I just returned from Chicago and I’m still processing everything I heard at Salesforce Connections 2026.
The energy at an event like this is hard to describe to someone who wasn’t there. Thousands of marketers, developers, and executives all operating on the shared assumption that the future of customer engagement has already arrived — and the only question is whether you can keep pace with the leaders. The conference theme, mastering agentic AI, wasn’t aspirational language. It was a statement of direction.
Presentations from brands like Bank of America, Omni Hotels & Resorts, and Mars Petcare were impressive. Not in a “look what’s possible someday” way, but in a “here’s what we’re actually doing right now” way.
For those of us who have spent years arguing that Zero-Party data and hyper-personalization at scale is the loyalty industry’s north star, watching those use cases come to life was genuinely energizing. For the first time, I felt like the tools might exist to match the ambition in an affordable way.
So yes, I’m optimistic that retailers and brands will benefit from Agentic Marketing, not be destroyed by it. But there are messy complications to be sorted before the potential is realized.
When the Vision Clicks
The Salesforce Connections theme organized around the promise that brands now have the power to deliver digital experiences tailored for individual customers and that agents will be most effectively fueled by contextual data. The efficiency of this delivery is central to the theme, meaning AI systems are capable to act on behalf of customers, execute decisions in real time, and deliver experiences no human team could produce at the same speed and scale.
The presentation from Omni Hotels & Resorts highlighted their vision for “Conversational Loyalty,” starting with something as simple as a check-in text. The use case shared centered on guest recognition and a matching of interests in real time to enhance their stay. Imagine an AI agent knows you’re a Select Guest Icon, finds a tee time available at the course you prefer, and asks whether you’d like to book it with your Omni Credits. That’s the moment the guest loyalty relationship shifts from transactional to genuinely responsive.
The next use case was sharper still: an AI agent that detects when a restaurant check has been printed, instantly texts an offer to cover the meal with loyalty credits and settles the bill when the guest replies “yes”. No hunting for a waiter. No fumbling with a terminal. Just a frictionless, well-timed magic moment.
Explaining their “Vision for Democratizing Experiences,” Omni shared the scenario where AI flagged a delayed flight on the staff console before the guest arrives at the desk. The agent authorizes an “empathy upgrade” in real time so the front desk team can lead with care.
A third example described AI autonomously triggering VIP perks for a guest who has golfed at an Omni property but never stayed overnight. The premise: agentic AI can expand who gets a premium experience, not just automate delivery for those already at the top.
Omni calls it “democratizing luxury” and I’m predicting the concept will travel well.
Excitement Tempered by Accountability
The smartest presentations at Connections went beyond pure evangelism. The practitioners who impressed the most acknowledged that governance matters enormously.
Guardrails aren’t optional and the power of an agentic system is exactly proportional to the damage it can do if it’s misconfigured, poorly supervised, or deployed before the organization is ready for it.
That balance — capability tempered by accountability — is the posture every marketer should bring to this technology right now.
The Courts Are Drawing Lines
Here’s the messier part of this idyllic future state. While Salesforce Connections was making the case for agentic marketing, the courts have been making a different kind of case.
Amazon recently won a preliminary injunction against Perplexity AI’s Comet browser, which was using AI agents to execute purchases inside Amazon’s platform on customers’ behalf. A federal judge ruled that user authorization doesn’t override the platform’s own prohibition on automated access, and Perplexity’s agents were found to have disguised themselves as standard human browsers to evade detection. That’s a significant line in the sand for the entire category of agentic commerce tools.
CNN has filed its first AI copyright action, alleging Perplexity is distributing CNN’s journalism to power its AI responses without compensation. And in Florida, the Attorney General has sued OpenAI directly, alleging the company put profit ahead of user safety.
Each case is different in its specifics. But they all point to the same reality: the legal and regulatory ecosystem is beginning to respond to technology that has outrun the frameworks meant to govern it. That’s not a reason to stop moving forward, though it is a reason to pay attention.
Even Anthropic Is Saying Slow Down
While processing all of this alongside the optimism of the Salesforce event, I read an article released by Anthropic just as the conference wrapped up.
Their paper, “When AI Builds Itself,” laid out a data-backed argument suggesting the market is approaching the early stages of recursive self-improvement, meaning the point at which AI systems can meaningfully contribute to designing their own successors. Their engineers are shipping roughly eight times as much code per quarter as they were just a few years ago and as of May 2026, more than 80% of the code merged into Anthropic’s own production systems was written by Claude.
Anthropic is explicit that full recursive self-improvement hasn’t arrived and isn’t inevitable. But they state plainly that it could come sooner than most institutions are prepared for — and they are calling for a mechanism to slow or pause frontier AI development if alignment research can’t keep pace.
When one of the most consequential AI companies in the world makes that argument at potential cost to its own competitive interests, it’s worth sitting with for a moment.
The Question That Matters Most
The opportunity in agentic marketing is real. The Omni Hotels vision of a guest receiving an empathy upgrade because an agent flagged their delayed flight before they walked through the door is a powerful precursor of the future of loyalty delivery, and it is within reach.
It’s rather clear that most businesses are not yet equipped to optimize these tools. The gap between what’s demonstrated on a Salesforce Connections main stage and what most marketing teams can actually implement, govern, and measure is enormous. That’s not a criticism — it’s a practical observation about how long it takes organizations to absorb transformational technology.
Recent data supports this statement. Consumer comfort with AI agents making purchases on their behalf dropped from 70% to 45% between Q4 2025 and Q1 2026 — even as industry investment in agentic tools accelerated sharply over the same period.
Adoption is outpacing trust, and the gap is not a temporary friction point. It is the central problem that brands and retailers must solve. (Source: Riskified Agentic Commerce Pulse, Q1 2026)
The compelling question is: how do brands and retailers build and maintain trust with their customers in an era when agents are acting on their behalf?
That is not a technology question. It is a relationship question. The answer requires governance, transparency, clear rules about what agents can and cannot do, and an honest acknowledgment that the customer — not the algorithm — is still the one whose experience matters most.
The train has left the station. I’m on board. But I’d feel a lot better knowing that someone has checked the track ahead.
Bill Hanifin is Managing Editor of The Wise Marketer and CEO Wise Marketer Group. Follow him on LinkedIn.