For the last two years, the AI conversation has been about which chatbot is smartest. ChatGPT versus Claude versus Gemini, who answers questions better, who writes a cleaner email.
That conversation is already out of date.
The companies building this technology have moved on to a different fight. It is no longer just about who has the best chatbot. It is about who builds the thing that does the work, not just talks about it. Industry analysts now expect a large share of business software to include specific AI agents by the end of this year, up from almost none twelve months ago.
I want to unpack what is actually happening, because the framing matters less than what it means for a small business in North London. Most of the coverage on this is written for Fortune 500 companies with IT departments and seven figure budgets. You do not need any of that. But you do need to understand the shift, because it changes what is worth your time and what is not.
Three Different Fights Happening At Once
There are three things going on at once in the AI industry right now, and they get blurred together.
The first is model quality. Which AI is best at reasoning, writing, coding. This still gets the headlines, but the gap between the leading models has narrowed enough that it is no longer the thing that decides who wins.
The second is distribution. Who owns the place where you actually start your work. This is why Google can put AI inside Search, Gmail and Android without you ever opening a new app, and why Meta can put it inside WhatsApp and Instagram where you already spend your time. It is also why the EU recently ordered Meta to let rival AI assistants back onto WhatsApp, after Meta had restricted access to favour its own. Whoever owns the road decides what you see on the way.
The third, and the one most people miss, is what kind of product AI is supposed to be. OpenAI’s approach looks like a bet on reach. Keep the free version huge, keep people using it everywhere, and work out the business model later. Anthropic, who make Claude, are betting the other way. Keep it free of advertising, build trust, and sell that trust to people and businesses who do not want a commercially driven assistant near their data or their customers.
Neither approach is right or wrong. But if you are a therapist, a charity, or an independent retailer deciding which AI tool to let near your client information or your customer list, that distinction is worth more to you than which one writes a slightly better paragraph.
From Answering Questions To Doing The Work
Here is the part that actually affects you. The shift happening across the industry is from AI that answers questions to AI that completes tasks.
A chatbot is something you ask. You type a question, it gives you an answer, you decide what to do with it. An agent is different. You give it a job, and it goes and does it. It checks a calendar, sends a message, updates a record, follows up if nothing happens, and only comes back to you when something needs a human decision.
This is not a small distinction. It is the difference between an AI that drafts an email for you to send, and one that drafts it, sends it, and tells you when the reply comes in.
The big platforms are racing to build this at enterprise scale. Systems with many agents working together, orchestration layers, governance frameworks, the works. None of that is relevant to you. What is relevant is the underlying shift. The tools that matter for a small business in the next year or two will not be chatbots you talk to. They will be small, quiet pieces of automation doing one job each, in the background, without you having to ask.
What This Actually Looks Like For A Small Business
Forget the enterprise version of this. Here is what it looks like at the scale you actually operate at.
A hairdresser or beauty salon. Right now, a cancellation costs you a slot you cannot get back. This is the territory of AI voice receptionists and tools that fill your waiting list, products built specifically to answer calls when you are in the middle of an appointment, send reminders, and offer a cancelled slot to the next person on your list automatically. You find out when it is already filled, not when it is empty.
A plumber or tradesperson. Quotes are where most small trade businesses lose time. AI quoting assistants, now built into several job management platforms built for trades, take a job description sent in by text or email and draft a price ready for you to check and send, instead of you doing it from scratch after a full day on the tools.
A small charity. Funding applications and donor follow up messages are repetitive but require accuracy. Several donor management platforms now offer AI drafted thank you and follow up messages triggered the moment a donation comes in, personalised with the amount and campaign, so nobody falls through the cracks during a busy appeal.
An independent retailer. Stock running low on a popular line is something you usually notice too late. Inventory tools with AI powered reorder alerts, increasingly built into platforms like Shopify and the tills many small shops already use, can flag low stock before you sell out and draft the reorder email to your supplier for you to approve.
I have not named specific products here on purpose. Most of the case studies and stats floating around for these tools right now come from the companies selling them, which is exactly the kind of confidently stated, unverified claim I would tell you to be wary of. The categories are real and worth a look. The marketing numbers attached to them are not something I would repeat as fact without checking myself, and neither should you.
None of this needs a developer team or a six figure budget. Most of it can also be built more simply with tools like Make.com or n8n, the same kind of setup I described for therapists in an earlier piece on AI tools for the profession. The pattern is always the same. Find the one repetitive, clearly defined task that eats your time every single week, and let something else do it quietly in the background.
The Risk Nobody Selling You This Will Mention
Every report on agentic AI right now also mentions the failure rate, and it is worth taking seriously. A significant share of agent projects in larger companies fail, mostly because organisations try to do too much at once, or plug an agent into a system that was never built to support it.
The lesson for a small business is the opposite of go big. Start with one task. One clearly defined, repetitive job that you understand completely, so you can tell straight away if the agent gets it wrong. Watch it for a few weeks. Only then think about a second one.
This is the same principle I keep coming back to whatever the AI trend of the month is. The tool is a small part of it. Knowing your own process well enough to hand a piece of it over safely is the rest.
Where This Leaves You
You do not need to pick a side in the AI wars. You do not need an opinion on OpenAI versus Anthropic versus Google’s distribution strategy. That is genuinely not your problem to solve.
What is worth paying attention to is the direction of travel. The next genuinely useful thing for your business probably will not be a smarter chatbot. It will be one small, specific task quietly handled for you, freeing up the hour or two a week you did not know you were losing.
Start there. One task. Not five.
If you want help working out which one task in your business is worth automating first, a one to one AI consultation is the place to start, or come along to one of my free AI workshops for North London professionals.
No jargon. No hard sell. Just a practical conversation about what would actually help.
Get in touch here or email joe@joeai.co.uk.
