The revenge of the consumer agents?
I stumbled upon this post on X some time ago, but it stuck. A person used the back then brand new Muse agent from meta to get a refund and re-booked a flight. A typical task that would cost you hours on the telephone hotline or filling our stupid form etc. And here the AI wins: It simply doesn’t care like a human filling out stupid things over and over again or listen to endless stupid waiting in the line music. Have we finally found the killer use case for broad AI adoption: fighting the customer service nightmare companies built up over the past decades in the endless race to cut costs and maximize shareholder value?
Before we look deeper into what the AI can do here for humans, we need to recap on what current customer centric systems were built upon. Mostly we have to thank Behavioral economics for all this, especially the negative nudges that are everywhere today in all sort of customer interactions hidden.
Behavioral economics and what is a negative nudge?
Behavioral economics is a field of economics that examines how people make decisions in real-life situations, recognizing that individuals do not always act rationally or make choices based purely on logic and self-interest. Instead, their decisions are often influenced by psychological factors, emotions, social norms, habits and cognitive biases. Unlike traditional economic theory, which assumes that individuals have stable preferences and make rational decisions, behavioral economics acknowledges that people are susceptible to predictable patterns of irrational behavior. One important concept within this field is the nudge, which refers to a subtle change in the way choices are presented that influences people’s behavior without restricting their freedom of choice or significantly changing economic incentives.
While positive nudges encourage individuals to make decisions that benefit themselves or society, negative nudges steer individuals towards less desirable behaviors or discourage certain actions through the way choices are structured. These can include making beneficial options less visible, creating inconvenience when accessing certain services, using social pressure, or framing information in a way that influences decisions through fear or guilt. Although negative nudges can sometimes be used to discourage harmful behaviors, they also raise important ethical concerns regarding manipulation, individual autonomy and the responsibility of institutions that design choice environments.
In the context of this post we see it as an anti pattern from service endpoints that are designed to annoy and demotivate humans. Resulting in less contacts with real agent that would cost money, less successful claims and on first sight: more money/saving for the company implementing those.
Human led AI agent against companies negative nudge barriers

With the example from the beginning we see the revenge of the consumers now bringing their very own AI agents to fight those systems. And they are good in such tasks every human would get frustrated very soon: agents don’t get bored, don’t get emotional and inpatient when letting them wait, they can fill the same data in dozens of forms over and over again, like in the example they can also access a users email and confirm real-time traps built in and can do all this all day and night long. Still one factor costs of the agent, meaning their tokens usage and other cost to run them (subscriptions etc.) Those costs compared to the time savings and stress reliefs are in no relation and the giving dropping tasks to my agent calculation is a simple one.
The revenge of the institutions…
On the other hand there are the companies that are now confronted with AI agent accessing their service and transaction tools in authorization of their users. Are they happy about it? Fuck no:

Big companies block AI agent access now as a result of the boom wave of easy to access personal agents like Muse from Meta or Duo from Anthropic or Dots from OpenAI. You can only spaltete about the reason why they are doing this, cause of course NO ONE gave plausible reasons for it. Security threat was the argument most used here. There are of course also economical reasons behind it of course as well:
- you cannot server adds to an agent (most likely #1 reason)
- agents are not prone on sell-up, hidden costs and other tricks in buying processes (like from booking a flight)
- agents cant be emotional triggered or manipulated like humans
- companies invested a lot of money into their systems with negative nudges they saved them previously as lot of money
Where is this heading…
Is this leading us into an ironical future where companies will hide behind anti pattern systems and their own agents to cut costs and cut human interactions. On the other side humans use their agents to fight the anti patterns and get to their results and outcomes?
This can’t be something any of both parties can’t be thriving through but the big hype of agentic commerce goes in that direction: It claims that in a near future there will be two target group categories with completely different needs: humans and agents. Agents then buy products on behalft of their humans and need to be addressed differently in the entire buying and transaction process – that’s what the folks say that want to boost this topic and make money around it. I am very skeptical all about this and not seeing this happen any soon with a substantial shift in buying behaviors. It is not compared with the invention of the internet end of the 90s and the invention of e commerce for me and also still this revolution didn’t make all psychical shops disappear yet tough.
As a marketeer I can no wish for such a future when agents decide based on simply transactional logic. Also when agents fight each other over outcome and there is no meaningful connection anymore between humans and brands/companies. This connection is the base of what our brand dominated consumer economic system holds together for so many years.

Leave a Reply