Getting a convincing answer out of AI can take seconds, but deciding whether to build a campaign, change a message or advise your CEO on the strength of it takes more judgment.
At Meltwater Summit on Tour London on October 1, 2026, speakers explored how PR and marketing teams can make AI useful in practice, from improving access to insights to understanding its growing role in brand discovery. Along the way, Professor Hannah Fry introduced some surprisingly talented pigeons. Here are six lessons from the day, and how to apply them.
Contents
Keep human judgment at the center of AI use
Start with the business question
Ground AI answers in reliable data
Make AI visibility part of reputation management
Create content for people and AI discovery
Use the time AI saves for strategic work
Put the lessons into practice
FAQs about AI in PR and marketing
Keep human judgment at the center of AI use
Professor Hannah Fry’s keynote gave the audience an unusual way to think about AI: pigeons trained to identify cancer in medical images. Their ability to recognize patterns was impressive, and it also illustrated how performing a specific task well differs from understanding the wider responsibilities of a profession.
She explored the problems that arise when machines interpret our instructions differently from how we intended them, and when people put too much faith in a machine’s output. For PR and marketing professionals, those problems can show up in a research summary, a campaign recommendation or a piece of copy that sounds plausible until someone checks it.
The live Data Malarkey podcast brought that lesson into the workplace, with host Sam Knowles, Canon EMEA’s Gareth Crew and Golin Ketchum’s John Brasington discussing the risk of handing over tasks without staying mentally involved in the work. Brasington described finding AI particularly useful in areas where he already had the expertise to evaluate its answers.
That distinction helps teams decide how to use AI responsibly, because a quick read for typos won’t reveal a flawed methodology or a recommendation built on the wrong assumptions. The person reviewing an AI output needs enough expertise to challenge it.
Build review into the workflow so people can check the evidence, assess the context and challenge the recommendation.
Start with the business question
Heineken’s Narek Garit offered a practical starting point for AI adoption: work backward from what a stakeholder needs to achieve.
It’s easy to start generating reports once you have data, dashboards and an AI assistant, but an impressive report can still spend its life unopened.
Garit recommended defining the use case first, then selecting the data, analysis and delivery method that serve it. He also stressed connecting social intelligence with relevant media, research and internal information, so teams can interpret what they find in context.
For a communications team, the question might concern whether an emerging narrative requires a response, while a marketing team might need to understand which audience concerns a campaign should address. Those are different decisions, even when they draw on some of the same conversations.
Define the decision before building the workflow. Agree on who needs the answer, what they will do with it and how quickly they need it, then determine where AI can help with gathering evidence, analyzing information or presenting the findings.
Garit’s advice was to start small and keep future scale in mind - one useful application gives you something concrete to improve and a reason for colleagues to support the next one.
Ground AI answers in reliable data
Meltwater’s Antony Cousins described a problem he called “shadow insights”: business users asking general AI assistants for answers without the data or methodological checks those questions require.
The convenience is understandable, because an assistant responds immediately, whereas an insights specialist might first ask which market, which audience and which time period the question concerns. Those clarifications improve the research, but a colleague facing a deadline may see them as another hurdle.
Cousins shared an example involving a question about women’s attitudes toward electric vehicles. The apparently useful response relied on a single, outdated blog post, but the answer looked much more comprehensive than the evidence behind it.
His session explored how Meltwater’s MCP connector can bring relevant data and existing searches into AI workflows. He also discussed capturing expert methods in reusable skills, so colleagues can draw on a consistent approach to recurring questions.
Make reliable research easier to access by agreeing on sources and methods for common requests, and include the limits of the evidence in the answer itself.
That last step is particularly useful, because social conversation can reveal themes worth investigating, while a representative survey answers a different kind of question.
Make AI visibility part of reputation management
Gener8 founder Sam Jones showed why AI deserves attention across the customer journey, and described using it to plan a family holiday and select a stroller, then buying the first stroller it recommended.
His presentation drew on an analysis of more than 100 million AI conversations from over 200,000 people across the US and UK. In the situations Gener8 compared, people were 3.6 times as likely to visit a brand’s website when it appeared in a commercially relevant AI response, compared with an answer that featured a competitor and omitted the brand.
For marketers, that makes AI recommendations a meaningful area to investigate. For PR professionals, the information inside those recommendations raises an equally important question: is the brand being represented accurately?
In the panel hosted by Meltwater’s Jenny Force, Snap Inc.’s Alicia Edmonds-Smith emphasized a strong, consistent narrative and factual accuracy, while Unilever’s Laura Thomas discussed the product attributes a brand needs to be known for, including the complications of managing different names and associations across markets.
The Data Malarkey conversation extended AI visibility to other stakeholders, explaining that employees, investors and others may also use AI to research an organization, so its answers can shape perceptions well beyond a product shortlist.
Start by reviewing questions relevant to your audiences, then examine where your brand appears, how it is described and which sources the answers cite. Include the markets and languages that are relevant to your business.
Accuracy deserves a place alongside visibility in your reporting. An increase in mentions is hard to celebrate if those mentions misstate your product, repeat an outdated claim or misrepresent your organization.
Create content for people and AI discovery
AI discovery gives content teams another reason to answer customer questions clearly.
Jenny Force shared Meltwater’s experience of rewriting key website pages in plain language around questions customers were already asking, and reported a 73% increase in AI citations across the pages measured within six weeks of the website launch.
Chloe Singleton of eight&four described work for Sandals Resorts that began with detailed customer questions and community discussions. The team then used GenAI Lens to benchmark AI responses and identify content gaps.
Both examples point toward a useful approach: learn what people need to know, publish clear answers and investigate whether AI reflects that information accurately.
Deloitte Digital’s Nick Purdon added a second dimension, discussing the rational information that helps people and machines assess a product, alongside the emotional connections that help brands participate in culture. Specifications, ingredients and evidence serve one purpose. Humor, taste and lived experience serve another.
His discussion of creators emphasized cultural relevance and lasting behavior change. Creators understand the communities they serve, and that knowledge can help a brand contribute something people value.
Use AI to assist with research, organization and analysis while giving people room to develop the idea and judge how it will land. A creator brief can set out the product facts and the desired outcome, while leaving space for the creator’s understanding of their audience.
Clear information helps discovery, while distinctive, relevant storytelling gives people a reason to care when they find you.
Use the time AI saves for strategic work
Meltwater CEO John Box and Chief Product Officer Chris Hackney opened Summit with a vision of connected intelligence across monitoring, discovery and action.
The demonstrations showed how AI can help teams investigate brand visibility, explore emerging conversations and work with Meltwater data through an AI assistant. In the keynote, an example from Havas Red described using custom Mira prompts to accelerate recurring reporting and monitoring. That is a valuable opportunity for busy PR and marketing teams, because gathering coverage, preparing briefings and organizing evidence can take up time that people need for planning, interpretation and relationships.
The Data Malarkey discussion raised a related concern: when people hand over too much of the thinking, they may lose the opportunity to develop their own understanding, which makes it important to decide which activities AI should handle and which experiences people still need.
Automate work where AI performs well and where you can assess the result, then decide how to use the capacity it creates. You might spend more time interviewing customers, testing a campaign idea, strengthening journalist relationships or helping leadership understand an emerging risk.
Purdon highlighted humor, taste and judgment as sources of competitive advantage - saving time on production creates room to develop those qualities.
Put the lessons into practice
In her closing remarks, Meltwater Chief Strategy Officer Alexandra Bjertnæs emphasized the value of curious, adaptable PR and marketing professionals. She encouraged the audience to keep asking why, going beyond the first answer to understand what sits underneath it.
That’s a useful habit for working with AI. Question the brief, examine the evidence and stay interested in the people behind the data.
Choose one application your team can test. Agree on the business question, establish a baseline and decide who will evaluate the output. Track whether the work becomes faster, whether the findings improve and whether people can make better decisions as a result.
An AI summary can be produced in seconds. Giving your team more confidence to act is the outcome worth working toward.
FAQs about AI in PR and marketing
How can PR and marketing teams use AI effectively?
PR and marketing teams can use AI for media analysis, audience research, campaign planning, content development and reporting. Effective use starts with a clear business question and suitable evidence. Define the task, provide context, establish how the output will be checked and connect the result to a decision or action.
What role should human judgment play in AI workflows?
Human judgment should guide the objectives, validate evidence and determine whether an AI recommendation is appropriate. PR and marketing professionals need to assess accuracy, cultural context, creative quality and reputation implications. Assign a reviewer with relevant expertise to challenge the output before it informs a decision or reaches an audience.
What are shadow insights?
“Shadow insights” was Antony Cousins’s term at Meltwater Summit on Tour London for business conclusions drawn from general AI assistants without appropriate research checks. An answer may sound comprehensive while relying on limited or outdated information. Teams can address this by making reliable data, clear research methods and guidance on interpreting results easier for colleagues to access.
How should teams measure AI brand visibility?
Measure whether your brand appears in relevant AI answers, how prominently it appears, how accurately it is described and which sources support the response. Compare results across relevant questions, competitors, models, markets and languages. Where measurable, examine subsequent audience behavior, such as website visits, so reporting connects AI visibility with the outcomes your business cares about.
What is a good first step for adopting AI in PR or marketing?
Start with a focused task that serves a business need, such as preparing a recurring briefing. Record the current process, time required and standard of quality. Test AI with appropriate data and human review, gather feedback from the people using the output and use the results to improve or expand the workflow.
