7 steps to guide your team on AI — and why "just start using it" isn't training

I recently came across two data points that perfectly illustrate what I see in the field: 92% of companies plan to increase their investment in AI over the coming year, but only 1% report actual mature implementation. And only about 30% of large companies are expected to mandate AI training for employees by the end of 2026. This gap, between investment and results, is exactly the gap I see in every implementation project I work on in Israel: there is no lack of budget, nor a lack of curiosity from the team. What's missing is a process.

This is exactly the same point I saw in research I conducted on the Israeli market: the team skills gap has become one of the main barriers to AI adoption — not management, not budget. The problem is that most companies translate "investment in training" to "we sent everyone a link to ChatGPT and told them to try it" — and that’s not training, that’s a prayer.

What actually happens when training is done wrong

From my experience, three things repeat themselves in every small business that doesn't include a full data team:

1. It starts with the business owner, not the employees. If the manager themselves doesn't use AI on a daily basis, there's no chance the team will take it seriously — employees don't adopt tools that management only "approves from above" without touching them themselves. It works much better as a small, real moment ("See? This saved me twenty minutes this morning") than as a mandate from the top.

2. AI doesn't fix a broken process — it just runs it faster. If your price quote takes a week because it passes between three people with no set order, AI won't shorten it — it will just create another version that also passes between three people with no order. The actual first step is to fix the process itself, and only then introduce AI into it.

3. "General" training doesn't work — because there is no such thing as "AI in general." I distinguish between three layers: a basic understanding of what AI does and doesn't do, role-specific training (what your bookkeeper needs is completely different from what your salesperson needs), and finally mastery of the specific data and procedures of the business. A small business doesn't need three separate programs — but it certainly needs to know that one generic training for everyone doesn't prepare anyone for their actual role.

What isn't talked about enough: Documenting knowledge, not just data

The point most relevant to small businesses, from my perspective: AI doesn't know things that only your experienced employee knows — why a certain client always asks for a flexible proposal, which phrasing in an email closes a deal and which deters. If this knowledge stays only in one person's head, the AI the team builds on will lack its most important part. This is the exact same point I wrote about in a previous post on "AI Workshop" — the content looks good, but the real knowledge is missing because no one fed it before it was written.

The practical step here is simpler than it sounds: you don't need a scary "knowledge documentation project" — it's enough to ask your most experienced employee one question ("How do you decide what to answer a client when they complain?") and document the answer, before asking AI to do it for them.

The counter-argument: "Employees are already using AI on their own, without training"

This is a real claim and not just an excuse to avoid a process — in many businesses, employees are already using ChatGPT or other tools privately, without anyone asking. And I think that's not bad, it's a positive sign. The problem is that each employee does it with their own method, without sharing, and without control — meaning the accumulated knowledge doesn't become an organizational capability; it stays with each employee individually. The right step isn't to "forbid and organize," but to channel this existing use into one shared space, so that what works for one employee becomes available to everyone.

What to do tomorrow morning

Without needing a full plan, three steps every small business can start with this week:

1. Fix one process before AI — Choose one process that is slow or confused, and organize it (who does what, in what order) before introducing AI into it.
2. One tool, not five — Choose one platform and stick to it, instead of each employee trying a different tool. Tool fragmentation creates more confusion than capability.
3. One question from the experienced employee — Document one piece of knowledge that only he/she holds, and build the AI usage around this knowledge, not without it.

The gap between the 92% who invest and the 1% who see mature results is not a gap of money. It's a gap of process — and that's exactly the part that is easy to fix, if you start with the right step.