Jonathan Beeston, Product Marketing Director for EMEA at Salesforce Marketing Cloud, on why AI-powered audience segmentation changes the velocity of marketing without replacing the marketer, why trust is the necessary companion to any AI deployment, and what a single customer view is actually for.
"I didn't get into marketing because I wanted to spend more time with spreadsheets. But here I am."
Why AI changes the velocity of audience segmentation without changing what great marketing actually is, and why trust is the only safe starting point
Jonathan Beeston has spent his career at the intersection of product and marketing. He started producing CD-ROMs, moved into search engine optimisation in 2002, and has been following the technology wherever it leads ever since. His current focus is the question every enterprise marketer is trying to answer: how do you use generative AI to accelerate marketing operations without inadvertently leaking confidential customer data into a large language model?
His answer involves a framework rather than a tool: retrieve only the relevant data, construct the prompt, send it to the model without retaining personally identifiable information, validate the output for toxicity and accuracy, and make the whole workflow auditable. The practical illustration is audience segmentation: a task that previously took a skilled operator considerable time can now be completed in fifteen seconds. But the fifteen seconds is only valuable if the segment is activated with the right message. Which is why a thousand segments with two messages is still just two segments.
"Imagine the Venn diagram between sales, field marketing, and product. Product marketing is right in the middle of that. My mantra for internal enablement is no AE left behind."
The no AE left behind principle matters at scale because enablement is the mechanism by which product innovations become commercial opportunities. If a product marketer builds excellent positioning but the sales team cannot communicate it, the investment is wasted.
"Every company is sitting on all sorts of interesting data. Generative AI is an amazing tool to pull all of that data together in a way that is accessible in real time. But that is your company confidential data. So how can you utilise generative AI without giving it that data?"
Beeston describes the framework Salesforce uses: retrieve the relevant data, construct the prompt, submit it to the model without the data being retained, validate the output, and make the whole flow auditable. CMOs across the industry identify this as both their number one opportunity and their number one concern.
"You can say, build me this segment, people who have bought my product in the last 14 days but have not activated the credit card. And it does it. It is doing it in about 15 seconds."
The fifteen-second example is deliberately concrete. What changes is the time between having the thought and having the segment. For teams managing hundreds of segments across multiple campaigns, that change in velocity is commercially significant.
"It is not my adage but if you can have a thousand audience segments but if you only have two creative messages, then you only have two segments."
Data infrastructure and segmentation capability are necessary conditions but not sufficient ones. The marketer still needs to develop message variation genuinely relevant to each segment. Beeston gives the example of a customer with an open service ticket: even with the best segmentation, marketing to them with an upsell at that moment is the wrong call.
"It is a mistake to put B2B on one side of the room and B2C on the other. There is an individual you want to put a message in front of. I do not think any channels are off limits. Connected TV is really interesting."
If the decision maker you want to reach watches Formula One, an ad in the broadcast makes sense. The identity-based targeting capabilities emerging, including cross-device matching and connected TV addressability, are making it possible to reach B2B buyers with consumer-quality precision at every point in their day.
"The single customer view is not the end. It's the means to activate the right kind of marketing message."
Host Tell me about the product marketing role.
Beeston Product marketing is at the Venn diagram between sales, field marketing, and product. It is trying to unify what our messaging is behind our products, how we articulate that to customers and prospects, and how we make sure our sales teams understand that. My mantra for internal enablement is no AE left behind.
Host What is the AI trust challenge?
Beeston Every company is sitting on interesting data. Generative AI is an amazing tool to pull all of that data together in real time. But that is your company confidential data. You need a framework. How can we retrieve the relevant data, put it into a prompt, give that to a large language model, but ensure none of that data is ultimately retained? And how can we make sure the output is auditable and free from toxicity?
Host What does AI-powered segmentation look like?
Beeston You want to create an audience segment: people who bought my product in the last 14 days but have not activated the credit card. Now you can just type that in and the system builds the segment in about 15 seconds. You can still edit it the old-fashioned way. But it accelerates enormously. Building audience segments was the number one thing marketers said they need technical help with.
Host Why do a thousand segments with two messages equal just two segments?
Beeston If you have a thousand audience segments but only two creative messages, you only have two segments. Being able to iterate on the right kind of message and put in the right kind of personalisation, that is where connecting with your data becomes really important. If a customer has an open service ticket, maybe I should not be trying to upsell them.
New leaders and insights. No noise.