How to implement AI in an SME: what large companies have done and what you can do without their budget

LESSON 1 — TECHNOLOGY IS NOT THE OBSTACLE

For large companies, the obstacle was never the tool itself, but rather the process and the culture.

"80% business, 20% technology."

💡 What do you copy:

Before committing to anything, map out the process you want to improve. If there’s no documentation of how it currently works — who does what, with what information, and what decisions are required — any technology introduced on top of that will only exacerbate the chaos.

This step doesn’t cost money. It requires a methodical approach.

LESSON 2 — START SMALL AND SUPERVISED

This is the experience of an industrial group in the packaging sector that has grown through acquisitions.

In order to align their strategy and culture, they set up a network of internal ‘champions’ to identify use cases across different departments and locations.

→ They identified around 200 use cases.
→ They rolled out 80. All of these were supervised assistants, not autonomous agents.
→ Just one year later, with results in hand, they began using autonomous agents for core processes.

Their exact conclusion: the bottleneck was not the tool. It was the cultural change.

💡 What do you copy:

You don't need 200 ideas or a network of champions. You need the same discipline on your scale: choose a process, assign an assistant with human review, measure, and expand autonomy when the data supports it.

Starting with supervision is what makes your team embrace it instead of fearing it.

LESSON 3 — PRIORITIZE BASED ON THE PROFIT AND LOSS BALANCE

Large companies assess each pilot project against internal criteria such as:

→ Alignment with strategy
→ Technical feasibility
→ Impact on the P&L
→ Improved competitiveness

Nothing is developed simply because it is fashionable, but rather for functionality and improvement (whether through cost savings or by generating new business models).

💡 What do you copy?

The key question is simply this: does this process have an impact on my bottom line? The first project is chosen based on the time saved and mistakes avoided, not on what appears on LinkedIn.

 

His admitted mistake, incidentally, was precisely that: implementing technology too soon, purely out of FOMO (fear of missing out).

LESSON 4 — MAKE YOUR OWN TOOLS; DON’T BUY THEM

The big players did not simply accumulate subscriptions. They built an asset:

→ Governance models
→ Repositories of in-house agents, accessible to the whole company
→ Trained teams (Damm trains its entire workforce so that each person can develop agents for their own tasks, with the IT team, of course, overseeing security)

💡 What do you copy?

The principle of ownership does not depend on size. A system built on open-source tools (self-hosted n8n, for example), which is documented and auditable, is owned by your company once the project is complete.

The alternative — a subscription for every problem — ends up creating silos that large organisations set out to avoid.

LESSON 5 — IT CANNOT BE DELEGATED (all)

Although agents decide on tasks, there are three things that remain an internal and non-negotiable responsibility:

→ Security and access control → Data quality → Business continuity

Without well-managed data, no agent can function.

💡 What do you copy?

If you have a trusted IT manager or IT specialist, bring them on board from day one. AI must be introduced through the front door, with proper governance — not as ‘shadow IT’ that someone signed up for using a company credit card.

 

And if you don’t have one: insist that your supplier provides open and auditable systems. That’s your guarantee of independence.

A SME’S PLAN, IN FIVE STEPS

1. Choose a process that impacts your P&L: manual orders, repetitive enquiries, invoice reconciliation, prospecting.
2. Document it exactly as it is today. An afternoon’s work is worth more than any tool.
3. Redesign before automating. Eliminate unnecessary steps. Pure Lean principles.
4. Automate the minimum that works, with human oversight from day one.
5. Measure the hours freed up and scale up. The next process is chosen based on data, not trends.

The difference between Damm and your company isn’t the method: it’s that they implement it with hundreds of people, whilst you can do it with a single supplier and a single process.

The method is free. And it accounts for 80 per cent.

Has your company already chosen its first process? Or are you still waiting to ‘see how this AI thing develops’?