The Cross-Functional Readiness Problem: Why Agentic Commerce Has No Single Owner
Which function in your organization owns whether an AI shopping agent recommends your products? If the answer takes more than a beat, that hesitation is the subject of this piece.
A growing share of product research, comparison, and purchasing now happens inside AI agents — ChatGPT, Perplexity, Claude, and the assistants embedded in retailer apps — that assemble a shortlist before a human reviews anything. Adobe Analytics recorded generative-AI-referred traffic to US retail sites growing roughly 690% year over year across the 2025 holiday season (Adobe Analytics, 2026). Most brands have no diagnostic for whether their products clear the bar to appear in that shortlist, and no measurement infrastructure that would catch it when they stop. A brand can be filtered out of consideration with no traffic decline, no campaign failure, nothing obvious to investigate.
The Brand Visibility for Agentic Commerce (BVAC) Framework is a diagnostic for the surface an agent actually reads when it decides. It evaluates a brand across two prerequisites — whether an agent can resolve the brand and product to stable entities, and whether the product attributes it needs are present — and six strategic dimensions sitting above them: whether differentiators exist as queryable fields, whether the brand operates an agent of its own, whether its trust surface is machine-readable, whether its protocol stack functions in machine time, whether data is fresh enough to be acted on, and whether the operating model that maintains all of it is owned. The framework scores each dimension on a five-level scale from Invisible to Agent-Native.
Seven of those eight dimensions describe what a brand exposes to agents. One — Governance Maturity — describes how the brand decides what to expose, who has the authority to change those decisions, and what accountability exists when something goes wrong. That distinction isn't academic. Without governance, the other dimensions degrade silently: protocol implementations drift out of spec, peer-agent verification decisions become inconsistent, the agent's authority scope expands without anyone reviewing it, and incidents go un-investigated because no function owns the post-mortem. The technical work gets done once and then erodes, because nothing maintains it.
The reason this is hard is structural rather than a matter of effort. Agentic commerce readiness spans marketing, product, IT, legal, security, customer experience, and, where it applies, commerce and merchandising. Most organizations have no function whose mandate covers that span, so the work either falls into the gaps between functions or gets adopted by whichever function it superficially resembled. This is the cross-functional readiness problem, and it is the one part of agentic commerce most reliably misdiagnosed as a tooling question when it is an operating-model question.
The default-ownership failure
The most common failure is the simplest. A single function owns agentic commerce because it sounded like that function's problem — marketing, because agent visibility resembled a marketing concern, or IT, because the protocol surface resembled a technical one. Neither function has the authority to make decisions that cross the others, so the decisions that cross functions do not get made. Marketing's brand-voice policy and legal's agent-authority-scope policy contradict at the edges and nobody owns the reconciliation. The agent's authority scope is set once, early, by whoever stood up the agent, and is never revisited, so it either commits to terms the business no longer wants to honor or escalates routine interactions to humans because the scope was drawn too narrow. Logs and transcripts accumulate and no function reviews them on a defined cadence, so drift and error patterns go undetected until they surface as something larger.
None of these is a failure of capability. Each is a failure of ownership, and they share a signature: the work that fell between functions was not refused, it was simply unassigned, and unassigned work in a cross-functional domain does not announce that it is not being done.
Why the other dimensions depend on governance
Governance Maturity sits at the intersection of every other dimension, which is why its absence is felt everywhere and attributed nowhere. The trust and security controls the briefing literature describes — actor classification, authentication, rate limiting, peer-agent verification — each have a technical implementation that lives in another dimension, and each requires a decision framework and a named owner to compose into actual protection. Assembled without governance, the controls stand inert: present in the architecture, unenforced in operation.
Authority scope is the cleanest illustration of the boundary. Brand-Agent Representation evaluates whether the agent exists and what it can technically do. Governance Maturity evaluates who decided what the agent is allowed to commit to, on what criteria, and with what review cadence. A brand can pass the technical dimension and fail the governance one with the same agent, because the agent being capable and the agent being governed are different properties. The framework keeps them on separate dimensions deliberately, so that publishing a capability is not mistaken for being able to operate it responsibly.
What running the assessment reveals
The assessment methodology makes the readiness problem concrete before any score is produced, because the assessment is itself a cross-functional process. Function representatives score the dimensions their function owns or has visibility into: marketing typically scores the brand-facing dimensions, IT and engineering the technical surface, and legal, security, and operations the governance dimension. A cross-functional readiness review then convenes those representatives to reconcile scoring disagreements, surface missing context, and validate the remediation priorities. An organization that cannot assemble those functions to run the assessment, or that runs it and discovers the representatives have contradictory pictures of who owns what, has already measured its governance maturity without needing the rubric.
The assessment lead role is where the problem concentrates. The role coordinates a process that every function feeds and whose output every function consumes, and no single function fully contains it. That is not an inconvenience to design around; it is the readiness problem stated precisely. The question of where the lead sits is the operating-model decision the rest of this depends on.
Where the ownership sits, and when it moves
The framework's position is directional rather than prescriptive, and it has a clear shape. Early in a brand's agentic commerce maturity, the assessment lead sits in marketing, because marketing carries the visibility into competitive positioning, agent-query simulation, and customer experience that the role draws on most heavily at that stage, and because marketing typically holds the convening authority for brand-level initiatives. As governance maturity rises, the role migrates. At a comparable level of maturity it remains marketing-led but acquires a defined cross-functional working group with named members and a set cadence. At a differentiated or agent-native level it moves to a dedicated function — a Head of Agentic Commerce or equivalent — reporting to a senior executive and operating with its own budget and authority.
The principle underneath the migration is the part executives should hold onto: the role's authority has to scale with the brand's agentic commerce maturity, and pinning it to a single function across every stage produces gaps as maturity rises. This is consistent with how the operating-model literature reads the broader shift. Boston Consulting Group and Bain both argue that siloed marketing organizations lack the speed and cross-functional integration agentic commerce requires, and that the response is cross-functional restructuring rather than a new tool, with the senior marketing role moving toward outcome accountability rather than campaign ownership (Wiener et al., 2026; Bhardwaj, Butler, & Fox, 2026). The framework's contribution is to make the migration conditional on a measured maturity stage rather than a calendar, so the structure changes when the governance does, not before it can be sustained and not after it has already failed.
A worked example
A brand has done credible technical work. Marketing ran a catalog and schema program that resolved identifiers and improved attribute coverage. IT built a functioning protocol surface. The differentiation work is real. On the dimensions that measure what the brand exposes, the brand would score well.
No function owns the operating model around any of it. Legal was consulted once, late, on the agent's authority scope and has not revisited it. No one owns first-party data quality, so credential stuffing has begun to pollute the data set the personalization layer trains on, and the degradation is invisible because no function reviews it. Observability outputs exist and are unread. Assessed in the framework, Governance Maturity scores Invisible. It does not trigger a floor, because the framework deliberately omits a floor on this dimension. The consequence is slower and harder to see than a floor would make it: the strong technical work is not failing today, it is decaying, because the authority scope drifts, the protocol surface returns stale data that no one investigates, and the first-party signal quality erodes underneath agents that keep training on it. A year later the brand's diagnostic scores have fallen and no single event explains why. The explanation is that the dimension that was supposed to keep the others current was never owned.
Where to start
The first diagnostic is an ownership map, and it is blunt: for each dimension in the framework, name the function that owns the brand's posture and the individual accountable for it, and mark every dimension where the answer is unclear or contested. The contested and blank entries are the readiness gap, stated more precisely than any maturity label states it. The second diagnostic is the coordination cadence — whether there is an actual operating rhythm in which the functions make joint decisions, or ownership exists on paper with no forum that turns it into decisions, which fails in its own way.
The decision an executive owns here is not which platform to buy. It is who holds the cross-cutting authority for agentic commerce, and at what cadence the functions coordinate, and how both scale as the brand's maturity rises. The other dimensions can be fixed by projects. This one is fixed by an operating model, and a brand that resolves every other dimension without resolving this one has built a set of capabilities with nothing assigned to keep them current.
References
Adobe Analytics. (2026). AI-driven traffic surges across industries, retail sees biggest gains [2025 holiday shopping recap]. Adobe.
Bhardwaj, S., Butler, L., & Fox, S. (2026). Rewiring demand generation in the age of AI agents. Bain & Company.
Wiener, L., Beaulieu, F., Kropp, M., Kelman, L., Iny, A., & Ho, V. (2026). Agentic scenarios every marketer must prepare for. Boston Consulting Group (BCG).