Marketing teams have spent years connecting customer data to the tools that activate it. Now the same integration work is becoming the gate between AI agents that merely draft recommendations and agents that can actually support marketing operations. On October 9, customer-data platform Hightouch announced that it had acquired Strata, whose AI agent builds and maintains custom connections between business applications.
The deal is a useful signal for business leaders: as marketing AI moves from content assistance toward analysis and execution, dependable access to governed, current data may matter more than adding another model. But an acquisition announcement is not proof that integrations are already available across every stack—or that automation will produce better campaigns by itself.
What Hightouch bought—and what remains unknown
Hightouch describes Strata as an agent-driven, no-code platform for building and maintaining custom API integrations. A team specifies what information should move and where; Strata’s agent maps fields and events, configures transformations, creates the integration, and keeps it working as connected systems change. Hightouch says the technology will expand its ability to connect marketing tools and supply its agents with more context. Strata CEO Kelsey Mullaney is joining Hightouch to help lead revenue programs and go-to-market strategy. (Hightouch’s announcement; the release carried by the Associated Press.)
The companies did not disclose financial terms, a product-integration schedule, or which connections will be available first. Treat promised onboarding and flexibility gains as the company’s expected outcomes, not independently measured results. The strategic point is the target: shrinking the backlog of bespoke data connections that can keep marketing ideas waiting on engineering.
Why data plumbing is becoming an AI marketing constraint
AI agents need more than a prompt and a model. To answer a performance question or prepare a customer segment, they need the right information from systems such as a CRM, analytics, advertising, commerce, and support—plus consistent definitions and permission to use it. A broken or stale connection can undermine the whole workflow, even when the model is capable.
Research points to that gap. In a survey of more than 500 U.S. technical leaders, Anthropic and research firm Material found that 46% cited integration with existing systems as an obstacle to agent implementation, while 42% cited data access and quality. The study also found 46% expected marketing and sales to be among the functions most affected by agents in the near term. These are self-reported, cross-industry expectations—not a forecast that every marketing team will see the same impact. (2026 State of AI Agents report.)
Marketing-specific evidence reinforces the operational challenge. Salesforce’s survey of 4,450 marketing decision-makers found that 98% reported at least one data-related barrier to personalization; just 56% said they had complete access to sales data and 51% to commerce data. These survey results do not prove that an integration platform alone resolves the problem. They do show why AI that cannot reach trustworthy customer context risks producing generic recommendations—or automating the wrong action. (Salesforce’s 2026 State of Marketing findings.)
More connectors do not automatically mean more control
Strata’s promise is faster connectivity, but speed is only one requirement. Each new path between systems raises questions about which records an agent may read, whether it can write or trigger actions, how customer consent and retention rules carry through, and who can review what happened. Field mapping also matters: if two platforms define “qualified lead,” “revenue,” or “active customer” differently, a seamless data transfer can still deliver misleading analysis.
For leaders evaluating agentic marketing platforms, ask vendors to demonstrate permissions inherited from source systems, data lineage, freshness monitoring, error handling, and audit logs—not just the size of an integration catalog. Start with an observable workflow and a human approval step. An agent that prepares a campaign audience for review is a safer first deployment than one that can silently change targeting or spend.
A 30-day plan to make your marketing stack agent-ready
- Choose one workflow with measurable friction. Examples include weekly performance reporting, lead enrichment, audience preparation, or summarizing campaign results. Record the current cycle time, handoffs, and error rate.
- Map the data path before buying a connector. List each source, owner, required fields, refresh frequency, access rules, and metric definitions. Identify where a missing or stale field would change the decision.
- Pilot read-only, then expand deliberately. Compare the agent’s output with a known answer for several cycles. Track accuracy, completeness, time saved, and exceptions. Add write access one action at a time, with named approval and rollback procedures.
- Evaluate integration promises in your own environment. Ask vendors to show how a connection is created, maintained, permissioned, monitored, and exported if you leave. Test a representative tool pair and include your security and data owners in acceptance.
Hightouch’s Strata acquisition is a bet that AI marketing will be limited less by idea generation than by the systems and controls that let agents use business data responsibly. Companies that get those foundations right can test automation with clearer evidence—and keep humans accountable for consequential decisions.
Real Internet Sales helps businesses turn AI and digital marketing into practical, measurable programs. Call 803-708-5514 or visit realinternetsales.com to discuss your strategy.