The programmatic stack is not going away. Billions of impressions trade through it every day. The infrastructure is well-established, the integrations are in place, and the major platforms have invested heavily in making it work at scale. The case for agent-to-agent trading is not a case for dismantling that infrastructure.

It is a case for adding something the existing stack was never designed to provide: a negotiation and governance layer that enables bilateral deal agreements between agents who may not have a prior integration and who need a shared record of what they committed to.

What was the programmatic stack designed to do, and what was it not?

The programmatic stack was designed to match buyers and sellers at scale through automated auction clearing. It is very good at this. An impression becomes available; the DSP evaluates it against targeting criteria and a bid; the SSP clears the auction; the creative serves. The whole cycle takes milliseconds. At tens of billions of impressions per day, the stack performs the task it was designed for with high efficiency.

What the stack was not designed to do is handle bilateral deal negotiation. It has no mechanism for a buyer and seller to exchange deal proposals, agree on terms that go beyond price and targeting, create a shared record of those terms, or manage compliance against that record after delivery. These functions simply do not exist in the programmatic infrastructure. They were handled, when they were handled at all, through separate direct-sales processes that ran alongside the programmatic stack rather than through it.

What specifically needs to be added to support agent-to-agent trading?

Three things that the current stack does not provide.

A negotiation protocol that allows agents on both sides to exchange structured deal proposals without requiring a prior bilateral integration. The protocol needs to be open enough that any agent implementing it can communicate with any other agent that implements the same protocol, independent of the underlying platform.

A shared deal record infrastructure that both parties can write to and read from. The record needs to be bilaterally owned: neither party should be able to amend it unilaterally after the deal is agreed. And it needs to be compatible with the audit requirements of both buyers and sellers, who may have different internal systems for recording commitments.

A human approval workflow that sits at the deal commitment point. Before the buy is executed, both parties' agents need to have routed the deal terms to the humans responsible for approving them. The approval needs to be recorded in the deal record. The entire subsequent delivery and reconciliation process references that approved record.

What stays the same in the programmatic stack?

The delivery infrastructure does not change. Creative trafficking, ad serving, impression counting, viewability measurement, and brand safety verification all continue to operate through the existing stack. Once a deal is agreed through the negotiation layer, the execution of that deal flows through the delivery infrastructure in the same way that any other deal would.

Data partnerships and audience infrastructure also remain relevant. The buyer agent still needs audience data to inform the brief. The seller agent still needs audience data to describe what it can deliver. The data layer does not disappear: it becomes an input to the negotiation.

The auction-based programmatic clearing mechanism remains the right model for commodity inventory. Not every buy warrants the overhead of a bilateral negotiated deal. For high-volume, low-specificity buys where price is the primary variable, programmatic auction clearing will continue to be the efficient choice.

Where does Alkimi fit in the changed stack?

Alkimi's role is the deal layer, not the delivery layer. The DealSheet provides the shared deal record infrastructure. The negotiation protocol enables agent-to-agent deal exchange without prior bilateral integration. The human approval workflow is built into the deal commitment process.

This positioning is deliberate. Alkimi does not intend to replace the programmatic stack. It does not handle ad serving, impression counting, or measurement. It handles the part of media trading that the programmatic stack was never built for: the deal negotiation and governance layer that bilateral agent-to-agent trading requires.

The distinction matters for how Alkimi integrates with the existing market. A buyer or publisher using Alkimi's deal governance layer does not need to replace their DSP or SSP. They need to add the negotiation and deal record layer on top of what they already have. The stack changes by addition, not by substitution.

When does this change happen, and who drives it?

The change is demand-led. Buyer-side adoption of agent infrastructure creates the demand for a negotiation layer that the current stack does not provide. As buyers deploy agents capable of executing bilateral deal negotiations, sellers face pressure to provide infrastructure compatible with that model. Publishers who can participate in agent-to-agent deals will capture the deal flow that buyers with agent infrastructure prefer. Publishers who cannot will find that buyers with agents default to programmatic for their inventory.

The pace of change will be determined by how quickly buyers develop the agent capability to drive demand for the negotiation layer. The infrastructure to support that demand exists. What is needed is enough buyer-side deployment to make the market for direct bilateral deals real rather than theoretical.

The programmatic stack will not look the same in five years as it does today. But it will not have been replaced. It will have had a layer added on top that handles something it was never able to handle: the bilateral commitment between a buyer and a seller, agreed before spend is committed, verified against a shared record after delivery completes.