AI can do this, AI can do that. Automate this, automate that.

You don't know. You come from another time, where most sales were done face to face. Based on trust. But this new technology is here. The news talks about it, your colleagues praise it, the whole society is obsessed with it. So you can't just stand by. You feel you need it in your business.

So you hear about agencies that help you grow. They don't talk about SEO anymore, that's old school - and you are old school, so you figure they know better. The new word is GEO. Generative engine optimization. They tell you that you need to be findable inside ChatGPT. You listen. You pay a few thousand. Then a few thousand more for the ad budget. Nothing happens. You start to doubt.

This is one version of the "AI does miracles" pitch. It's been sold over and over.

Over the last year it has gotten to most CEOs, especially the ones who built their business the old way. Established small and medium companies feel pushed to grow fast. Faster. Their marketing and IT budgets have never been this big.

The pressure comes from the top too. In larger companies, investors and boards want an "AI strategy". So a company takes a feature it already had, throws money at an AI agency, and puts an "AI-powered" sticker on it. Just to show it's keeping up.

People who spent their whole careers selling face to face are now throwing money at AI projects, treating it like a miracle. Why? Information asymmetry. The big companies that own the models get bigger. The agencies that sell AI as the cure for everything take their cut. For your business, nothing changes.

So, the question is: What should you know before putting AI into your business.

One. AI is only as good as the data you feed it. And the data inside most small businesses is living hell. Scattered across spreadsheets, half of it outdated, no one fully owns it. You can't find patterns in broken data. Or worse, you find the wrong patterns, which AI is also very good at. What the agency does is charge a fat retainer to build a "custom AI strategy" for a business that just needs a clean, organized CRM.

Two. AI is a tool. Most businesses can get value from it on small, repetitive work. Drafting emails, first versions of job descriptions, basic marketing copy, and a human still has to read it and fix it. Turning meeting audio into notes works well. Simple sorting of data works. That's roughly it for now. Here, the best example is Microsoft. They own the software the whole world runs on, they have endless money, and they pushed their AI assistant into every product they have. Two years in, most companies are still stuck running small pilots, and asking if it's worth the price. If the biggest software company on earth can't make AI fast, you should not expect to either. Go slow. In small steps. And don't assume you understand your own processes as well as you think, or that they repeat cleanly enough for a machine to take over.

Three. AI mostly produces more of everything. More text, more drafts, more output. All of it still has to be checked and processed by a person. You are essentially just moving the problem to someone/something else. If you don't already understand your own process clearly, and most companies don't, then all that extra output has almost no value. It just becomes more to clean up.

Four. Think about what it does to your people. There is more and more research on this now, and it points one way. The more someone trusts the AI, the less they check it, so mistakes slip through unnoticed. The work also gets flatter. Everyone feeds the machine similar prompts, so everyone gets back similar emails, similar posts, similar ideas. And AI has a habit of agreeing with you, telling you what you want to hear. A whole team can end up confident about something stupid.

The people selling you AI call themselves growth gurus. In practice they oversell and overpromise until it becomes obvious. They lean on the hype because the hype sells. Sell first, find excuses later.

The companies that build these models just admitted the same thing. In the last weeks, OpenAI and Anthropic both spent billions to build their own consulting. OpenAI even bought a consulting firm outright. They are hiring engineers and sending them to sit inside companies, in person, to make the AI actually work. The hard part is coming, the reality of a real business, data, people, and habits.

People don't want polished, empty, and automated content. Know your customer's name. Remembering what they bought last year. Pick up the phone. Stand behind your work with your own name on it. And please, don't produce slop. 

Sources:
- https://www.axios.com/2026/05/11/openai-deployco-private-equity
- https://thenextweb.com/news/tomoro-openai-deployment-company-consulting
- https://www.webpronews.com/microsofts-copilot-ambitions-collide-with-reality-as-enterprise-adoption-stalls/
- https://samexpert.com/microsoft-copilot-reorganisation-march-2026/
- https://www.livescience.com/technology/artificial-intelligence/using-ai-reduces-your-critical-thinking-skills-microsoft-study-warns
- https://www.youtube.com/watch?v=5yohuMdhUcs
- https://www.youtube.com/watch?v=bn_1Tji35fk