Inside Wpromote’s Four-Week Sprint With Kargo’s Agentic SSP
Kargo’s Project Kera slipped quietly into closed beta this spring, promising to collapse the friction between biddable and non-biddable supply into a single chat window. The mobile-first SSP is betting that an agentic layer—an AI that can plan, budget, and format on command—will finally let buyers treat premium in-app inventory like a self-serve vending machine instead of a RFP black hole.
Wpromote, the 700-person indie agency out of El Segundo, raised its hand first. It joined Hershey and the travel-media network Navigator as one of only three external testers, clocking four weeks of live campaigns inside Kera’s sandbox. “We didn’t want to wait for the deck to be perfect,” says Skyler McGill, Head of Programmatic & Video. “We wanted to see if the agent could survive our real-world chaos.”
How Kera Fits Inside Wpromote’s “AI Operating System”
Most agencies still treat new tools like shiny objects. Wpromote is trying to orchestrate them. The shop’s internal stack—nicknamed Polaris IQ—already pipes spend across Meta, TikTok, Pinterest, and open-web programmatic. Kera is being wired in as a sub-routine, not a side quest.
“Technology and agents are really about enabling marketers to be more strategic in decision-making and orchestration across systems. It’s less about just moving faster and more about creating a step change in how we operate.”
—Skyler McGill
Inside the beta, that translates into three concrete actions:
- Auto-generated media plans that blend Kargo’s rich-display units, opt-in rewarded video, and conventional open-exchange placements into one budget line.
- Dynamic creative assembly—the agent sizes, headlines, and CTA-tests variants in minutes instead of days.
- Real-time re-allocation when viewability or attention KPIs drift, something the team previously handled in twice-weekly optimization sprints.
Early returns are workflow, not performance. Brief-to-live cycles that averaged 10–12 days in Q1 are now clocking under 72 hours for Kargo inventory. “Measurable lift? Ask us after the next refresh,” McGill laughs. “But the brief-to-campaign compression alone pays for the pilot.”
The Multi-Agent Playbook: Kera, PubMatic, Yahoo, Amazon
Wpromote’s philosophy is poly-agent. While Kera gets the headlines, the agency is simultaneously:
- Piloting PubMatic’s upcoming agentic console, focused on connected-TV supply-path optimization.
- Monitoring Yahoo’s DSP agents that rewrite private-marketplace deals on the fly.
- Watching Amazon’s rumored “Austin” project, which is expected to let marketers brief campaigns in plain English and receive back a ready-to-flight media plan that spans Amazon-owned and third-party inventory.
“It really comes down to where you can drive the best outcomes and reach the right audiences. You’ll have a lot of optionality, and the systems will need to connect.”
—Skyler McGill
That last clause—“the systems will need to connect”—is the hidden tax most vendors ignore. Wpromote is building a universal ingestion layer that normalizes objectives, KPIs, and creative assets so the same brief can be pushed to Kera, PubMatic, or Yahoo without re-keying. Think of it as the Switzerland of agentic SSPs, an approach that could insulate the agency if one platform wins or disappears.
Guardrails, Not Gatekeepers
The specter of runaway automation keeps McGill’s team up at night. Kera’s first draft once allocated 40 % of spend to a Kargo unit that delivered stellar CTR but hadn’t cleared the client’s viewability threshold. The fix: hard-coded governance rules inside Polaris IQ—minimum viewability, brand-safety floor, and spend-cap per format.
“We’re being really mindful of overreliance on automation. You still need the right data inputs, governance and QA processes in place to understand what’s actually happening.”
—Skyler McGill
Weekly “human-in-the-loop” reviews pit media, creative, and data science leads against the agent’s choices. Discrepancies above 5 % trigger a retro where the team retrains the model or rewrites the guardrails. It’s a living compliance layer that looks a lot like the Ad-Tech Compliance Guillotine: 5 Rules That Decide Who Survives—only automated.
Creative at Machine Speed
The most tantalizing promise is scalable creative iteration. Kera ingests a single 15-second product video and spits out:
- Six vertical snippets for Kargo’s fullscreen “Flip” unit.
- Three interactive end-cards with shoppable hotspots.
- A rewarded-video companion overlay localized for DMA-level weather triggers.
That’s work that used to eat three days of studio time. Now it’s rendered overnight and A/B-tested while the account team sleeps. McGill calls it “operationalizing a campaign brief”—a shift from human-driven production to machine-driven permutation.
“What’s really exciting is the ability to operationalize a campaign brief and build creativity at scale, then optimize in real time. That’s a big shift from the manual processes we’ve had.”
—Skyler McGill
If this feels familiar, it’s because packaged-goods upstart Nutpods pulled a similar stunt last year—using AI to spin thousands of audio variants and beat walled-garden fragmentation. Wpromote is extending the logic to visual and interactive formats, betting that creative diversity will matter more than ever when cookies finally go dark.
What Happens When the Dust Settles?
Harry Kargman, Kargo’s CEO, won’t commit to a public launch date. The roadmap calls for identity bridging (Kera will soon ingest UID 2.0, RampID, and first-party brand data) and outcome-based buying where the agent guarantees CPA or ROAS, not just media delivery. If that sounds like the performance promises DSPs made in 2015, you’re not wrong—except this time the inventory is premium, the supply path is transparent, and the agent is doing the math in real time.
For Wpromote, the endgame is consolidation without lock-in. If Kera, PubMatic, and Yahoo all hit their stride, the agency could collapse six point-solution platforms into one orchestration layer, saving 15 % in tech fees and clawing back 200 basis points in working media. The risk: agents become commodity arbitrageurs, and the only moat left is who writes the better prompt.
McGill is comfortable with that future—as long as the prompts stay human.
💡 Deep Dive: Don’t miss our Ultimate Industry Guide for advanced strategies.
