A data-driven analysis of how agentic architectures, protocols and AI skills are entering programmatic advertising, based on live AdTech hiring signals.
The question is no longer only whether AI will automate AdTech.
It is how an AI system will connect to buying, selling, data, measurement and execution—and which layers hiring teams are already investing in.
Today, a person configures targeting, budgets, bids and deals in a platform UI. In an agentic scenario, intent could pass through an AI agent and a protocol or governance layer before reaching the same programmatic execution machinery. Hiring signals do not prove that architecture, but they reveal where organizations are placing technical bets.
Live market signal
All headline numbers are computed from currently published, non-expired jobs. Filters change the detailed analyses, while the market-size headline remains anchored to the full dataset.
Active jobs analyzed
3,733
89 companies
Agentic AdTech jobs
?348
AI or agentic signal + advertising context
Share of active jobs
?9.3%
Full unfiltered dataset
Companies hiring
?60
At least one classified role
Distinct detected signals
?30
Agentic, protocol and AI infrastructure
Source: AdTechTalent jobs database · Updated August 20, 2026
Architecture scenarios
These diagrams are technology scenarios. The measured hiring evidence begins in the following section.
Architecture A
A general assistant becomes the frontend while MCP exposes tools and context from specialized platforms.
Architecture B
AAMP could carry advertising semantics, trust and buyer–seller workflows while coexisting with a general agent protocol.
Architecture C
Large platforms can keep differentiated intelligence and workflows inside their own product boundaries while selectively exposing interfaces.
Protocol and technology mentions
Counts represent active classified jobs whose title, description, requirements, responsibilities, tags or tech stack explicitly mention the term. Zero-value categories are hidden.
Source: AdTechTalent jobs database · Updated August 20, 2026
What the data says now: MCP currently appears far more often than AAMP in active AdTech hiring (13 jobs versus 1), although both remain smaller than broader agentic and AI-infrastructure signals.
Company adoption
Only companies with at least 5 active jobs are eligible. Hiring rate equals classified Agentic AdTech jobs divided by all active jobs at that company.
| Company | Agentic jobs | Hiring rate | AI jobs | All jobs | Top signal | Top protocol |
|---|---|---|---|---|---|---|
| Yahoo DSP | 5 | 100% | 5 | 5 | Gemini | OpenRTB |
| Kayzen | 10 | 52.6% | 10 | 19 | Claude | — |
| OpenAI | 13 | 39.4% | 13 | 33 | OpenAI | — |
| PubMatic | 27 | 37% | 27 | 73 | Claude | OpenRTB |
| Tatari | 24 | 36.9% | 24 | 65 | Claude | — |
| Sovrn | 3 | 33.3% | 3 | 9 | Agentic | OpenRTB |
| Klever | 2 | 33.3% | 2 | 6 | AI Agents | A2A |
| RTB House | 38 | 31.4% | 38 | 121 | Agentic | — |
| BidMachine | 3 | 30% | 3 | 10 | Claude | — |
| VideoAmp | 5 | 29.4% | 5 | 17 | Agentic | MCP |
| Moloco | 12 | 24.5% | 12 | 49 | LLMs | — |
| Appodeal, Inc. | 3 | 23.1% | 3 | 13 | — | — |
| Simulmedia | 2 | 22.2% | 2 | 9 | AI evaluation | — |
| Smadex | 10 | 20.4% | 10 | 49 | LLMs | MCP |
| 7 | 18.9% | 7 | 37 | Agentic | — |
Source: AdTechTalent jobs database · Updated August 20, 2026
Workforce formation
Similar titles are grouped into deterministic role families. The residual “Other” family is shown rather than forcing ambiguous titles into a named AI role.
| Role family | Jobs | Companies | Median seniority | Remote | Top signals |
|---|---|---|---|---|---|
| Other agentic role | 212 | 44 | Manager | 32.5% | Claude, Agentic, LLMs |
| Software Engineer | 44 | 20 | Senior | 36.4% | LLMs, Agentic, OpenRTB |
| Data Scientist | 31 | 20 | Senior | 25.8% | AI evaluation, Claude, Fine-tuning |
| Product Manager | 29 | 17 | Manager | 0% | LLMs, Agentic, AI evaluation |
| ML Engineer | 20 | 11 | Senior | 40% | LLMs, Agentic, Model serving |
| AI Product Manager | 7 | 5 | Manager | 14.3% | Agentic, AI evaluation, Multi-agent |
| AI Engineer | 4 | 3 | Principal | 50% | LLMs, LangChain, Agentic |
| AI Solutions Engineer | 1 | 1 | Senior | 0% | Agentic, LLMs, OpenAI |
Source: AdTechTalent jobs database · Updated August 20, 2026
83.3% of classified jobs are Senior, Staff, Principal, Lead, Manager or Director+.
Hiring is concentrated in senior roles, consistent with an infrastructure-building phase.
Source: AdTechTalent jobs database · Updated August 20, 2026
Existing AdTechTalent role categories, normalized into comparable groups.
Source: AdTechTalent jobs database · Updated August 20, 2026
Reuses the job board’s remote / hybrid / on-site classification.
Source: AdTechTalent jobs database · Updated August 20, 2026
Architecture + measured signals
The labels describe a possible technical architecture. The counts inside each layer are current hiring signals—not claims about product adoption or technical capability.
AI interface
Top signal: LLMs
122
active jobs
39
companies
Examples in dataset
Epsilon, RTB House, Skai
MCP / A2A
Top signal: MCP
13
active jobs
9
companies
Examples in dataset
The Trade Desk, Amazon Web Services (AWS), Epsilon
AAMP / advertising semantics
Top signal: Agentic
2
active jobs
2
companies
Examples in dataset
Epsilon, PubMatic
Vendor intelligence
Top signal: Agentic
85
active jobs
37
companies
Examples in dataset
Skai, Teads, TripleLift
DSP / SSP / Data / Measurement
Top signal: Claude
143
active jobs
48
companies
Examples in dataset
Epsilon, Skai, Kayzen
OpenRTB / OpenDirect / Deals
Top signal: OpenRTB
29
active jobs
11
companies
Examples in dataset
Epsilon, Teads, TripleLift
Source: AdTechTalent jobs database · Updated August 20, 2026
Evidence-based conclusion
The hiring market does not support a clean MCP-versus-AAMP winner-take-all story. MCP appears in 13 active classified jobs, A2A in 4, and AAMP in 1. Broader agentic and AI-infrastructure terminology is materially more common than any single interoperability protocol.
The present evidence is most consistent with a layered transition: general AI infrastructure and vendor-specific intelligence are being staffed now; general protocols have an early but measurable footprint; advertising-specific semantics remain nascent in explicit job descriptions. That could lead to general protocols dominating, AAMP becoming a semantics layer, proprietary systems exposing standard interfaces, or all three coexisting. Hiring data cannot yet distinguish decisively among those four scenarios.
3,733 active jobs from 89 companies, extracted across 34 countries represented in classified job locations. Data updated August 20, 2026.
Case-aware regular expressions inspect title, description, summary, requirements, responsibilities, tags and normalized tech stack. A classified job needs both an AI/agentic signal and an advertising/programmatic signal. Existing category, seniority and workplace fields are reused, then normalized for this report.
Descriptions may not expose internal technology. Companies may use protocols without hiring for them. Named protocols can be rare even when similar concepts are used. Hiring indicates organizational investment, not adoption, product quality or headcount. One posting may represent multiple openings.
Continue exploring
Work on the agentic AdTech stack
Browse engineering, product, data and programmatic opportunities at companies building the next generation of advertising technology.