This article is based on a Q&A with Kana Head of Marketing Jessica Vose and Greg Kihlström for The Agile Brand Guide’s Expert Mode series.
Every few years the industry sells marketing a layer that will finally make the rest of the stack behave. The marketing cloud did it in 2012. The customer data platform did it in 2018. Both promised coordination, both arrived with a diagram showing everything else plugged neatly underneath, and both ended up as one more box on that diagram. So when a vendor turns up in 2026 saying marketing doesn’t need more tools, it needs an operating layer, the honest reaction from most of us is that we’ve heard this one before.
Jessica Vose, who runs marketing at Kana — an agentic AI marketing platform founded by the team behind Krux and Habu — doesn’t argue with the pattern. She grants it, then argues that the earlier waves were solving a different problem than the one they were sold against, and that there’s a test for whether this wave is any different. The test has nothing to do with the demo.
Consolidation Was Never the Same Job as Coordination
The distinction Vose draws is a narrow one, and it’s worth holding onto because most vendor pitches blur it deliberately. A marketing cloud consolidated logins. A CDP consolidated records. Neither one made a call. Asked what her own category’s predecessors actually delivered, she didn’t soften it.
“The marketing cloud and the CDP both sold coordination and both mostly delivered consolidation; for example, a single vendor’s login screen instead of six, a single database instead of six. That’s progress, but it isn’t the same job as coordination.”
A record, as she puts it, “is not a recommendation or a decision.” That’s the sentence to carry into your next four vendor meetings. When a system is described to you, work out which of the two problems it’s solving — visibility, or the gap between insight and decision — because a pitch that keeps returning to the unified customer view is a CDP with an agent bolted to the side, and you probably already own one of those. Vose’s version of coordination is narrower and more falsifiable: the system senses drift in a channel mix, recommends a change, acts inside guardrails you set, then logs what it did for you to check.
The Expensive Waste Never Shows Up on a Media Invoice
Say “wasted spend” to a CMO and the mind goes to media — frequency burn, bad targeting, the fraud line finance already flags every quarter. Vose agrees that waste is real. She also thinks it’s the smaller pile, and the example she reached for isn’t a media example at all. It’s a promotion in enterprise CPG working its way through an approval chain.
“A national promotion needs sign-off from brand, legal, media, and between a couple of regional teams. By the time it clears every desk, the market conditions the promotion was built for have shifted … and the team runs it anyway, because the cost of restarting the review is higher than the cost of running a now underperforming promotion.”
Nobody expenses that. It shows up as a promotion that underdelivered, and the post-mortem blames the creative. Her broader list of culprits is specific in a way that should feel uncomfortably familiar: the audience analysis rebuilt four separate times by four agency teams because nobody trusts or can find the last version, the promotional calendar that lives correctly in a planning tool and then gets hand-rebuilt as a slide for Tuesday’s leadership review. What’s notable is that Vose draws a line she has no commercial reason to draw. Waste sitting in review architecture, she says, is not a data problem, and no amount of better audience data touches it. A layer can compress decision latency and duplicated manual work. It can’t renegotiate who signs off on a national promo.
Three Teams Shouldn’t Buy This, and One Question Sorts the Real Ones
Kana sells the layer, so we asked Vose to make the case against buying one. She named three organizations that shouldn’t, and the first should give pause to a lot of enterprise marketing teams who think they’re further along than they are.
“A team without governed, first-party access to its own data. Coordinating disorganized inputs faster just produces disorganized outputs faster, at higher volume.”
The second is a lean team — two or three people, one ESP, one paid channel — where the constraint is hours, not coordination, and software doesn’t change that math. The third is any organization mid-reorg or mid-integration, where a system built to enforce coordination will, in her words, “automate the disagreement faster.” All three should settle the argument about which decisions they want automated before buying something built to automate decisions. As for sorting a real coordination layer from a point solution with new slides, she’d lead with a question about memory rather than autonomy: “Show me a decision your system made three months ago. What’s different about how it makes that same decision today?” If an experienced marketer overrides the recommendation and nothing about the system changes, you’ve bought automation with a subscription. Ask what happens to the override. It’s a better question than anything on the security questionnaire, and it takes about eleven words.
Marketing Should Own the Call, Not the Guardrails
Kana’s own research, The Agentic Divide, surveyed 225 senior marketing, data, and AI leaders at U.S. enterprises and found that 40% of the full sample put ownership of agentic marketing strategy with the Chief AI Officer — 52% among the AI leaders in the sample. Marketing executives were the only cohort leaning toward their own function. Vendor-commissioned research pointing toward marketing feeling under-empowered is exactly the finding a vendor selling to marketing would want, and it’s worth reading it that way. What’s more interesting is that Vose, a former cybersecurity CMO, spent most of her answer arguing the case against her own function.
“Agentic systems create genuine infrastructure, security, and legal exposure that marketing organizations have historically not been resourced or incentivized to own … A marketing team that insists on total ownership without partnering on those guardrails will ultimately fail the brand by exposing the enterprise to something like a massive data breach.”
The split she lands on is that domain-specific calls belong to marketing, operating inside a governance framework the CAIO and the data organization set. A discount landing the week of a competitor’s product recall isn’t a model-risk question. It’s a brand judgment made under time pressure by someone who knows the category. What a CMO loses if the whole territory settles on the Chief AI Officer isn’t the technology — it’s the muscle. Marketing becomes, as Vose describes it, “a requirements-writer instead of a decision-maker,” still accountable for pipeline and brand health while another function controls the mechanism producing both. Authority without accountability is a familiar org-chart failure, and it tends to cost you the people you least want to lose. Nobody good stays to write specifications for someone else’s system.
So what actually changes six months in? Vose points first at budget, and specifically at timing rather than amount — pulling dollars out of a channel that’s still hitting its target but showing a visibly weaker incremental return than it did a month ago. A quarterly attribution model reports that after the quarter’s already spent. Campaign and asset volume counts quietly stop earning a slide. The weekly cross-functional sync whose real purpose was reconciling five dashboards before anyone was allowed to decide anything becomes an exceptions review, on whatever cadence the exceptions warrant. On her own team at Kana, the change she’d tell a peer to copy has nothing to do with software: fewer static reporting decks defended in a room, and messaging pressure-tested against real objections continuously instead of locked at the start of a quarter.
Which brings us back to the box on the diagram. Vose’s own test for whether any of this worked is unglamorous and hard to fake — ninety days in, name what you turned off. A meeting that stopped happening, a license that got revoked, a line of spend that got cut. Put it in the contract discussion now, before signature, while the vendor still wants the deal. If the honest answer at day ninety is that you’ve got a new tab open and nothing else changed, you bought the sixteenth tool.
