending better.

The principle matters more than the particular software. Good personalisation reduces irrelevance. It helps a customer find useful information, pick up where they left off or receive an offer at a sensible moment. If it merely allows a business to send more messages, more quickly, the customer gains very little.

Dotdigital’s July update extends this approach into loyalty. Customers can earn points for purchases, reviews, birthdays and anniversaries, social follows and SMS sign-ups, progress through up to four tiers automatically, and redeem rewards ranging from discounts and gifts to free delivery. More usefully still, those points, tiers, purchase histories, engagement signals and marketing preferences all live in one place, so retention work no longer has to lean on transaction value and nothing else. The chance here is to stop running loyalty, email, website behaviour and customer service as separate campaigns out of separate systems. To the customer they were always one relationship.

Distinctiveness still comes before optimisation

The opening keynote from Phill Agnew, host of the Nudge podcast, brought behavioural science into a day otherwise dominated by data and technology. His examples included the Von Restorff effect, our tendency to remember whatever stands out from its surroundings, and the IKEA effect, the extra value people attach to products they have helped create. Dyson’s transparent bin provided a familiar illustration of visible effort: rather than merely claiming that the vacuum works, the design lets customers watch it working.

The broader lesson is blunt. People notice difference, evidence and participation. Yet many organisations still build campaigns by studying competitors and reproducing the same formats, claims and visual conventions. Optimisation cannot solve that problem. A perfectly segmented campaign may reach exactly the right audience and still be ignored because the creative idea looks like everything surrounding it.

Before asking how to produce more content, marketers should ask whether the content gives anyone a reason to remember the brand. That might come from a distinctive point of view, visible expertise, a recognisable voice or an experience in which the customer plays an active part.

Trust cannot be automated

Trust ran through the summit in several forms: trust between brands and customers, trust within teams and confidence in making decisions without complete certainty.

Stephen Libby, who alongside fellow Traitor Rachel Duffy became a joint winner of The Traitors UK in 2026, spoke about consistency, vulnerability and taking opportunities. There was a certain irony in a winning Traitor (a cyber security consultant before the show) delivering one of the day’s clearest messages on authenticity: audiences notice quickly when a person, or a brand, moves away from previously stated principles simply because a commercial opportunity has appeared.

That warning lands harder as generative tools make it easy to produce endless variations of superficially polished content. Volume can increase while conviction disappears. A brand voice stitched together from generic prompts will read well enough. What it won’t do is tell anyone what the business stands for, or give a customer one good reason to choose it over the next tab along.

Michelle Corp, Ecommerce Director at Lily’s Kitchen, turned the same question inwards with a session on resilience and on leading a team through uncertainty. The closing slot belonged to Dr Jo Salter MBE, Britain’s first female fast-jet pilot, on challenging the status quo and holding your nerve under pressure.

None of this sat apart from the technology conversation. Rolling out AI well is a management problem as much as a technical one. Teams have to be free to experiment, and to look hard at what came of it afterwards. Above all, it has to be safe for somebody to say out loud that the automated recommendation was wrong.

What should UK marketing teams do next?

The summit offered plenty of new technology to consider, but the most useful response is not to adopt everything at once. Four steps, in order:

1. Sort the data first. Check whether consent records, contact details, behavioural information and channel preferences are reliable enough to support meaningful targeting. Improving that foundation will usually create more value than adding another disconnected tool.
2. Pick one process, not ten. Choose something repetitive but commercially relevant: building campaign segments, producing initial content variations or setting up a post-purchase programme. Define what the technology may do, where human approval is required and how success will be measured.
3. Measure what matters. A tally of campaigns sent, or of hours saved, proves nothing by itself. Watch conversion, repeat purchase, retention and unsubscribes, and ask whether the enquiries coming in are ones the sales team would thank you for.
4. Guard the brand’s voice. Write down the point of view, the vocabulary, the creative standards. AI is good at applying rules like those once they exist. Inventing them was never its job.

The next audience is not even human

One more shift is worth planning for. A growing share of product discovery now happens inside AI assistants and answer engines rather than on a traditional results page, and those systems reward much the same things the summit kept returning to: clean structured data, consistent product information and content that answers real questions plainly. The brands that get quoted and recommended by AI tools are the ones whose sites are easy for machines to read and worth citing. That makes it a search job as much as a content one, and it is where everything above starts to overlap. The same tidy data that feeds a Segment Agent also feeds the answer engines your customers are learning to trust.

The real lesson from Dotdigital Summit 2026

Dotdigital Summit 2026 caught a marketing industry moving beyond the early excitement, and the early anxiety, surrounding generative AI. The mature question is no longer whether marketers should use AI, but where automation improves the customer experience, which decisions still require human responsibility and whether the organisation’s data is strong enough to support either.

The businesses that benefit most will not necessarily be those producing the greatest volume of automated content. They will be the ones using technology to understand customers more clearly, remove unnecessary work and make better creative decisions. Marketing automation earns its place when it makes people feel better understood, not simply contacted more often.

If you are weighing up where AI fits within your email, CRM or paid media programme, talk to AIWIZ and we will help you find the answer.