Your martech stack is more sophisticated than it has ever been. Your campaigns still take three weeks to launch.
This is the story of modern marketing. Organizations bought the data platform, the CDP, the creative tools, and the orchestration layer. But the tools were never the bottleneck. Coordination was.
A campaign passes through planning, analytics, creative, data, legal, and channel teams. Each handoff adds time: a brief to write, a spec to clarify, an approval to chase. Work is queued. AI agents promise to compress that cycle, but most stall where people do: they act without full context. The fix is not better agents. It's giving them the whole picture.
Why agents may produce mediocre work
Marketing context lives in pieces. Customer data sits in the warehouse. Performance data sits in analytics. Brand rules sit in a design system. Compliance status sits in a ticketing tool. No single person sees all of it, and neither does an agent asked to draft a campaign.
Feed an agent a fragment, and it returns a fragment: copy that ignores past performance, an audience that violates a suppression rule, or a concept that drifts off brand. The output looks plausible but fails when it reaches real customers.
The real work is not “adopting AI.” It’s closing three specific gaps:
- No shared context. Agents guess at brand voice, prior results, and market conditions.
- No governance. Decisions can't be audited, so no one trusts them enough to ship.
- Locked-up knowledge. Answers live in separate teams and tools, so people burn time checking what an agent should already know.
Close those three gaps and coordination time shrinks. That’s the real prize, and it changes what marketing roles look like day-to-day.
Context as the foundation, not the afterthought
Most major platforms are racing to build this context layer, including Hightouch. Hightouch’s AI Decisioning reads directly from a customer's data platform, such as Databricks, and connects to activation tools like Salesforce Marketing Cloud for channel and content context. The principle is simple: gather everything an agent needs (customer data, channel performance, market signals, and brand context) in a single, well-governed place. This allows the decisioning agent to reason over real results, not a guess.
An increasing number of CDPs include a decisioning component. This agent picks the message, offer, channel, and timing per customer. But execution is only half the problem. The other half is trust. A human needs to understand why an agent makes a decision before anything ships. That’s why Hightouch surfaces reasoning through companion agents and insights views so that marketers can interrogate decisions in natural language.
Data platforms like Databricks are also evolving to embed governance. Every agent action is logged and tied to a real permission, using the same rules that apply to a person. You know which agent did what, on whose behalf, and when. Reasoning happens inside a secure data environment, so sensitive customer data is never copied to a separate tool for analysis. Marketers describe the outcome, and agents execute inside purposeful guardrails, preventing the off-brand output that erodes trust.
The last mile: talking to your whole stack
Context allows AI agents to make better decisions, but it may not reduce back-and-forth between humans.
Model Context Protocol (MCP) servers help close this gap. MCP is an open standard for connecting AI assistants like Claude and ChatGPT to tools and data through connectors and plug-ins, allowing marketers to converse with their platforms in natural language. The number of connectors is growing fast, and many are already available for applications such as JIRA, Databricks, and Adobe Experience Platform.
This democratizes knowledge once locked in separate teams, but does not negate the need for experts. A “human in the loop” must still confirm the output is accurate and on brand. Public mistakes can cost far more than time saved in this process.
Put the two together and timelines shrink: agentic platforms address coordination, MCP connectors expose locked-up knowledge, and the three-week relay becomes a conversation that resolves in days or even hours.
How marketing roles change, and why that’s the point
When execution logistics move to agents, the people around them move up a level. Teams worry about this shift most. What will the doers do? The doing moves to higher-value work, creating real opportunity for people willing to adapt.
Tactical assembly moves to agents, while humans handle judgment about strategy, brand, and what resonates. Marketing operations groups lean into governance: defining objectives, setting guardrails, and monitoring the quality of what agents produce. Teams that own CRM and data operations become context engineers, focusing on building engaging, adaptive customer journeys instead of shipping static segments on request. Their output becomes the foundation agents reason on.
This is an operations decision
The temptation is to treat all this as a purchase: buy the platform, connect the servers, done. That misses the work that determines success. The gains come from redesigning how marketing happens.
That’s what Merkle’s AI-Powered Delivery is built for. We help marketing teams rebuild how work moves by setting governance and reshaping roles so agents act on full context. Talk to our team.
Start with the work, not the tools. The speed follows.
