If certain tech giants realise their vision, we will soon enter a Black Mirror-esque reality. Where machines that know us very well will buy the things we need with us barely thinking about it. We will not have to search and compare, check availability, ring a phoneline that nobody answers or fill out an application form for a service. We will just ask for something and it will appear like magic, all of that tiresome research done for us in the background by an AI. For jaded consumers who just want a garden sprinkler that works or a suitable present for a 10-year-old’s birthday party, that might sound like heaven. For small producers and service providers trying to survive, it may sound more like hell.
Agents are the next big evolution of AI. In the beginning, ChatGPT and its competitors were just chatbots. Clever chatbots that had consumed the internet without asking anybody if that was OK, but chatbots nonetheless. They started off speaking human languages, some could speak in code but they never “went” anywhere, they couldn’t “do” anything except have a conversation. They lived inside the genie’s lamp.
In just two and a half years – a blink of an eye in terms of research – we have models that can combine awesome reasoning skills with tool use. OpenAI’s o3 model, Deepseek, and Claude Sonnet 3.7 can create images, analyse databases, draw graphs and collaborate on writing code or documents. Some of these reasoning models appear to even show us how they approach a task: we can “watch” their reasoning as it happens (although they don’t always reason like they say they do).
I dream of Genie
But until now, humans were very much in the driving seat – we would instruct an AI on how to do something, then it would try to do it, with varying results, much like the sulky teenager living in my house. We are now entering a new era – the dawn of the age of agents. New systems are being built where the AI no longer needs to be told how to execute a task; instead it wields reasoning and tool use to make its own decisions on how to arrive at an outcome. To facilitate this, tech companies like Anthropic and OpenAI are building a world to allow “agents” to connect to any service on the internet that has an API – unleashing limitless potential. This is the moment where the genie comes out of the lamp.
I recently gave China’s Manus, a closed beta agent that has caused deserved hype, a difficult task – sorting over 1,000 talks for an upcoming conference into a meaningful program where each session was grouped thematically, had the same amount of speakers and each subject flowed into the next in a logical way throughout the session.
It would have taken an expert in that field hours to do this. Manus read the database, wrote some code, fixed the errors in its own code, compiled a Python script, made a program and ran it. The program read the database, analysed the information and gave me a database with new columns sorting all of the talks exactly as I had requested. This whole process took eight minutes. I have seen this same AI generate fully working educational websites, interactive maps and even basic video games with a single prompt.
In this stunning demo at a Google Cloud Next event, Patrick Marlow shows an AI customer service agent that is next level.
Marlow is on a gardening store website. He needs to buy some fertiliser for plants he has at home. The agent requests access to his webcam to identify the plants, changes the products in his shopping cart for more suitable ones, recommends and then books a planting service, and even queries a human manager about whether or not it can apply a requested discount. And while yes, this is a demo at a tech event, it demonstrates the versatility soon to be available to sites offering e-commerce.
Other things are happening in this space too. Amazon’s Alexa+ will apparently soon be able to fully complete complex services for you like repairs or returns. Visa has announced a partnership with OpenAI to allow authorised agents to purchase on behalf of users leading to speculation that soon, ChatGPT search will allow shopping directly in the app.
Alarm bells
While all of this might sound fantastic to the end user, major warning bells should be ringing for small businesses in retail and service provision. While the culturally more legislative EU will work hard to protect rights and fair access to the market, there are a lot of developments happening at the same time and it will be difficult to keep up.
Already, Google’s AI Overview that pops up when you run a traditional search is stopping many customers from visiting third-party sites. It could get much harder to have your product seen online very soon.
Think about it. Right now, a lot of inbound sales rely on search and advertising. Humans Google a product, do a bit of research and make a decision. Along the way, they will see marketing and advertising designed for human eyeballs. That research could soon mostly be undertaken by an AI agent designed by Amazon and tailored very specifically to the customer. It knows you are vegetarian, that you typically prefer quality over price. It knows that you wear a size six running shoe and that you’ve read most of the Aisling books and probably fancy Paul Mescal. It knows you’ve just done an extension to the house and might be expecting. Alexa+ knows all of this because it will remember almost everything you ever ask it.
So as a business owner, how do you get ‘found’ by this new breed of agent? The answer to this question is both complicated and of course, not fully clear.
There are some standard things you can do to help – ensure your product database is fully up to date, that you have performed SEO on your entire online offering, but there are new things to consider, too. New market research suggests that agentic search favours task relevance and structured content over keyword matching. Your website needs to be formatted using schema markup so that AI agents can parse it efficiently. Essentially, you need to make it as easy as possible for an agent to get the answers it needs in real time.
Where there is a threat, there is opportunity though. For forward‑thinking, agile SMEs this upheaval is less a threat than an invitation: those who embrace agentic AI early can become the very “preferred suppliers” these digital concierges seek out, by designing with machine discoverability in mind, partnering with emerging agent marketplaces and keeping an eye on emerging AI technologies.
In Ireland, we need quickly not just to become AI literate, but AI friendly, because removing a little friction in business always yields better outcomes. Ironic, then, that it is a little friction that got us there – the rubbing of the genie’s lamp.
Further information on Jonathan’s Get Started with AI here.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
