AI Agents Are Becoming a New Sales Channel. Most Organisations Have Nobody Accountable for It.
By Malin Gaertig · September 9, 2026
Search created a new distribution channel. Social created another. AI agents are creating the next one — but unlike search or social, agents don't sit neatly inside an existing function. Marketing owns acquisition. Ecommerce owns the website. IT owns infrastructure. But who owns the moment when an AI agent decides your brand is not the one to recommend? That is the Agentic Accountability Gap, and it points to a commercial ownership problem most organisations have not answered yet.
- —AI agents are a new distribution channel, but unlike search or social they don't sit inside any one function — so the moment an agent chooses a competitor often has no internal owner.
- —The commercial battle has shifted from 'Can the customer find us?' to 'Will the agent select us?' — a brand can be highly visible and still miss the recommendation that shapes the purchase.
- —Interceptors (marketplaces, aggregators, comparison sites) can capture demand even when your brand created it, and your traditional competitor set won't show them.
- —The Agentic Accountability Gap is easy to miss because traffic, conversion and awareness can look stable while revenue quietly slips upstream of every metric you already track.
- —Start with three questions: Are we being selected? Who is intercepting the demand? What happens when an agent reaches us? — measured by MARK's Share of Model, interceptor mapping and Ghost Shopper.
The channel is bigger than ChatGPT
When most organisations talk about AI, they still tend to mean a chatbot, but that's increasingly too narrow. AI can now influence the customer at three points in the journey:
Selection — An agent can decide which brands make the shortlist.
Evaluation — It can compare products, prices, availability, reviews and alternatives.
Transaction — It can increasingly take action on the customer's behalf, from finding the right product to completing a purchase.
The same shift is happening on the other side of the transaction. Agents are being built to help merchants manage catalogues, pricing, advertising and customer interactions. Payment companies are building infrastructure so agents can transact without a person manually completing every step.
So this is not one new chatbot, but an entirely new layer between demand and supply. This creates a simple commercial question: When an agent stands between your brand and the customer, who makes sure your brand wins? Most organisations don't have an answer yet.
The problem is not just visibility
The first instinct is to treat this as another version of search, but that misses a critical step. In traditional search, the objective is largely to get found and earn the click. But an agent makes a decision and influences the shortlist before the click happens.
Ask an agent for the best running shoe under $200. It may return three brands. Ask for a hotel near the CBD with a late check-in. It may choose one. The commercial battle has therefore changed from 'Can the customer find us?' to 'Will the agent select us?' That distinction matters because a brand can be highly visible in traditional channels and still be missing from the recommendation that actually shapes the purchase.
Worse, the agent may select a competitor, marketplace, comparison site or aggregator instead. That's what we call interceptors — a brand that captures the demand even when your own brand created it.
The new competitive set is not the one on your strategy slide
Most companies know who their competitors are, but AI agents may have a broader interpretation. Ask an agent for a product in your category and the recommended brands may include companies you would never put on a conventional competitor list. That is because the agent is entirely focused on delivering on the customer's request.
If a marketplace has better product information, a retailer has clearer availability or an intermediary is easier for the agent to cite or transact through, that business will enter the consideration set. This is why traditional competitive monitoring is becoming incomplete — it misses who the agent deems most relevant and adds to the shortlist.
And nobody is watching the whole journey
This is where the accountability problem becomes practical. The ecommerce team can measure conversion. The SEO team can measure rankings and the media team can measure paid acquisition. But none of those reports necessarily tells you that an AI agent recommended a competitor instead of your brand, sent the customer to an intermediary or abandoned the purchase because your site could not provide what it needed.
The loss often happens upstream of every metric you already track. That is the Agentic Accountability Gap: the space that opens when an AI agent makes a commercial decision on a customer's behalf and no one inside the organisation is accountable for the outcome.
And it is easy to miss because nothing necessarily looks broken initially. Traffic may be stable, conversions may be stable and brand awareness may be stable. But gradually traffic declines for no reason. Conversion volume is lower and revenue starts slipping.
Where should an organisation start?
Start by answering three questions:
Are we being selected? Identify who agents recommend in your category. Measure how often your brand appears, where it appears and who appears instead or alongside.
Who is intercepting the demand? Look beyond your named competitors. Identify the retailers, marketplaces, aggregators and other intermediaries entering the recommendation and potentially capturing the transaction.
What happens when an agent reaches us? Test the journey from the agent's perspective. Can it find the right information? Compare products? Establish price and availability? Complete the task? Or does it fail and send the customer elsewhere?
Those three questions give a leadership team something more useful than an AI trend report. They give it a view of where the organisation stands, who is winning the demand and where ownership needs to sit. This is exactly what MARK measures — Share of Model (are you being selected), competitive and interceptor mapping (who is intercepting), and Ghost Shopper (what happens when an agent reaches you).
Ask ChatGPT, Gemini or Claude to recommend a product, service or supplier and it influences the sale before your customer reaches your website. It can recommend your brand, a competitor or send the customer to a marketplace instead. Increasingly, it does more than recommend: it can research, compare, select and transact on the customer's behalf. Your existing channel reporting may never show you why you lost that customer.
Search created a new distribution channel. Social created another. Organisations eventually built teams, budgets and metrics around both. AI agents are creating the next one, but the difference is that agents don't sit neatly inside an existing function. Marketing owns acquisition. Ecommerce owns the website. IT owns infrastructure. But who owns the moment when an AI agent decides your brand is not the one to recommend? That is the Agentic Accountability Gap and it points to a commercial ownership problem.
See how this is measured in the MARK product suite, or read the latest AI Commerce briefings.