When AI agents become your customers: a CIO's guide to staying in control

Why identity, intent and accountability are becoming critical CIO concerns.

Nov 18, 2026

Dirk Jan van der Pol, Head of Commerce, Empathy Lab
Brian Gilmore, Global Head of Commerce GTM, EPAM

AI agents are changing more than how customers discover and buy products. They're changing the nature of the customer relationship itself.

Unlike human customers, agents can search at scale, compare options without emotion and complete transactions in seconds. They may also be acting on behalf of multiple people at once, with their own commercial logic influencing every interaction. 

For CIOs, this creates a new challenge. Your systems may record a successful sale without knowing who the agent represents, what it's authorized to do or whose interests it's serving.

The intent problem: you made the sale and learned nothing

When a human customer buys from you, the relationship is relatively easy to see. One person, one session, a login or at least a cookie. You know who they are or have some way of identifying them. Everything downstream depends on that: loyalty, personalization, service, returns and the next campaign.

A browser-based agent can break that assumption.
It may arrive looking like any other session, without identifying who it represents or what it has been authorized to do. It adds items, applies a code, pays and leaves. Your systems log a successful transaction. What they can't necessarily tell you is who you just sold to.

That's the real intent problem. The risk isn't necessarily that agents are malicious. It's that their intent may be invisible to you. Your promotions, pricing and loyalty rules were designed for a known or identifiable customer. Applied to an anonymous agent session, those same rules may still work, but with unintended consequences.

4 ways agent activity can disrupt business as usual

 

Promotions
Offers intended for new customers may be redeemed by intermediaries with no direct relationship to your business.

Loyalty
Rewards may be collected or consolidated across customers in ways your program wasn't designed to support.

Pricing
Agents can systematically test and compare prices without any intention of buying.

Service and recalls
Without a reliable link to the end customer, follow-up communication and support become more difficult.

None of this requires bad actors. We've seen this before with coupon extensions capturing affiliate credit and bots draining limited releases. When an intermediary's interests don't align with those of the customer or retailer, the results aren't always what anyone intended. An agent doesn't have to be disloyal to its user to leave you with a commercial outcome you never intended.

“Yesterday, CIOs needed to know who had access to their systems. Today, they need to know who agents are acting for, what they're allowed to do and whose interests they serve.”
Dirk Jan van der Pol, Head of Commerce, Empathy Lab

From anonymous to accountable

 

This isn't something marketing can solve alone. Marketing can design the promotion, but IT needs to help establish who's on the other end of the transaction. Specifically, the agent's identity, its authority to act, how that authority is verified and the audit trail linking the transaction to the customer it represents.

The goal isn't to block agents indiscriminately. It's to understand who they represent, what they're authorized to do and whose interests they serve. With that visibility, CIOs can make informed decisions about which agent interactions to enable, restrict or block. And build trusted relationships with agents that bring genuine value to the business and its customers.

Detection is necessary but not sufficient

According to the July 2026 edition of Adobe Digital Insights, AI-referred retail visitors converted at a rate 60% higher than non-AI traffic and generated 53% more revenue per visit. The commercial opportunity is real, even if AI-referred visitors aren't necessarily the same as autonomous shopping agents.

But knowing that an agent is interacting with your business is only the starting point. CIOs need to understand what it's trying to do, on whose behalf and whether that aligns with the commercial relationship the business intended.

That means rethinking the rules for how agents interact with your business:

 

  • Volume discounts may need safeguards against aggregation across agent-managed accounts.
  • Limited promotions may need controls to prevent automated abuse.
  • Loyalty programs need ways to distinguish legitimate customer activity from agents exploiting rewards across multiple accounts.

These controls can't live exclusively in an API or an agent-specific integration. Browser-based agents interact with your website, checkout and promotional systems directly.

“Governance has to meet the agent where it is: in your UI, your checkout flow, your promotional engine. Not where you would prefer it to be.”
Brian Gilmore, Global Head of Commerce GTM, EPAM

Who decides the rules of engagement?

 

A CIO who treats agent governance purely as a security challenge may focus on detecting and blocking. But treating it as a commercial challenge opens up a more useful question: which agent interactions should we welcome, which should we restrict and which should we stop?

That requires close collaboration between the CIO, the CMO and the commercial team. The CIO brings the technical capabilities to detect and analyze agent intent. The CMO and commercial team define the customer relationships and business rules that guide how the enterprise responds. Neither can build an effective governance framework without the other.

But understanding agent intent is only part of the challenge. CIOs also need to decide how open their enterprise should be to agent-initiated interactions, and where that openness becomes exposure. We’ve put together a playbook for that – read more in our next article.

You can't govern what you haven't seen

 

Before deciding which agents to welcome, what to protect and where to draw the line, you need to understand where your enterprise stands. Our Agentic Commerce Readiness Starter Assessment helps CIOs and commercial teams evaluate agent activity, assess data and content readiness, and identify gaps in governance, access and accountability.

The result? A clear picture of your readiness and a prioritized roadmap for what comes next.

Contributors in this article

Dirk Jan van der Pol
Head of Commerce, Empathy Lab
Brian Gilmore
Global Head of Commerce GTM, EPAM
Stay up-to-date with everything Empathy Lab
Follow Empathy Lab on LinkedIn