A brand can buy an agentic capability tomorrow. Yahoo, PubMatic, Amazon, Google, The Trade Desk and a dozen others now sell versions of it, integrated into the existing stack and live in production. The capability is available. What the capability does to the brand depends almost entirely on whether the brand is ready to be optimised against by a machine, and the honest answer for most brands writing 2026 budgets is that they are not.
"Not ready" is not a competence judgement. It is a structural observation. Agentic systems operate on different inputs from the human teams they are increasingly replacing in the operational layer. They make decisions faster, on more transactions, against criteria a brand has not had to articulate cleanly because a human buyer absorbed the ambiguity. What used to live in the head of an experienced media buyer (the soft judgements about brand fit, risk tolerance, channel mix, creative variability) now has to live somewhere the agent can read.
For most brands, it does not. The counterintuitive part is that this is more useful to know than the question most brands are currently asking, which is whether to deploy agentic capabilities at all. The deployment question has a vendor answer. The readiness question has a brand answer. The first decides what the brand buys. The second decides what the agent does with what was bought, and the second number is larger.
The Readiness Confusion
Most agentic readiness conversations happen at the wrong layer.
The vendor-led version is technical. Are your data pipes connected? Do your DSPs and SSPs support the latest protocols? Are your audience segments available in agent-readable formats? Have you adopted MCP, or AdCP, or ARTF, or whichever acronym the vendor has stake in?
These are real questions. They are also, mostly, vendor problems. A brand can answer them with a procurement cycle and a check. They do not, in any meaningful way, change whether the brand is ready to be sold by an agent.
The harder version, the one most brands are deferring, sits underneath the tech stack. It is about whether the brand itself is legible to a machine in the way it needs to be when a human buyer is no longer translating between the brief and the buy.
Three dimensions matter. Most brands are weak in at least two.
Brand Decision Logic, Articulated
A media buyer with five years on the account knows things about the brand that have never been written down. They know the CMO hates anything that looks like clickbait, even when the data shows it converts. They know last year's connected-TV experiment underdelivered because the inventory mix tilted toward late-night programming the brand was uncomfortable with. They know the founder personally vetoed a placement on a competitor-adjacent site once and the policy was never formalised. They know the brand says "we are a premium player" in the brief and then routinely pushes for tier-three inventory at quarter end.
None of this is in the campaign brief. All of it is in the buyer's head. All of it is exactly the kind of judgement an agent does not have unless it has been written down in a form the agent can read.
The first test of readiness is mechanical. Can the brand produce, in fewer than thirty pages, the criteria that actually govern its media decisions? Not the formal positioning. The operational rules. What the brand will and will not run next to. What it values and what it punishes. What its risk tolerance looks like in a specific scenario. What the failure mode of "wrong inventory" actually costs.
Most brands cannot. Most brands have never had to. The agency absorbed the ambiguity, and the cost of absorbing it was hidden inside the agency retainer.
When an agent is making the decisions, the ambiguity becomes a default setting. If the brand has not specified its rules, the platform's defaults apply. The platform's defaults are not aligned with the brand's interests. They are aligned with the platform's optimisation function, which is some combination of inventory clearance and platform revenue. The cost of being unready in this dimension is paid in inventory the agent buys that a human buyer would have rejected.
Measurement Infrastructure Trustable by Something Other Than You
The second dimension is harder for brands to evaluate honestly because it sits between marketing and finance.
An agent does not optimise against what the brand believes happened in a campaign. It optimises against what the data says happened. The integrity of the underlying measurement layer becomes the constraint on the agent's intelligence.
In practice, most brand measurement infrastructure is a layered set of partial reports stitched together with attribution models built for a previous era. Platform self-reports. Server-side tracking. Third-party verification, sometimes. MRC-accredited measurement, sometimes. An attribution model that allocates credit across channels according to assumptions that were defensible in 2018 and have been worked around ever since.
For a human buyer, this is workable. The buyer triangulates, applies judgement, knows where the numbers are softer than they look. For an agent, it is dangerous. The agent will optimise against whatever signal is strongest, regardless of whether the signal reflects what actually happened or what the platform reporting the signal wants the brand to see.
The readiness question here is concrete. Can the brand independently verify the key inputs the agent will be optimising against, against a source of truth that neither the platform nor the brand controls? For most brands the question is uncomfortable because the answer is no, and the path to yes runs through either a meaningful change in measurement infrastructure or a meaningful change in the supply path where the data originates.
The cost of being unready in this dimension is paid in budget the agent routes toward whatever the platform is most willing to report well on. Which is usually whatever the platform makes the most margin on. Which is usually not what the brand most wants.
Brand Story Stable Enough to Survive Velocity
This is the dimension most brands underweight, because it sounds like a comms problem rather than a strategy problem.
When campaigns ran on weekly creative cycles, brand inconsistency was absorbed by the planning rhythm. There was time for the strategy team to review, the creative team to align, the legal team to sign off. The agency layer normalised the output. The brand showed up coherently because the human process produced coherence.
In an agentic environment, the production cycle compresses to hours and the deployment cycle to seconds. PubMatic's early data showed an 87% reduction in campaign setup time. Amazon's Full Funnel Campaigns can launch multi-format buys from a single prompt. Creative variation, audience targeting, channel selection and bid logic all move faster than the brand's traditional review processes can keep up with.
Brands with stable, well-documented identity systems (clear voice guidelines, defined visual ranges, articulated audience priorities, scenario-tested risk positions) can run at agentic speed because the agent is operating inside a corridor the brand has already defined. Brands without that infrastructure run at agentic speed by accident, and the cumulative drift of small inconsistencies across thousands of micro-decisions adds up to a brand that means something subtly different at the end of a quarter than it did at the start.
The first signal that this is happening is hard for most brands to detect, because it shows up in AI search visibility before it shows up in conventional metrics. The AI's representation of the brand is a running average of how the brand has been described, by the brand and by others, across the recent corpus the model is grounded on. A brand that has been internally inconsistent across a quarter of agentic deployment surfaces in AI responses with diluted associations, conflicting qualifiers and competitor brands borrowing the language that used to belong to it.
Most brand tracking systems will not detect this for two to four quarters. By then, the position has shifted.
What the Readiness Assessment Actually Looks Like
The three dimensions above can be assessed without buying any new technology, attending any vendor briefing or signing any pilot agreement. What it produces is not a score. It is one page, four sections, the brand's own numbers.
Where the brand stands today. Three lines, one for each dimension. Brand decision logic articulated: yes or no, with the named document. Measurement infrastructure independently verifiable: yes or no, with the named source of truth. Brand story stable enough for agentic velocity: yes or no, with the named identity system in place. Three yeses means deployment. Two yeses means staged deployment. Fewer than two means the readiness work is the project, not the agentic capability.
What each "no" costs. Unwritten decision logic costs inventory the agent buys that a human buyer would have rejected. Unverified measurement costs budget the agent routes toward whatever the platform reports best on, which is usually whatever the platform makes the most margin on. Unstable brand story costs AI visibility share to competitors borrowing language that used to belong to the brand. The costs compound across every transaction the agent processes. In a CTV campaign running through an agentic stack, that is thousands per day.
The gap-closing programme. For each "no" answer, name an owner, a quarter and a definition of done. Decision logic articulated by Q3, owned by brand strategy, finished when the agency can hand a written document to a vendor without further clarification. Measurement infrastructure verifiable by Q4, owned by analytics, finished when the agent reads against a source of truth neither the brand nor the platform controls. Brand story stable by Q1 of next year, owned by brand, finished when the quarterly review cycle is operational. No "soon." No "in progress." Defined quarters or it is not on the page.
The deployment posture. Where the brand is ready, an agentic pilot launches this quarter with a named DSP, a named inventory partner, a named budget and three success metrics. Where the brand is not ready, the readiness programme runs first and the pilot waits.
A useful pattern from the campaign data sits underneath this. The Polestar connected-TV activation Alkimi ran with Nielsen-measured brand uplift produced a 34% increase in sales intent and a 24% lift in brand association with "100% electric". Those are the outputs an agent should be optimising against. They are also the outputs the agent cannot read unless the measurement infrastructure dimension is satisfied, because Nielsen-measured uplift is not in any platform's default report.
The Cost of Acting Anyway
Some brands will read the above and conclude, reasonably, that the path forward is to wait. Build readiness. Resource the gaps. Deploy agentic capabilities when the brand is ready to host them.
The conclusion is partially correct. The reason it is only partially correct is that competitors who are also unready will not all reach the same conclusion. Some will deploy anyway. The brands deploying agentic capabilities against an unready brand pay the costs described above, but they also accumulate operational learning that the cautious brands do not. They discover where their decision logic is unwritten because the agent's defaults produce outputs that surface the gap. They discover where their measurement is weak because the agent's optimisation rewards the wrong signals visibly. They discover where their brand story is unstable because the AI representations drift visibly within a quarter.
The cautious brands know less. The aggressive unready brands learn more, at a real cost to the brand in the meantime.
The defensible posture for most brands sits between the two. Run the readiness assessment honestly. Resource the gaps where they are critical. Deploy agentic capabilities in the dimensions where the brand is ready, in pilots small enough to learn from without committing the full media budget. The absence of an agentic deployment is not an excuse for the absence of the readiness work itself.
The infrastructure question, whether the agent can operate on a verifiable record of what is happening across the campaign, is separate from the brand readiness question. The agent is only as good as both the brand briefing it and the substrate it is reading. The two halves of the problem are independent in their solution and joint in their effect.
What Most Brands Will Find When They Look
Few brands will run the assessment and come back with three yeses. That is not a problem. The problem starts when a brand deploys agentic capabilities against three nos and treats the disappointing campaign data as a vendor failure rather than a readiness failure.
The brands that walk into 2027 with the readiness work already done will have paid the cost on their own timetable. The brands that walk in without it will pay it anyway, in real time, against an agent that has already started routing their budget.
---
Related reading from Alkimi Research
What Your Clients Are Actually Asking When They Ask About Agentic. The agency-side complement to the brand-readiness conversation above.
Map your readiness across the three dimensions
The Alkimi research team can supply the readiness assessment template as a structured document, including the prompt set used in the AI brand-visibility check that maps to the brand-story dimension. lauren@alkimi.org or marco@alkimi.org.