For almost thirty years, the internet had only two types of visitors: people and bots. People bought, booked and complained; bots crawled, indexed and, when they became too much of a nuisance, were blocked. That boundary is being erased with the arrival of AI agents: programs that act on behalf of a real customer to search for a flight, compare prices, book an appointment or dispute an invoice. They are not people, but they are not conventional bots either. And your website, your online store and your customer service department still don’t know what to do with them.
The issue is very well framed in a recent article from MIT Technology Review, "Making the internet better for AI agents might make it worse for everyone else", which warns of an uncomfortable paradox: if we redesign the internet so that machines can work comfortably, it may end up being a considerably worse place for people. It is worth looking at this from the business side, which is where we operate at Tisa.
Neither person nor bot: a third type of visitor
The underlying problem is one of classification. The systems we use to separate humans from machines —captchas, request limits, access rules— were born for a binary world. An agent that browses at machine speed, but with a legitimate intent and a customer behind it, doesn’t fit into any box.
The initial reaction of many platforms has been the simplest one: block. According to the aforementioned article, major e-commerce and booking brands have prevented these assistants from accessing their sites, and there have even been lawsuits between companies over it. The text also mentions very rapid growth in traffic generated by agents and download figures for the new applications
The argument of those who block is not arbitrary: an agent that requests refunds, negotiates prices or floods a support form can render unusable a channel designed for people. Nor is the argument of those who open up: if the agent brings a customer with card in hand, blocking it means closing the door on a sale.
The risk: a web optimised for machines
Here is the warning issued by MIT Technology Review, and it is worth taking seriously. If automated traffic ends up being the majority, the incentives shift: interfaces for humans get neglected, identity verifications multiply, prices stop being single and public and become different depending on who (or what) is asking, and you never know whether the recommendation your assistant gives you is influenced by a commission.
For a private company this is not philosophy: it is commercial strategy. The question is not whether you are for or against AI agents, but what policy you want to have when they knock on your door. And it is best to decide that before the first odd traffic spike on a Friday afternoon decides it for you.
What this means for your business
Customer service and after-sales
If your claims or returns processes depend on the volume being "reasonable", automation will put them to the test. It is worth reviewing which procedures can be resolved automatically with clear rules (a return within the deadline, a duplicate invoice) and which require human judgement. What is well defined scales; what depends on the goodwill of a person does not.
Prices, catalogue and channels
An agent compares in seconds what used to take a customer an entire afternoon to review. That rewards whoever has a clean catalogue, real stock and updated delivery times, and penalises whoever publishes information that doesn’t match their ERP. In retail and distribution —sectors we have been working in for years— that consistency between online store, warehouse and management system was already important; now it is downright competitive.
Data, security and governance
Neither you nor your customers want a card or a personal data item circulating uncontrolled. Before opening any automated channel you need to know what information is exposed, with what permissions and with what record of what happened. This is exactly the data governance work we propose in our AI projects: without organised data and clear rules, no automation performs well or can be defended.
How to prepare your business for AI agents
There is no need to reinvent anything tomorrow. But it is worth having a position, and it can be built in very down-to-earth steps:
- Measure. Review your traffic and separate the human from the automated. You can’t make decisions about something you can’t see.
- Decide your policy. What do you allow an external assistant to do: check prices, book, buy, complain? What not, and why?
- Sort out the data before the channel. Catalogue, stock, rates and terms must come from a single reliable source: your ERP.
- Separate lanes. Handling automated requests through a controlled route (an integration, an API) protects the experience of the people browsing your website.
- Define the exceptions. What gets escalated to a person, under what criteria and within what time frame.
- Review compliance. Data protection, traceability and basic security. Better to think about it before rather than after the incident.
The other side of the counter
It is worth not forgetting the other half of the story: your company can also use AI agents for its own benefit. Comparing supplier offers, preparing case files, cross-checking delivery notes with invoices, summarising service incidents or preparing reports that today consume hours of a team’s time. The same technology that forces you to rethink your website can take administrative work off your hands.
The key, as always, is to start with a measurable use case, not with the tool. A specific process, with available data, someone responsible and a way of knowing whether it has improved. That is how AI delivers margin and not just headlines.
Let’s talk about your case
At Tisa we have been helping companies in retail, distribution, construction, the food industry, hospitality and services bring order to their management since 1987, with Microsoft Dynamics 365 Business Central, Power Platform and custom development. That order is, today, the best possible preparation for an internet that is changing tenants.
If you want to assess how all of this affects your business and where to start, write to us at info@grupotisa.com or call us on (+34) 971 305 885. First conversation with no commitment.
Reference source: MIT Technology Review, "Making the internet better for AI agents might make it worse for everyone else" (The Algorithm).
