This article was based on the interview with Treasure AI CEO Kaz Ohta on keeping autonomous agents on-brand by Greg Kihlström, MarTech advisor and expert for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
As marketing leaders, we’ve become accustomed to tectonic shifts. We’ve navigated the move from mass media to digital, the explosion of social channels, the rise of big data, and the subsequent, necessary wrangling of that data into Customer Data Platforms (CDPs). Each evolution has demanded new skills, new processes, and a new way of thinking about the customer relationship. Now, we stand at the precipice of another transformation, one that makes the previous shifts feel like mere prologues. The era of the autonomous AI agent is not on the horizon; it is here, and it’s poised to fundamentally change the nature of brand interaction.
The central challenge is no longer just about collecting and understanding customer data. It’s about empowering technology to act on that data, autonomously and at scale. When your brand is represented by AI agents making thousands, or even millions, of micro-decisions every day, a critical question emerges: How do you ensure those decisions are not just intelligent, but are also fundamentally *yours*? How do you keep an autonomous agent on-brand, on-strategy, and in compliance? In a recent conversation with Kaz Ohta, CEO and co-founder of Treasure AI, we explored this very issue. His company’s recent evolution from Treasure Data, a CDP leader, to Treasure AI is a microcosm of the industry’s pivot from data management to intelligent action, and his insights provide a practical framework for navigating this new agentic world.
The New Reality: An Internet of Agents
The first concept we must internalize is that the interaction model is changing. For years, the goal has been to bring customers to our owned properties—our websites, our apps—where we control the experience. However, the rise of consumer-side AI assistants is rapidly decentralizing the top of the funnel. Customers are no longer starting their journey with a Google search that leads to a blog post; they’re asking their personal AI assistant for a recommendation. This creates a scenario where your brand’s AI must interact with a customer’s AI, long before a human is ever involved.
As Ohta explains, this isn’t a future-state prediction; it’s an observable trend with massive implications.
“Imagine we have 5 billion internet connected people having AI assistant for each. The only way to handle those AI assistants, billions of AI assistants, is to have an AI for you… it’s not just about data. It’s really the decisioning the AI on top of it that matters next.”
This shift from a human-to-brand to an agent-to-agent model is profound. It means our meticulously crafted landing pages and SEO strategies, while still important, are becoming insufficient. The new top of the funnel is a conversation between bots. For a brand to even be considered in this new paradigm, it must have its own agents capable of understanding context, negotiating terms, and persuasively communicating value to another machine. The companies that thrive will be those who equip their agents with the best data, the clearest instructions, and the most robust decision-making capabilities. This isn’t just a new channel; it’s a new internet.
The Foundation: Your Data Platform is Now the AI’s Brain
With the rush to embrace agentic AI, it’s easy to get mesmerized by the capabilities of large language models and forget the fundamentals. An autonomous agent, no matter how sophisticated, is only as effective as the data and context it’s given. The adage “garbage in, garbage out” has never been more relevant, or the stakes higher. This is where the years of hard work standardizing, cleaning, and unifying customer data in platforms like CDPs pays its most significant dividend. That infrastructure is no longer just a system of record; it becomes the centralized brain and memory for your entire fleet of AI agents.
Ohta emphasizes that the glamour of AI implementation is built on the unglamorous, essential work of data governance.
“In reality of implementing AI agent, 90 to 95% of the work is actually having clean data and context… what AI needs is you actually need governed customer data, and that’s within CDP.”
This is a critical point for marketing leaders. Your investment in a CDP or a modern data warehouse is not legacy tech; it is the essential launchpad for your AI strategy. Without a governed, reliable source of truth, an autonomous agent is operating in a vacuum. It can’t personalize effectively because it lacks customer history. It can’t make ROI-driven decisions because it lacks performance data. And, most dangerously, it can’t be trusted because it lacks guardrails. The concept of governance must be baked in from the start. Ohta suggests treating AI safety as an engineering problem, starting agents with “zero permission” and gradually granting them capabilities within a secure, containerized environment. This ensures that autonomy doesn’t devolve into chaos, and every action can be audited and traced back to the core data and rules you’ve established.
The Leadership Challenge: Shifting from Doers to Orchestrators
The introduction of capable AI agents will inevitably change the structure and function of our marketing teams. This isn’t a simple story of automation replacing jobs. Rather, it’s a redefinition of roles. Repetitive, tactical work—building segments, manually creating journey flows, A/B testing copy variations—will increasingly be handled by agents. This frees up human talent for higher-order tasks: strategy, creative oversight, and, most importantly, the orchestration of their new AI colleagues.
The new role of the marketer, and the marketing leader, is to become an expert conductor of an AI-powered orchestra. You are still accountable for the final performance, even if you aren’t playing every instrument.
“Your job basically is to orchestrate your AI agents, and then let it work together, but in fact, you’re responsible for every single outputs they’re generating… I’m actually start observing some sort of laziness in terms of AI, which is, okay, you ask some work, and you just delegate to AI, and then you just respond with the output without any review. And that’s the new type of laziness I think that’s happening in the world.”
Ohta’s observation about “AI laziness” is astute and serves as a warning. The temptation to blindly trust and forward the output of an AI is a significant risk. The effective marketing teams of the future will be those that develop a new muscle for human-AI collaboration. This involves learning how to write effective prompts, critically evaluate AI-generated output, and provide feedback that improves the model over time. Leaders must foster a culture of responsible delegation, where AI is treated as a powerful, but junior, team member that requires oversight and mentorship. As Ohta puts it, “people without AI will be replaced by people with AI.” The mandate for leaders is clear: equip your teams with the skills and mindset to be the latter.
This transition from data custodians to AI orchestrators is the defining challenge and opportunity for marketing leaders today. The technology is no longer a bottleneck; the compute power and AI models are becoming commoditized at a staggering rate. The true differentiators will be the quality of your proprietary data, the clarity of your strategic direction, and the ability of your team to effectively manage a hybrid workforce of humans and AI agents.
The path forward requires a dual focus. First, we must continue to invest in the foundational layer of clean, governed, and accessible customer data. This is the fuel for everything that follows. Second, we must begin to redesign our teams and processes around the concept of orchestration, not just execution. This means training our people to be AI-native thinkers, managers, and strategists. The agentic world is coming fast, but for those who have laid the groundwork and are prepared to adapt their leadership style, it represents an unprecedented opportunity to engage customers with a level of personalization, relevance, and efficiency we could only have imagined a few short years ago.
