AdvertisingProgrammatic & Display
Chalice AI Injects Custom Models Into DSPs and SSPs
Chalice AI's Ali Manning says the startup lets brands run proprietary models inside walled gardens, DSPs and SSPs, chasing outcomes the CFO cares about instead of clicks and viewability.
Key takeaways
- Chalice AI lets brands deploy their own models directly inside walled gardens, DSPs and SSPs rather than relying on platform-default bidding signals.
- COO Ali Manning says most DSPs still optimize for easy-to-measure proxies like clicks, conversions and viewability instead of brand-specific business outcomes.
- Bayer's One A Day brand, where over 80% of sales happen in stores, used Chalice to redirect its Amazon DSP toward offline 'new to brand' shoppers.
- Manning is steering away from the term 'custom algorithm' after holding companies grew wary of the category following questionable practices from some rival providers.
- Chalice points to agentic bidding and containerization as the next phase for making programmatic optimize for real business outcomes.
The News
Chalice AI, a startup run by COO and co-founder Ali Manning, is pitching a direct fix for one of programmatic's oldest complaints: DSPs that optimize for whatever is easiest to measure rather than what actually moves a brand's business. Manning laid out the mechanics on the August 4, 2026 episode of AdExchanger Talks, explaining that Chalice lets brands run their own proprietary models inside walled gardens, DSPs and SSPs instead of accepting platform-default bidding logic.
Manning calls the pitch "models built for me." Most DSPs, she argues, chase clicks, conversions and viewability because those are simple to track, even when they're weak proxies for what a CMO or CFO actually cares about.
Why It Matters: The Bayer Case
Bayer's One A Day vitamin brand is Manning's proof point. Its Amazon DSP setup had been tuned for ecommerce and upper-funnel signals like video completion, a mismatch for a brand that sells mostly in physical stores.
Share of One A Day vitamin sales that happen in stores, not online
Ali Manning, Chalice AI, via AdExchanger Talks, August 2026
Chalice paired AI modeling with a panel partner to identify shoppers most likely to be "new to brand" for One A Day, then bid higher against that group inside Amazon DSP. Manning says Bayer has seen a significant lift in new-to-brand shopping as a result, effectively using a retail media DSP to chase an offline sales metric it was never built to prioritize.
What To Watch
Manning points to containerization and agentic bidding as the next step for letting brand-specific models run natively across ad stacks, a direction that echoes what's happening with Google's own agentic ad stack. For CMOs, the practical move is auditing whether your DSP setup is quietly optimizing for the wrong proxy metric, the same gap that shows up across attribution models built for 2026 budgets.
Explore more programmatic and CTV strategy on CMO Mag's Advertising hub.
Advertiser disclosure: some links in our articles are affiliate links, and CMO Mag may earn a commission or referral fee if you sign up or buy through them, at no cost to you. It never affects our editorial coverage. See our advertising & affiliate policy.
More in Advertising
View all
Snapchat Bans AI-Generated Videos From Its Spotlight Feed
Snap says Spotlight's algorithm will no longer surface fully AI-generated videos, betting that Gen Z wants human-made content over synthetic feeds.
Retail Media CPMs in 2026: The Benchmark Table Buyers Need
No retailer publishes a rate card, so we built one from the agency benchmark reports and public filings that do exist, comparing what brands actually pay across the top retail media networks.
PubMatic, Optable Agents Now Monetize Publisher Data Live
PubMatic and Optable have taken agentic ad buying out of the lab: about 30 programmatic campaigns are already running on AI-curated first-party publisher data, and half of them flow through their new joint integration.
Discussion
No comments yet. Be the first to say something worth reading.