Salesforce lets marketers measure their AI reputation

Salesforce’s marketing keynote at Dreamforce in San Francisco centred on three questions: what do AI assistants say about your brand, can a campaign adjust itself, and how do you turn B2B signals into pipeline?

The third theme builds on the agents that Salesforce previously gave a name and a function. The first two brought news for marketers.

AI as the first impression

The opening speaker began with a personal example. Claude helped choose a wine for their own wedding, matching the menu and staying within budget, and did not provide a list of links but an answer. The question for marketers is whether their brand appears in such an answer and how. A brand’s first impression increasingly comes not from its own website or latest email campaign, but from an answer engine, powered by reviews, communities and sources over which the brand has no control. Salesforce concludes that marketers will now be working for two audiences: people and AI.

Contentful is now part of the picture

Salesforce completed its acquisition of Contentful on September 1, earlier than the end of the third quarter of the fiscal year that had been indicated when the deal was announced. The purchase price was approximately $1.5 billion in cash. The Berlin-founded platform manages content separately from its presentation layer, allowing the same content to be delivered to a website, app, email or AI agent. Within Marketing Cloud, Contentful is intended to connect Salesforce customer data with content to create personalised experiences.

Palmata measures AI reputation

Contentful launched Palmata in June, a standalone product that examines how answer engines such as ChatGPT, Gemini, Claude and Perplexity portray a brand and identifies which content shapes that perception. A first report is free. More comprehensive subscriptions cost up to $600 per month.

The on-stage demo used a fictional sports brand. It was mentioned in 62 percent of the AI responses analysed but cited as a source in only 26 percent. Of the perception, 23 percent was positive. Palmata looks not only at a company’s own website but at everything on the open web: press releases, news articles, competitors’ pages and blogs. In the demo, the warranty and service pages were actually found to be negatively influencing the brand’s image because AI systems were misinterpreting them.

For every recommended action, Palmata provides supporting evidence and a simulated impact. To do this, the system creates its own text variations, submits them to answer engines and calculates which performs best. In the demo, the recommendation was to add a short summary above the blog articles.

Execution through Claude

The change was not made in the CMS but in Claude, through MCP connections with Contentful, Palmata and Salesforce. Claude retrieved the recommendation, found the content in Contentful, modified it and published it. This took about a minute. After jumping several weeks forward in time, a new report showed more mentions, more frequent source citations and a 6 percent improvement in AI perception. This was a simulated demo, not a customer result.

Salesforce also presented MCP connectors for Marketing Cloud Engagement and Marketing Cloud Next. These allow marketers to operate the products from other AI environments and agents. This fits with the direction of AIforce, the CRM without its own interface.

Formula 1 gets to know its fans

Matt Kemp, Head of Customer Data Operations at Formula 1: “Everyone thinks Formula 1 has been around for 75 years. Until eight years ago, we were primarily a B2B company and didn’t know who our fans were.”

Since 2018, the number of fans has grown from just over half a billion to more than 830 million. Formula 1 began collecting data through its own products, services and email, building fan profiles rather than sending one email to everyone. In the first half of 2026, the organisation recorded 1.6 billion social media interactions, 48 percent more than a year earlier. The audience has also changed. A few years ago, it consisted almost entirely of men over 45. Today, 43 percent are women and 42 percent are under 34.

At Formula 1, data and content are spread across many channels and sources. Marketing Cloud and Contentful bring these together into a single profile for each fan. The first step towards conversations with fans is the assistant Formula 1 launched in March, which answers questions about the new technical regulations. The next step is to add those conversations to the fan profile and connect them with content, video, live data and the loyalty programme. The goal is one relationship per fan rather than one message for an entire group.

Campaigns built around a goal

Marketers still spend weeks working on segments, channels, timing, approvals and predefined if-then journeys. Meanwhile, they are judged on pipeline and revenue, rather than the number of messages or variants. Marketing Cloud’s new campaign agent therefore starts with a goal.

In the demo, the marketer did not begin in the segmentation screen or with a multi-step customer journey, but with the marketing plan as context and a brief for each campaign. The brief contains the primary campaign objective, which the agent optimises for. If information is missing, the agent asks for it, such as the budget. The demo showed an email campaign. The agent works with all channels supported by Marketing Cloud.

Who gets which campaign

Under this approach, a segment no longer determines who receives a campaign, but who is eligible for one. With Adaptive Campaign Audiences, the agent assesses each person to determine which ongoing campaigns they are eligible for and selects the campaign that best fits them at that moment. This can also mean that someone receives no message. In the demo, about 16 percent of the target audience was moved to a different campaign.

After launch, the agent continues testing and shifts volume towards the content that performs best. Goals can include sign-ups, events or the loyalty programme.

Content from Contentful

The content is created in the content canvas, which Eric Zenz from Salesforce previously discussed. What is new is the connection with Contentful, which will become available as a pilot in October. The agent can use existing content from the Contentful library as the basis for new variants. The canvas is not simply an image generator but a complete email editor with personalisation based on data in Marketing Cloud. Translating content for other markets takes just a few clicks, after which a colleague can review the translation. Published content automatically appears in the campaign.

Pilot at Under Armour

Under Armour is testing the campaign agent in a pilot. The brand has more than 50 million subscribers. Because of creative approvals and production processes, it previously took more than two weeks to get a campaign live. According to Salesforce, the time required to build an email fell from hours to minutes in the pilot.

www.salesforce.com

Follow Marketing Report on LinkedIn