Adform and Vertoz Put MCP Into Production, Turning AI Assistants Into Programmatic Interfaces
Executive Summary
Programmatic platforms can now be operated through ChatGPT, Claude, Gemini and Perplexity over an open standard, and Omnicom Media and Publicis already run Adform's MCP Server in production. The proprietary dashboard stops being a moat, and trader hours spent on pulls and pacing become compressible. Run a 30-day read-only pilot before year-end. If savings exceed 15 percent, write MCP support into every Q1 2027 vendor RFP.
The Signal
Two ad tech vendors moved Model Context Protocol (MCP) from concept to live product on October 28, 2026. Adform announced results from the live rollout of its MCP Server, reporting that Omnicom Media and Publicis are using it to change how programmatic campaigns are managed and optimised. The source did not publish the underlying performance metrics. Separately, Vertoz (NSE: VERTOZ) launched MCP Connectors globally for two platforms: IngeniousPlex, its performance media buying platform, and IncrementX, its publisher monetization platform. The connectors work with ChatGPT, Claude, Gemini, and Perplexity, using OAuth authentication and existing account permissions. At launch, Vertoz's connectors are limited to retrieving campaign, account, inventory, and monetization data through natural-language queries. They do not execute campaigns or trade media. Both run alongside existing dashboards.
What Changed
Programmatic platforms can now be queried and operated through general-purpose AI assistants over an open standard, without custom API integrations or dashboard navigation. Adform's server is already in production use at two of the largest holding-company agency networks. Vertoz extends this to four major assistants with permission-governed access. Campaign and inventory data becomes retrievable in plain language, and the same interface layer that enables retrieval can support autonomous execution once vendors enable write access.
Why It Matters
The sharpest implication: the dashboard stops being the moat. For fifteen years, ad tech vendors defended retention with proprietary interfaces and the training cost of mastering them. MCP converts that interface into a commodity. When Claude or ChatGPT can pull campaign and inventory data in plain language, the vendor's UI becomes an optional layer, and switching costs built on user familiarity start to erode. The commercial gains accrue first to agency holding companies. Omnicom Media and Publicis running Adform's server in production means analyst hours spent on pulls, pacing checks, and ad hoc reporting are now compressible. That work is a large share of traders' time and a quiet source of agency margin. Whoever redeploys it into optimisation rather than pocketing the saving gains a performance edge over agencies still hiring headcount to navigate dashboards. Devalued: bespoke API integration work, BI-layer middleware whose job was reconciling platform data, and platform-specific certification as a hiring signal. The why behind the why is vendor positioning. Adform and Vertoz are not being generous; they are hedging against disintermediation. If assistants become the interface, platforms that expose clean MCP endpoints stay in the workflow, and those that don't get routed around. Vertoz's choice to launch read-only across four assistants is a land grab for default status. Write access follows once permission governance proves out. The new competitive move is data gravity. Whichever assistant a team standardises on becomes the control plane, and the vendors beneath it compete on the quality and breadth of what they expose. Platforms that treat MCP as a checkbox will find themselves ranked by how well an agent can use them.
Marketing Impact
media
Traders and planners can pull pacing, spend and inventory data in plain language instead of navigating dashboards or waiting on analyst reports. Routine checks compress from hours to minutes. Performance advantage goes to teams that redeploy the saved time into optimisation, not to those who simply pocket the efficiency.
martech
Platform selection criteria change. Stack evaluations must now score how well an agent can query and eventually operate a tool, not only how usable its interface is. MCP endpoint quality, OAuth permission granularity and cross-assistant compatibility become procurement line items. Vendors with weak endpoints get routed around in agent-led workflows.
marketing ops
Bespoke API integration projects and BI middleware built to reconcile platform data lose their business case when assistants query platforms directly. Ops teams must decide which assistant becomes the control plane, then govern access, since vendor permissions now define what any agent can see across the stack.
research
Ad hoc analysis that once required a ticket to an analyst, such as inventory availability or campaign comparisons, becomes self-serve for strategists and researchers. Read-only access keeps risk low. The constraint shifts to data literacy and validating assistant-generated answers against source dashboards.
The Exploit
Opportunity
Agencies and in-house media teams can compress the manual layer of programmatic operations: data pulls, pacing checks, ad hoc reporting. Connect an assistant to one platform's MCP endpoint and redeploy freed trader hours into optimisation. Teams on Adform or the Vertoz platforms can pilot now, measure hours reclaimed in a single quarter, and reset staffing and margin models before 2027 budgets lock.
Risk
Read-only access limits early upside, and natural-language queries can return confidently wrong numbers. Granting assistants account permissions creates data-governance exposure. Betting on one assistant as control plane risks lock-in to a layer that is still shifting.
The Move
Head of media operations should run a 30-day pilot before year-end: connect one assistant to one MCP-enabled platform under read-only permissions, then benchmark hours spent on pulls and pacing checks against a control team. If measured savings exceed 15 percent, write MCP support into every Q1 2027 vendor RFP.
First-Mover Advantage
Gains
Teams that standardise early build prompt libraries, permission structures and workflow habits while rivals still navigate dashboards. They also gain leverage in vendor negotiations, with proof of which platforms perform best when operated by an agent.
Risks
Read-only access limits early upside, and natural-language queries can return confidently wrong numbers. Granting assistants account permissions creates data-governance exposure. Betting on one assistant as control plane risks lock-in to a layer that is still shifting.
Window
Roughly six to nine months, until write access ships and MCP endpoints become standard procurement requirements. The window closes when holding companies publish efficiency gains and clients demand the same fee reductions.
Winners & Losers
Winners↑
Holding-company trading desks running MCP in production
Omnicom Media and Publicis are already live on Adform's MCP Server, which compresses the analyst hours spent on data pulls, pacing checks and ad hoc reporting, a large and quiet share of trader time. The advantage accrues to whoever redeploys that capacity into optimisation and client strategy rather than banking it as margin. Desks that move first should rewrite trader job design now, because the performance gap over dashboard-bound competitors opens before the savings show up on a P&L.
Ad platforms with clean, well-governed MCP endpoints
Vendors like Adform and Vertoz keep their place in the workflow even if the assistant becomes the interface, because agents route to platforms they can query reliably. Vertoz's read-only launch across ChatGPT, Claude, Gemini and Perplexity is a deliberate bid for default status before write access arrives. Platforms that ship breadth and quality of exposed data, with OAuth permissioning that security teams will approve, convert agent-friendliness into a selection criterion at RFP.
Assistant-native media and analytics teams
Teams that standardise on a general-purpose assistant as their control plane gain plain-language access to campaign and inventory data across multiple platforms without waiting on BI tickets or custom API work. The mechanism is speed of question-to-answer, which compounds in pacing decisions and mid-flight reallocation. Move now to define permission scopes and approved query patterns, so that when vendors enable write access the governance is already in place and execution follows without a rebuild.
Publishers using AI-accessible monetization stacks
Vertoz's IncrementX connector lets publishers query inventory and monetization data conversationally, which lowers the analyst overhead of yield and sell-through reviews for lean ad ops teams. Publishers that expose their data through MCP also become easier for buy-side agents to evaluate and transact with later. Prioritise partners with MCP endpoints in stack decisions this quarter, since agent-readable inventory is likely to be favoured as agentic buying matures.
Losers↓
Ad tech vendors whose retention rests on proprietary dashboards
For fifteen years, interface familiarity and the training cost of mastering a UI have defended switching costs. When an assistant can pull the same data in plain language, the dashboard becomes an optional layer and that lock-in erodes. Vendors need to re-found their moat on data quality, execution performance and endpoint reliability, because platforms that treat MCP as a checkbox will be ranked by how well an agent can use them.
BI-layer middleware and bespoke API integration shops
A large part of this segment's value is reconciling platform data and building custom connectors so humans can query it. An open standard with first-party vendor connectors removes much of that work, and permission-governed access arrives without a services engagement. Survivors will need to move up-stack into cross-platform measurement, governance tooling and agent orchestration rather than selling data plumbing.
Agencies staffing for dashboard navigation and manual reporting
Agencies that scale headcount to handle pulls, pacing checks and reporting carry a cost structure that holding companies running MCP in production are actively compressing. The pressure shows up in pitch pricing and in talent expectations, as platform-specific certification loses value as a hiring signal. Defensive moves are to retrain traders toward optimisation and analysis, and to renegotiate scope with clients before competitors reprice the same service.
Strategic Outlook
Expect write access to follow read access within two to three quarters. Vertoz's read-only launch is a land grab for default status. Once permission governance holds up in production, execution endpoints for budget shifts, bid changes and pause/resume actions follow. Adform's live use at Omnicom Media and Publicis sets the bar competitors must match in Q1 2027 RFPs, where MCP support will appear as a standard requirement. Holding companies will standardise on one or two assistants as control planes, and the DSPs, SSPs and publisher platforms beneath them will compete on endpoint depth and reliability. Laggards will not lose on features; they will lose by being invisible to the agents making recommendations. The first significant incident of an assistant executing an unauthorised or erroneous change will force a governance standard on permissions and audit trails before mid-2027.