---
title: AI needs a data foundation first
description: Most SMEs aren't held back by AI, they're held back by their data. Why the foundation comes first, and the three steps from scattered data to useful AI.
---

[Insights: Data, CRM and AI for SMEs | Vidalico Digital](https://www.vidalico.com/blog)

# [AI needs a data foundation first](https://www.vidalico.com/blog/harnessing-the-power-of-data-infrastructure-for-effective-ai)

 Written by [Peter Tasala](https://www.vidalico.com/blog/author/peter-tasala) | Feb 23, 2024, 6:29:18 PM

*Why most AI projects in SMEs stall, and what to do before you start one*

Every software vendor now has an AI button. Summarise this, predict that, ask anything. For a Swiss SME, the pressure to "do something with AI" is real, and so is the worry of falling behind.

But when we look at AI pilots that quietly stopped, the reason is almost never the AI. It's the data underneath. Customer records spread across Excel, inboxes and two systems. Processes that live in people's heads. Tools that don't talk to each other. An AI tool on top of that doesn't fix it. It makes the mess faster and more confident.

## What AI actually needs from you

Strip away the marketing and most useful business AI does one of three things: it answers questions from your data, it automates a step in a process, or it drafts something based on what it knows about a customer or a case. All three depend on the same things:

- **Data in a known place.** If the answer lives in a salesperson's personal spreadsheet, no AI will find it.
- **Data you can trust.** Duplicates, outdated contacts and free-text fields lead to wrong answers that look right.
- **Processes that are written down.** You can't automate a step nobody can describe.
- **Clear rules on access and protection.** Which data may go into which tool, and who may see what. Under the revised Swiss data protection law, this isn't optional.

None of this is new or glamorous. It's also exactly what separates the AI projects that run from the ones that stop after the demo.

## The real challenge isn't the dashboard

For years, the answer to "we need better insight" was another dashboard or another BI tool. Many companies have several. Still, the question "how are our top customers doing?" takes days to answer, because the dashboard is only as good as the data behind it, and someone still has to know where to look.

AI changes how you ask the question. You can type it in plain language instead of building a report. It doesn't change what the answer depends on. Clean, connected data was the bottleneck for dashboards, and it's the bottleneck for AI.

## Three steps from scattered data to useful AI

**1. Understand.** Map where your data lives, how it moves between systems, and which processes cost the most time. Talk to the people who do the work, not just the org chart. The result is an honest picture and a ranked list of what's worth changing.

**2. Build.** Fix the foundation where it matters: one place for customer data, owners for the records, rules built into the system, the important tools connected. Then add AI where the foundation is solid, starting with one use case that has a clear benefit, like [asking your CRM questions in plain language](https://www.vidalico.com/chat-with-your-data).

**3. Grow.** Measure what the first use case changed, keep the data clean, and only then pick the next one. Small, working steps beat a big programme that never leaves the pilot phase.

## A quick self-check

Before you buy an AI tool, try answering these honestly:

- Can you produce a list of your 20 best customers and what they bought last year in minutes, not days?
- Does every customer exist exactly once, in one system everyone uses?
- Do you know which AI tools your team already uses, and what data goes into them?
- Do you know where your company data is stored, by country and provider?

If you answered no to two or more, the most valuable AI project for you right now is the foundation. That's not a setback. It's the part that makes everything after it work.

Want a fuller picture than the five questions above? Take our free [AI-Readiness Check](https://www.vidalico.com/ai-readiness-check): twelve questions, about three minutes, and you see your score in five areas straight away.

## Where to start

Our [Digitalization Audit](https://www.vidalico.com/ai-and-digitalization-for-swiss-smes) looks at your data, processes, systems, team and data protection, and tells you honestly where AI can help now and what needs fixing first. You get a roadmap you can run with your own team, with us, or not at all.

If you're planning a CRM move as part of this, read [clean your data before you move it](https://www.vidalico.com/blog/clean-your-data-before-you-move-it) first.

**[Start with the Digitalization Audit →](https://www.vidalico.com/ai-and-digitalization-for-swiss-smes)**

[View full post](https://www.vidalico.com/blog/harnessing-the-power-of-data-infrastructure-for-effective-ai)

```json
{
  "@context" : "http://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Peter Tasala"
  },
  "dateModified" : "2026-11-03T08:25:57.379Z",
  "datePublished" : "2024-02-23T18:29:18Z",
  "headline" : "AI needs a data foundation first",
  "image" : {
    "@type" : "ImageObject",
    "height" : 627,
    "url" : "https://2977752.fs1.hubspotusercontent-na1.net/hubfs/2977752/Bannerwhat-are-good-use-cases-for-data-mesh.png",
    "width" : 1200
  },
  "mainEntityOfPage" : "https://www.vidalico.com/blog/harnessing-the-power-of-data-infrastructure-for-effective-ai",
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "height" : 60.0,
      "url" : "https://2977752.fs1.hubspotusercontent-na1.net/hubfs/2977752/Vidalico_Logo_Horizontal_color-2.png",
      "width" : 172.74193
    },
    "name" : "Thoughts & ideas on digital marketing"
  }
}
```