In our new series, Jonathan McCrea looks beyond the headlines of how AI is transforming businesses, work, education and research. This week faces AI agents and their involvement for retailers.
If certain technological giants realize their vision, we will soon enter a black reality in the style of the mirror. Where the machines that know us very well will buy the things we need with us just thinking about it. We will not have to search and compare, verify the availability, call a telephone line that nobody Andra or complete a application form for a service. We will simply ask for something and it will seem magic, all that abundant research conducted by us in the background by an AI. For tired consumers who only want a garden sprayer that works or an adequate gift for a 10 -year -old boy’s birthday party, who can sound like heaven. For small producers and service providers trying to survive, it may sound more like hell.
Agents are the next great evolution of AI. At first, Chatgpt and its competitors were only chatbots. The smart chatbots who had consumed the Internet with no one if that was fine, but chatbots without power. They started speaking human language, some could speak in code, but they never “went” anywhere, they could not “do” anything except to have a conversation. They live inside the genius lamp.
In just two and a half years, the blink of an eye in terms of research, we have models that can combine impressive reasoning skills with the use of tools. OPENAI’s O3 model, Deepseek and Claude Sonnet 3.7 can create images, analysis databases, draw graphics and collaborate in the writing code or documents. Some of the reasoning models seem to even show us how they address a task: we can “see” their reasoning as it happens (although they do not always reason as they say they do).
I dream of genius
But until now, humans were very in the driving seat: we would instruct an AI on how to do something, then we would try to do it, with variable results, just like the moody adolescent who lives in my house. Now we are entering a new era: the dawn of the age of the agents. New systems are being built where AI should no longer be said how to execute a task; Instead, exercise reasons and the use of the tool to make your own decisions on how to get to a result. To facilitate this, technological companies such as Anthrope and Openai are building a world to allow “agents” to connect to any Internet service that has an API, unlimiteding unlimited potential. This is the time when the genius leaves the lamp.
Recently, they cooled to Manus from China, a closed beta agent who has caused a deserved exaggeration, a Devifitult task: classifying more than 1,000 conversations for an upcoming conference on a significant program where each session the richness of the subject was grouped and the extensors had the same Eage Eage in the account of the speakers through the session.
I would have tasks of an expert in that field time to do this. Manus read the database, wrote some code, fixed the errors in his own code, compiled a Python script, conducted a program and executed it. The program read the database, analyzed the information and gave me a database with new columns that classify all the conversations exactly as I had requested. This whole process touches eight minutes. I have seen this same fully generated work websites, interactive maps and equally basic video games with a single message.
In this impressive demonstration in a Google Cloud Next event, Patrick Marlow shows a AI customer service agent that is the next level.
Marlow is on the website of a garden store. You need to buy some fertilizer for the plants you have at home. The agent requests access to your webcam to identify the plants, change the products in your shopping cart for others more appropriate, recommend and then reserve a planting service and even consult a human manager about anyone or cannot get a discount. And while this is a demonstration in a technological event, it shows that the child of versatility is aviaxable to sites that offer electronic commerce.
Other things are also happening in this space. Amazon Alexa+ can apparently soon complete complex services, such as repairs or returns. Visa has announced an association with OpenI to allow authorized agents to buy in excesses of users that lead to speculation that the search for Son, Chatgpt, will allow purchases directly in the application.
Alarm bells
While all this may sound fantastic for the end user, the important warning bells should sound for small businesses in the retail and services provision. While the culturally more legislative EU will work hard to protect rights and just access to the market, there is a lot of development at the same time and it will be difficult to keep up.
The general description of the Google AI that appears when it executes a traditional search is preventing many customs from visiting third -party sites. It could be much more difficult for your product to look very soon.
Think it. At this time, many incoming sales depend on search and advertising. Humans are looking for a product on Google, investigate a bit and make a decision. On the way, you will see marketing and advertising designed for human eyballs. This research could soon be carried out mainly by an AFA agent designed by Amazon and adapt very specifically to the client. He knows that you are a vegetarian, that you normally prefer the quality of the price. He knows that you wear a SIZ SIX race shoe and that you have read most of the aisling books and probable elegant Paul Mescal. He knows that he has just made an extension to the house and could be waiting. Alexa+ knows all this because he will remember almost everything you ask.
So, as owner of a business, how is it ‘found’ by this new or agent? The answer to this question is complicated and, of course, is not completely clear.
There are some standard things that can be done to help: make sure that your product database is fully updated, which SEO has done in all its online offer, but there are also new things to consider. New market research suggests that agent’s search favors the relevance of the task and the content structured in the coincidence of keywords. Its website should be formatted using programming marking so that IA agents can analyze it efficiently. Essentially, it must make the easiest possible thing that an agent obtains the answers you need in real time.
Where there is a threat, you think about the opportunity. For the future, agile SMEs, this agitation is less a threat than an invitation: those who adopt the early agent can become the “favorite suppliers” that these digital janitors seek, when designing with capacity for the machine in mind, associating with emerging agents markets and maintaining an eye in emerging AI technologies.
In Ireland, we quickly need not only being literate from AI, but AI friendly, because eliminating some business friction always produces better results. Ionic, then it is a bit of friction that took us there: the rubbing of the genius lamp.
More information about Jonathan begins with AI here.
Do not miss the knowledge you need to succeed. Register for him Daily letterSilicon Digest Republic of Science Fiction Technology News.
