San Francisco startup Kana emerged from stealth with $15M in seed funding to deliver flexible, human-in-the-loop AI agents that plan, run, and optimize marketing campaigns. The round was led by Mayfield, and managing partner Navin Chaddha is joining the board—an early signal that investors see room for agentic systems that go far beyond copy generation.
Co-founders Tom Chavez and Vivek Vaidya are returning to familiar terrain. Their prior companies Rapt (acquired by Microsoft) and Krux (acquired by Salesforce) were formative to modern ad and data platforms. Kana is their bid to bring composable “agents,” rather than monolithic tools, to the messy realities of media planning, audience targeting, and cross-channel reporting.
A Fourth Act From Veteran Ad Tech Builders
Chavez and Vaidya incubated Kana out of their venture studio super{set}, positioning it to integrate with the stacks they helped shape. That matters: marketers rarely rip and replace—most orchestrate across CRMs, CDPs, DSPs, analytics suites, and walled gardens. Veteran founders with deep integrations and pragmatic roadmaps tend to win trust where net-new platforms often stall.
The company pitches speed and adaptability as its edge. Rather than a rigid workflow, Kana’s agents can be assembled “on the fly,” then tailored to a brand’s taxonomy, data permissions, and measurement framework—an approach closer to microservices than to an all-in-one suite.
What Flexible AI Agents Mean In Practice
Kana describes “loosely coupled” agents assigned to distinct jobs: ingest a media brief, clarify objectives, identify audiences, draft a cross-channel plan, pull inventory and research, propose budgets, and then run autonomous tracking, optimization, and reporting—always with human approval gates.
Think of it as an orchestration layer atop existing systems. Agents call out to ad platforms, analytics, and data providers, then reason over the responses to recommend next steps. Marketers can approve, edit, or reroute tasks, while role-based controls ensure brand safety and compliance with internal policies.
Example: a CPG team uploads a seasonal brief. One agent refines goals and KPIs; another evaluates first-party signals in the CDP; a planning agent proposes channel mix with reach and frequency targets; an optimization agent sets experiments and shifts budgets as performance data arrives. The net effect is compressing weeks of coordination into days without ceding final say.
Synthetic Data For Targeting And Research
Beyond agents, Kana emphasizes synthetic data generation to augment sparse or expensive third-party datasets. Used well, synthetic panels can accelerate market research, scenario testing, and early audience hypotheses while keeping PII off-limits—timely as marketers navigate signal loss and privacy restrictions.
Care is warranted. Industry groups like the Association of National Advertisers have flagged waste and quality concerns in programmatic markets, and synthetic data adds its own risks around representativeness and bias. Best practice—endorsed in frameworks from NIST’s AI Risk Management guidance to emerging enterprise guardrails—calls for clear provenance, validation against ground truth, and continuous monitoring to avoid spurious lift.
Kana’s human-in-the-loop stance is a practical hedge: analysts can review synthetic assumptions, reject weak segments, and promote only those findings that replicate in live campaigns.
Battling Incumbents With Configurability
Kana is entering a crowded field. Big suites from Adobe, Salesforce, Google, and Meta now bundle AI across insights, creative, and bidding. Content-focused startups like Jasper and Copy.ai cover ideation, while agencies are experimenting with custom agentic workflows atop orchestration frameworks.
Kana’s bet is that configurability—“build with” rather than “build or buy”—becomes the moat. Instead of bespoke one-offs, the company says it can rapidly compose reusable agents that fit a customer’s governance model, data contracts, and channel mix. The challenge will be scalability: showing that fast-tailored deployments can standardize without drifting into services-heavy margin traps.
Reliability will also be a proving ground. Marketers will expect transparent logs, test harnesses, and shadow-mode rollouts before agents touch real budgets. Observability, safety guardrails, and clear attribution of agent decisions are likely to be as decisive as raw model quality.
Funding And What To Watch Next From Kana
The seed capital will fuel hiring across engineering, product, and go-to-market, as Kana lines up pilots with brands that need orchestration across legacy systems. Pricing hasn’t been disclosed, but usage-based models tied to workloads and outcomes would align with how teams evaluate ROI on automation.
Markers of early success will be straightforward: faster campaign cycle times, fewer handoffs between planning and activation, measurable lift validated by independent analytics, and clean interoperability with major ad platforms and measurement partners. If Kana can convert its founders’ track record into repeatable, safe agent deployments, it will have a credible shot at redefining how modern marketing gets done.