Dept doesn’t want to sell marketers AI tools. It wants to sell them growth.
That’s the pitch, anyway, and it comes with a deliberate sacrifice: the agency won’t charge for the tech layer of its new AI system at all, only for the outcomes that come out the other end.
The system is called Deptify, but the way people interact with it is through a persistent assistant called D, where Dept’s bet on orchestration over an operating system plays out day to day. Instead of having to learn a new interface, or wait while an agency rebuilds every partner tool to fit inside one of its own, D pulls context from whatever workflow tools the client already runs, whether that’s Adobe Workfront, Asana or something else entirely, and routes a person to the right one for the task at hand.
So far Dept says three, unnamed clients have said they’re ready to use it, with another agreeing to do so following a pitch. Others will get to test this out themselves this week at the Cannes Lions Festival of Creativity. But the tool is far from embryonic. In fact, it has been deployed across clients representing around 20% of the agency’s global revenue, and the company expects to exceed 80% by the first quarter of 2027. All new business wins are now onboarded into Deptify from the start, meaning every new client relationship starts inside the system Dept is still rolling out to the rest of its roster.
In a demo, Roy Armale, Dept’s chief product officer, asked D what he needed to deliver that day. D checked the work management tool the team uses and identified the task and the tool needed for it. Telling D which task he wanted to work on, rather than naming a client directly, was enough to trigger a switch. D identified the client itself, in this case one of its largest, and told him so directly, then cleared away what it knew about every other client he works with, so none of that other context could leak into the account. What D retained was its own history with the demo client specifically, including a campaign he had been working on called the spring sale.
Armale compared the effect to the character shifts in the Spider-Man “Spider-Verse” series, where each version of D keeps learning within its own client relationship over time but the different versions are never meant to meet or share what they know.
“This is a multiversal assistant,” he explained. “Every time we go to a new universe within the multiverse, it is a new assistant.”
That persistence is also what Armale said lets D do more than retrieve information. Showing a report from a previous engagement, he described being able to see what a viewer would and wouldn’t focus on in a piece of creative, including that it lacked an explicit call to action. He said this kind of check is what allows Dept to charge for an asset at all: an asset only qualifies for output-based billing once it has passed this kind of effectiveness test, run through a third-party tool called Optimal rather than something Dept owns itself, so that, as he put it, Dept isn’t “grading its own homework.”
The decision to build D this way, rather than as part of a single branded interface, is what Armale says shaped Deptify’s economics overall. Agencies building AI systems have largely taken one of two approaches, he said: build an interface and require every partner tool to redeploy inside it, which the industry calls an operating system, or connect existing tools to each other using the same underlying technology, MCP, without absorbing them, which Armale calls an orchestration system. He said he ran into the first approach directly with Adobe over Workfront. “You want to run Workfront from your operating system?” he said Adobe asked him. “Workfront is an operating system,” he said he told them, describing the disagreement that followed.
He’s not speaking hypothetically. Before joining Dept, he was chief product officer on WPP Open, the operating system he’s now describing the limits of.
**Connect don’t conquer **
Building Deptify as an orchestration layer, Armale said, means Dept can’t charge for that layer itself, since a client won’t pay separately for both the underlying tools and a system connecting them.
That revenue goes to the partner tools, not to Dept. He linked the tradeoff to Dept’s size, contrasting it with what he described as a $15 billion company carrying high opportunity costs relative to revenue. Dept, he said, is lean by comparison, with himself personally client-facing, client-billable and building the product at the same time — a combination he said a larger, less lean organization charging for growth the same way would struggle to sustain without its margins suffering.
What Dept charges for instead is split into three tiers: input, output and outcome. Input is billed for time and materials. Armale compared it to the way Uber prices Black above standard, not because the ride costs more but because the driver is rated and trained to a higher standard.
Dept applies the same logic to a person working with D, charging for an “augmented human” rather than time alone. He said this is the easiest tier to charge for and the one least connected to outcomes. Output is the asset-based tier described above. Outcome, tied to growth numbers, sits at the top and is currently treated as a bonus rather than a primary fee. Armale said the balance between the tiers shifts as a client relationship deepens. The more end-to-end control Dept has over a process, the more it’s willing to be paid against outcomes rather than time or assets.
Token costs are not passed on to clients under any of the three tiers, according to Armale, who described this as a deliberate choice rather than a temporary one.
“If I pass token costs on, I have removed the value of the person using the token completely,” he said, arguing that billing separately for compute turns a client conversation toward technology costs and away from growth, the opposite of what Dept is trying to sell.
More in Marketing
[
WTF is containerization? ](https://digiday.com/marketing/wtf-is-containerization/)
Containerization is suddenly everywhere in ad tech conversations.
[
The AI search boom is fueling content budgets before media buys ](https://digiday.com/marketing/the-ai-search-boom-is-fueling-content-budgets-before-media-buys/)
As brands chase visibility in AI search, many are reallocating SEO, content and creator budgets before investing in AI search advertising.
[
Cannes Briefing: The ad industry’s power shifts, mapped in sand ](https://digiday.com/marketing/cannes-briefing-the-ad-industrys-power-shifts-mapped-in-sand/)
An investment bank walks into Cannes. A holdco walks out.