The story. TIME is using two AI agents from the startup Kana: one reads RFPs and drafts proposals, the other watches live campaigns and recommends fixes. Its COO says sellers now spend their time selling, not coordinating. (AdExchanger, 9 November 2026)
01What happened
- Two agents, two jobs. One works with the sales team, reading RFPs and drafting proposals. The other works with ad ops and recommends changes to trafficking, pacing or budget when a campaign falls behind.
- They read from three systems. Kana builds a shared “knowledge graph” from TIME’s Salesforce CRM, its order management system and its ad server, direct and programmatic, plus data from live campaigns.
- The old process was the bottleneck. Mark Howard, TIME’s chief operating officer, described RFPs setting off Slack alerts, Google Meets with product experts and days of manual stitching. “Sellers lose too much time playing coordinator,” he said.
- The relationship is a hard line, with a caveat. “We don’t do it with our journalism, and we’re not going to do it with our sales,” Howard said, then added that this “isn’t the plan for now.”
- The buyer-agent door is technically open. Kana co-founder Vivek Vaidya says an MCP server in front of the proposal agent would let it talk to an advertiser’s agents. TIME is not planning to do that.
Kana came out of stealth in February with a $15 million seed round, so this is an early product with one named publisher customer in the piece.
02What it means inside a GAM network
The useful detail is the data source, not the agent. A proposal agent is only as good as the order history, the delivery data and the inventory definitions it can read. For a GAM publisher that means line item naming, ad unit structure and the CRM fields reps actually fill in. If those are inconsistent, an agent will draft confident proposals from inconsistent inputs.
The second point is where the time goes. Howard’s complaint is days of coordination across tools, not a lack of sales skill. Most of that work sits between the CRM, the order system and ad ops, which is where a small team loses the most hours and has the least slack.
The pacing agent targets a cost every direct seller knows: a shortfall that turns into a make-good. Ad ops already has dashboards that say a campaign is behind. The new part is a recommendation for what to change. Whether those recommendations are right is not shown in the article, so treat them as claims to test against your own delivery.
The third point is the one TIME chose not to cross. Howard drew the line at the relationship, and his “for now” shows how provisional it is. Vaidya calls connecting to buyer agents “a technology problem, which is easy to solve.” The decision that remains is a policy one, and each publisher should make it before a buyer asks.
TIME is showing one defensible split: agents take the coordination, people keep the buyer.
03What publishers should do about it
04The bottom line
TIME is showing one defensible split: agents take the coordination, people keep the buyer. The next question is who owns the rules that keep that split in place. Publishers with clean order and delivery data can try the same split on their own terms.
If you want a second pair of eyes on your direct sales set-up, see our publisher services. A related read on what happens when buyers’ automation meets publisher supply: AI Buyers Skip 86% of the Auctions.