A fireside chat on the revenue case for humanistic design in digital commerce.

Let’s be clear: ISG Research asserts, by 2028, one-third of enterprises will rely on agentic artificial intelligence (Al)-driven digital commerce to proactively manage customer interactions and transactions end-to-end, improving operational efficiency and increasing conversion rates. Case in point, during a recent hotel stay, I wanted to book a spa service for myself, friends and family. The resort directed me to an app.

I wanted a conversation. Someone who knew the services, could explain our options and could help me choose an experience we would enjoy together. I had questions about the environment for neurodivergence and sensory‑sensitive clients. Instead of easing into a relaxing afternoon, I found myself on sensory overload, navigating a process when I wanted guidance.

It left me wondering: Had the hotel forgotten one of its most important advantages over a vacation home? Hospitality.

The expertise, the welcome, the feeling that someone is paying attention to what you need. Those are part of what I am buying when I choose a resort. I wanted that care to begin while I was deciding what to purchase.

That experience captures two questions I want chief revenue officers to ask as bots and AI agents take on more of the customer relationship: How much work are we asking customers to do to give us their business? Is this new digital experience diminishing our core differentiation in the market?

We can make an interaction less expensive to deliver while making it harder for the customer to complete. That possibility belongs in the automation business case.

When More Engagement Becomes More Work

The constant dings get to me. Each one asks for a little more of my attention. Add moving content, proactive messages, and prompts that appear before I have finished reading and the experience can become exhausting. So, I sat down with psychologist and professor Dr. Charmeka Newton during a recent American Psychological Association event to understand more. She puts words to what happens in those moments: “Attention is a finite psychological resource. Every notification, interruption, and prompt adds to the mental effort an interaction requires.”

That is worth sitting with. We often talk about engagement as though more of it is always better. But a customer responding to a fifth prompt may still be trying to get past the first problem.

Research in the American Psychological Association’s Journal of Experimental Psychology: Human Perception and Performance finds that phone notifications disrupt performance on an attention-demanding task even when participants did not interact with their phones. The study did not measure sales; however, it does give us a reason to question the assumption that an ignored notification has no cost.

These demands deserve particular attention when designing for neurodivergent customers and people with different sensory, attention and reading needs. There is no single experience to design around. One person may appreciate a conversational interface. Another may need a quiet page, time to read or the opportunity to speak with someone. “When those demands exceed a person’s capacity in that moment, engagement can give way to overwhelm and withdrawal,” Dr. Newton explains. “Giving people control over the experience supports their sense of agency.” Digital commerce platforms must work to protect user agency, not as a nice-to-have, but as an imperative.

For revenue leaders, that raises a question worth investigating: Did the customer lose interest in the purchase or did the process become too demanding? A record of abandonment cannot answer that on its own.

Revenue leaders can investigate that question in three ways:

1. Look at what happened before the customer left. Identify repeated steps, errors, interruptions and prompts that appeared while the customer was still completing a task. These patterns can reveal demands that an abandonment rate alone cannot explain.
2. Ask customers what got in their way. Offer a brief, optional way to explain why they stopped, including whether they were browsing, reconsidering the purchase, having difficulty or feeling overwhelmed. Include people with different sensory, attention and reading needs in customer research.
3. Test an experience that gives customers more control. Let people pause prompts, reduce motion, read at their own pace and choose how to get help. Compare task completion and reported effort with the existing experience. If customers find it easier to continue, that is evidence that the process itself may have been a barrier.

The Commercial Stakes Are Real

There is evidence that the way customers experience automation can affect purchasing. A more recent example concerns checkout design. In March 2026, Walmart executive Daniel Danker told WIRED that products sold through ChatGPT’s Instant Checkout converted at roughly one-third the rate of products requiring a click through to Walmart’s website. He pointed to individual-item checkout and concerns about separate shipments. The product selections differed, so this was not a controlled comparison or proof that shoppers rejected bots.

The example brings me back to the same management question: Is this particular experience helping people complete a purchase? Message counts and reduced human contact are insufficient answers. A bot can keep a conversation going without bringing the customer any closer to a decision.

Sometimes the Missing Feature Is a Person

At the resort, I wanted someone to help me understand what would work for us. That expertise was part of the value of the service.

Research by Chiara Longoni and Luca Cian in the Journal of Marketing helps explain why the type of decision matters. Their research on AI and human recommendations finds a preference for AI when functional attributes were emphasized and for people when taste and enjoyment mattered. When recommendations needed to reflect unique personal preferences, participants preferred human recommenders even for functional decisions.

Checking whether a spa appointment is available and deciding which experience is right for a parent and child involve different needs. Our service choices should make room for both. As Dr. Newton explains, “Human contact matters psychologically: a knowledgeable person who listens, acknowledges a concern, and responds to its context can restore a sense of control and build trust.” This is what I mean by humanistic design: putting people’s agency, dignity and individual needs at the center of the experience. In practice, customers should be able to understand their options, choose how they engage and get useful help.

A friendly bot name and sympathetic language can only do so much. The Journal of Marketing study “Blame the Bot” finds that humanlike chatbots worsened satisfaction, company evaluations and subsequent purchase intentions among customers who entered the interaction angry. Their presentation raised expectations of helpfulness that the interaction did not fulfill.

For me, the implication is straightforward. Match the promise of help with the ability to provide it. Give the system the knowledge and authority it needs and provide a clear route to someone who can handle what it cannot.

Let Customers Stay in Control

As agents move from answering questions to taking action, control becomes even more consequential.

A British Journal of Management study of AI shopping assistants finds that greater autonomy reduced adoption intentions through a diminished sense of personal power. The effect varied with product scarcity, so autonomy was not a barrier in every setting. It does suggest that leaders should pay attention to how delegation feels to the customer.

Can I see what the agent is about to do? Can I change my mind? Can I stop it without losing everything I have already entered?

Those are reasonable expectations for a service acting on someone’s behalf.

Recovery matters, too. In research published in APA’s Journal of Experimental Psychology: General, people lost confidence more quickly in algorithmic forecasters than in human forecasters after seeing them make errors, even when the algorithms performed better overall. Those experiments concerned forecasting, rather than customer service, but they illustrate how fragile confidence in an automated system can be.

I would translate that into a practical requirement: correct the mistake, preserve the customer’s information and offer another way to continue. Dr. Newton describes the stakes clearly: “When a customer service bot repeats questions, offers irrelevant answers, or makes it difficult to reach a person, it adds cognitive load at the very moment someone is seeking help.” The customer came with a problem. We should be careful about adding another one.

And this is the line I would put in front of every team designing these experiences: “Customers should not have to prove how frustrated they are before they deserve to be heard.”

Five Actions for CXOs

Here is where I would start the leadership conversation.

1. CRO and CFO: Measure whether the customer succeeds. Connect automated interactions to completed purchases, abandonment, repeat business and contribution after technology and support costs. Before declaring an interaction without human involvement a success, establish whether the customer’s need was resolved. Ask whether lower service costs are being offset by lost purchases or repeat contacts.
2. Chief customer officer and product leader: Let customers set the pace. Make nonessential sounds optional, let people dismiss proactive messages and respect those choices. Provide clear navigation and enough time to read. Include neurodivergent customers and people with varied sensory and processing needs in design testing. Ask what helps them complete the task, and measure where the experience gets in their way.
3. Customer service leader: Preserve access to expertise. Offer a direct route to a person for questions requiring judgment, reassurance or exceptions. Transfer context so customers do not have to repeat their story. Evaluate whether these interactions resolve concerns and help customers complete purchases. A useful human conversation deserves credit in the revenue model.
4. CIO and commerce leader: Make delegation explicit. Explain what an agent can recommend, change or purchase. Provide meaningful review before consequential actions and a clear way to revise or stop instructions. Permission for one transaction should not silently become permission for future ones. Design recovery with the same care as the initial interaction.
5. Executive sponsor and analytics leader: Test the whole journey before expanding. Use controlled comparisons where feasible, accounting for customer intent and purchase complexity. Establish thresholds for conversion, errors and failed handoffs that trigger review or rollback. Assign an owner for the commercial outcome across technology, service and commerce.

I would ask software providers to demonstrate the difficult moments, too. Show me what happens when a customer declines assistance, asks for another channel or raises a question the system cannot resolve. Show me whether the customer can still move forward.

Hospitality Begins Before the Booking

I keep coming back to that spa experience. I wanted to spend time with my daughter, and I wanted someone knowledgeable to help me choose. That conversation was part of the service I was seeking.

As leaders, we should ask whether our technology preserves the qualities that make customers choose our businesses. For a resort, hospitality belongs in the booking process as much as it belongs in the treatment room.

Your bot may be engaging. Make sure your customer still wants to stay.

Regards,

Barika Pace