AI for businesses: the challenge is 80% business and only 20% technology.

1. WHAT AI CAN DO FOR YOUR COMPANY

Forget generic promises. These are real examples of what a well-designed AI system does in a company today:

→ Sales and prospecting: researches target companies, prepares personalized emails with real information about your company and research on each prospect, and manages follow-ups that currently die in inboxes due to lack of time.

→ Customer service: categorizes each inquiry, responds in the customer's language to those that don't require expert advice, and escalates those that do to your team.

→ Administration: extracts data from invoices, orders, and delivery notes, matches them, and leaves only the exceptions for your team. If you work with systems like Holded or HubSpot, they update automatically, and you monitor them.

→ Operations: converts orders received by email into ERP records without anyone having to type them in.

Notice the pattern: in all cases, AI prepares, and your team decides.

THE TYPES OF AI THAT COMPANIES USE (AND WHEN TO USE EACH ONE)

Another recurring search: "What kind of AI do companies use?" These are the different levels of AI being implemented in companies today:

Level 1 — Fixed Rule Automation. If X happens, do Y. It's not AI, but it handles a huge portion of repetitive work. It's the best starting point.
Level 2 — Supervised Conversational AI. Understands natural language: customer inquiries, emails, documents. With human review for sensitive matters.
Level 3 — Tooled Agents. Systems that execute complete processes — search, extract, draft, record — within defined limits and rules.
Level 4 — Predictive AI. Anticipates demand, maintenance, and pricing. Requires a well-organized data history. It's almost never the first step.

💡 The key:

Most SMEs achieve the highest return by combining levels 1 to 3 on one or two specific processes. Implementing AI where a fixed rule was sufficient is a waste of money.

3. THE NEW QUESTION: CAN YOUR CLIENTS' AGENTS FIND YOU?

There's a reason this article mentions what people ask AIs (like Claude or Gemini), and not just Google:<br />
<br />
<strong>Increasingly, the person searching for, comparing, and shortlisting a supplier isn't a person. It's their AI agent.</strong><br />
<br />
Questions like "Where can I hire AI services to improve business logistics?" are now being asked directly to an assistant.<br />
<br />
Traditional SEO is no longer enough. Your company needs to be discoverable—and responsive—to agents as well. Responding quickly and with real data is what makes the difference when the question is being asked by a machine that compares ten suppliers in seconds.

What you can do right now:

• Organize your public information: catalog, deadlines, capacities, indicative prices
• Publish content that answers your customers' direct questions (in metadata or organic posts)
• Measure how long it currently takes you to respond to a request for a quote

It's a silent change and it's happening faster than you might think.

4. HOW TO START WITHOUT MAKING MISTAKES

From that same roundtable discussion, the mistakes that large companies acknowledge having made were identified. So you don't repeat them:

→ Don't implement out of FOMO. Technology implemented prematurely, without a proper process behind it, is the mistake Celsa admits to. Prioritize based on real impact on your bottom line.
→Don't start with autonomous agents. Those leading the way began with supervised assistants and expanded autonomy when the data supported it.
→ Don't accumulate black boxes. Each AI tool or program contracted separately is another subscription, another silo, and your data in a system that no one internally maintains. The alternative: build a system on open tools that your company owns—auditable, documented, and yours when the project is finished.

IN SUMMARY

The best AI for your business isn’t a product.

It’s the result of mapping out a process, redesigning it and applying just the right technology — with your team overseeing what matters and with ownership of the system firmly in your hands.

That’s why at KAIzen we don’t sell AI. We design processes that run themselves.

What about your organisation? Are you still trialling AI ‘in a corner’, or are you already embedding it at the heart of your processes?