Every marketing executive I talk to has the same story: their company has adopted AI tools, their teams are experimenting and somewhere in a slide deck there’s a roadmap that says “AI-powered” next to nearly every initiative. Access to AI is no longer the challenge. Confidence is.
Because here’s what that same executive is also thinking about: What happens to our customer data when it touches an AI model? Who can see it? Where does it go? If a regulator asks us to explain our AI activity six months from now, can we?
For marketing teams, this tension is sharper than almost anywhere else in the enterprise. Marketers sit on first-party buyer data — purchase history, behavioral signals, personal identifiers, engagement patterns, and sensitive demographic and firmographic data — that is both the fuel for great AI-powered experiences and the biggest liability if it’s mishandled. The promise of AI in marketing is real. But so is the risk of getting the foundation wrong.
Optimize your B2B marketing with connected analytics
Learn three strategies to bridge the divide between marketing and sales reporting, using next-gen analytics to tie cross-channel engagement directly to pipeline and revenue.
Simply put, the AI revolution is changing the way we work. Soon, people will not only manage other employees but also oversee digital agents. Companies are operating in an era where governance, data privacy, security and responsible AI use are paramount. The ones moving forward with confidence aren’t the ones with the most sophisticated models. They’re the ones who solved the trust problem first.
Values have to be operational, not aspirational
At Salesforce, trust is our first value. It comes before innovation, before customer success, before everything else — because we believe that nothing else works if trust breaks down.
But values only matter if they’re operational. It’s easy to put “trust” on a website. It’s harder to architect an entire AI platform around it.
We know that our customers are trusting us with their hard-earned data: every interaction, every contact, every buyer they are engaging. And part of our trust mission is to keep that data safe. Your data belongs to you.
When we made the decision to bring AI into the enterprise — at scale, across every product — we didn’t start with the model. We started with the question: how do we make it safe for our customers to actually use this? The answer became the Einstein Trust Layer.
It is critical for brands to be compliant, ranging from following laws like GDPR to regulated industries like finance and healthcare. It takes a lot of hard work across teams to accumulate first-party data and make sure it is compliant in an efficient and thorough way. That being said, our decision wasn’t a compliance exercise or a legal requirement. It was a product philosophy.
Our values don’t sit in a mission statement — they show up in how we build. And for AI, that means governance, security and data protection aren’t features you add after the fact. They’re the foundation everything else is built on.
What the Einstein Trust Layer actually does for your data
Let me make this concrete, because “enterprise-grade AI safety” can sound like jargon until you see what it means for your marketing data specifically.
Zero LLM data retention. When your data is used to power an AI interaction — a personalized email, a recommended next step, a generated insight — it is never stored by the model provider. Your customer data touches the model and comes back. It doesn’t stay there, it doesn’t train anything, and it doesn’t become someone else’s asset.
PII masking and tokenization. Before any data reaches the large language model, the Trust Layer detects and masks sensitive information — names, email addresses, phone numbers, any personally identifiable information. Those fields are replaced with tokens. The model never sees the raw PII. When the output comes back, the Trust Layer re-maps the tokens to the real values for your team. Your customers’ information never travels unprotected.
Full audit trail. Every AI action is logged. Every prompt, every output, every interaction — complete visibility for governance, compliance, and accountability. If your legal or compliance team ever needs to reconstruct what happened and why, that record exists. This is what responsible AI looks like in practice.
Governance built in, not bolted on. This is the one that matters most at the executive level. Many AI deployments treat governance as a layer you add once something goes wrong. The Einstein Trust Layer inverts that model — governance is built into the platform architecture from day one. That means your marketing team operates within a guardrail system that protects them, your customers, and your company without slowing anyone down.
For marketing leaders specifically, this translates directly: your campaigns, your AI-generated content, your customer segmentation, your personalization logic — all of it runs on a foundation that is secure, auditable, and compliant with industry and governmental regulations by design.
What this means when marketing puts it to work
When you’ve solved the trust problem at the infrastructure level, something interesting happens: your marketing team stops asking permission and starts moving.
This is only possible because the data foundation is secure. Marketers can go deeper into personalization, move faster on opportunities and scale programs they couldn’t have run manually because they’re not carrying the anxiety of “what are we doing with this data?”
The most powerful marketing experiences are built on the deepest customer understanding. The deepest customer understanding requires the most sensitive data. And the most sensitive data demands the most robust trust architecture. These aren’t competing priorities. They’re a chain.
Modernize your email program
Learn how to audit your lifecycle maturity and shift from low-impact batch sends to high-performance, behavior-based orchestration.
Remember: the trust foundation has to come first
The AI conversation in most boardrooms is still focused on capability: what can the model do, how fast can we deploy, what’s the ROI? Those are the right questions eventually. But the first question, the one that determines whether any of it is sustainable, is simpler.
Can we trust it with our customers’ data?
The companies that answer that question confidently, at the infrastructure level, are the ones that will build a lasting AI advantage. Not because they moved the fastest, but because they built on the right foundation.
At Salesforce, we start with trust. Everything else follows.
How is your organization thinking about the trust foundation for AI? I’d love to hear where executives are finding the biggest gaps, and what’s giving them the most confidence. Connect with me or drop a comment below.
