I don’t trust people who predict the future with precision. What I do believe in is the exercise of writing down your expectations at the start of the year and checking them honestly in December. I have spent years working with startups and small companies for equity rather than fees, so when I get a forecast wrong, I pay for it directly. That tends to sharpen your thinking.
Below are ten things I expect to see in 2026. They are opinions, not facts, and certainly not certainties. I wrote them for business owners, and for each one I added what I would actually do if I ran a small or mid-sized business.
1. AI agents move from demos to narrow, measurable tasks
About a year and a half ago I wrote about AI agents as the next step after chatbots. The excitement was huge; the real-world results, more modest. I think 2026 is the year agents stop trying to “do everything” and start doing one thing well: triaging support emails, reconciling invoices, drafting standard quotes.
What to do: pick one repetitive process, measure how long it takes today and put an agent on that alone. If you can’t measure the outcome, it isn’t a project yet. It’s a toy. I saw this clearly in the conversations at AI Expo Europe 2025.
2. AI budgets get audited
After two or three years of subscriptions bought “so we don’t fall behind”, someone in the company will ask the obvious question: what did we actually get? I expect 2026 to be the year AI is treated like any other investment, judged on returns rather than applause.
What to do: list every paid AI tool you use and, next to each one, write a single sentence on what it saves or earns. Whatever you can’t justify, cancel.
3. Small teams with AI compete with much bigger companies
I already see this in the startups I advise: three or four well-organized people with good tools ship what a whole department shipped a few years ago. Size matters less in marketing, content, coding and analysis than it used to.
What to do: before you hire, ask whether your problem is a lack of people or a lack of process. Sometimes the answer is a new hire. Sometimes it’s a tool and an hour of reorganizing.
4. Being found is no longer enough. You need to be cited
More and more searches end with an AI-generated answer rather than a click. I wrote at length about how SEO is changing in the age of AI answers, and my view is that the trend continues. Brands that show up as sources in those answers earn trust; the rest quietly disappear.
What to do: publish clear content with direct answers, verifiable facts and a named author. And build your name beyond your own website: press, podcasts, reviews.
5. The first purchases made by AI assistants on people’s behalf
So-called agentic commerce, where an assistant compares, chooses and even pays for you, is still very early. I don’t think it becomes mainstream in 2026, but I do expect the first real steps to become visible, especially for simple, repeat purchases.
What to do: if you sell online, clean up your product data: clear names, complete attributes, accurate stock and prices, return policies that are easy to read. An AI agent isn’t impressed by pretty photos. It’s impressed by good data.
6. Trust becomes part of the product
Fake content, cloned voices and increasingly convincing fraud are no longer just a political problem. I wrote about deepfakes and how we protect ourselves in an election context, but the same techniques are now showing up in fake invoices, calls “from the CEO” and fabricated reviews.
What to do: set simple verification rules for payments and bank-detail changes, with confirmation on a second channel. And with customers, be open about who you are, what you check and how you protect them.
7. Regulation: the AI Act moves ahead, but simplification is on the table
Obligations under the EU AI Act keep phasing in gradually. At the same time, in November 2025 the European Commission proposed a “digital omnibus” package aimed at simplifying some digital rules and adjusting some AI Act timelines. What the final version will look like, I don’t know, and I doubt anyone can say for sure right now.
What to do: don’t wait for full clarity before getting organized. Map where your company uses AI, especially where it makes decisions about people, and keep an eye on what Brussels decides.
8. In Romania, a tighter fiscal year
After the budget-consolidation measures Romania adopted in 2025 (the government raised several taxes to bring down a large public deficit), the fiscal environment is heavier for many businesses. I won’t go into rates and details, as they change too often to be useful in an article. My point is simple: margins will be thinner and waste will show up faster.
What to do: rebuild your budget on the cautious scenario, talk to your accountant early and look at every recurring cost as if you were approving it for the first time today.
9. Talent: the path for juniors changes
Many of the tasks a beginner used to get, such as research, first drafts and simple reports, are now handled by AI. In my piece on AI and jobs in the companies I work with, I argued that people aren’t disappearing, but the way they learn the trade is changing. I think 2026 is when this becomes a real problem: where will tomorrow’s seniors come from if nobody trains juniors?
What to do: keep hiring people at the start of their careers and teach them to work with AI, not be replaced by it. Companies that train people will have an edge that money can’t easily buy.
10. Back to fundamentals
My last prediction is also the least spectacular. I believe 2026 will reward companies with cash, customers who come back and real profit, rather than growth at any cost. I learned this the hard way in 2009 and wrote about it in my lessons from 20 years of entrepreneurship. Technology changes; the rules of money, not so much.
What to do: track cash flow weekly, find out why your best customers return and build on that before buying growth with expensive ads.
How I’ll check them
I’m leaving these exactly as they are, no edits. In December I’ll come back and grade them simply: happened, partly, not at all. And I’ll write openly about where I got it wrong, because a wrong prediction teaches you more than one you got right by luck.
Until then, I’m curious what you think: which of these ten seems most likely to you, and which do you think is completely wrong?