AI, data & systems

Short answer: how does this all fit together?

AI is the umbrella term: technology that carries out tasks that would normally require human intelligence. Machine learning is the underlying technique, where systems learn from data. Generative AI is the application that produces text, images or code. AI agents chain steps together — researching, summarising, preparing something for review.

None of that delivers much without good data. A CRM keeps track of what happens with customers and leads; an ERP keeps business processes connected. Those systems are the fuel: the cleaner your data, the more useful anything built on top of it becomes.

• Understanding: what AI is, how it learns, and where it goes wrong (hallucinations)
• Application: agents that take over repetitive work, with a check afterwards
• Foundation: a CRM for customer data, an ERP for processes
• Framework: regulation classifies applications by risk level

1. What AI is — and what it isn't

Artificial intelligence isn't a thinking machine. It's software that recognises patterns in data and applies those patterns to new situations. A language model predicts, word by word, what's likely to come next; that produces fluent text, but not verified facts. That's why a model can confidently state something that simply isn't true — a hallucination.

All AI in use today is narrow AI: specialised in a specific task. General, human-like intelligence doesn't exist yet. Read more in our explainer on artificial intelligence.

2. AI agents: where they actually work

An AI agent carries out a sequence of steps on its own, instead of returning a single answer. That works well for repetitive work with a clear start and end: classifying enquiries, summarising conversations, drafting a first version of content, enriching data.

It works badly when the outcome is legally, financially or medically binding without review, or when the information it needs isn't recorded anywhere. In those cases, put a person in the loop who checks the output. See AI agents in practice.

3. Getting your data in order: CRM and ERP

A CRM (Customer Relationship Management system) records who your customers are and what happens with them: contact moments, enquiries, deals and follow-up. An ERP (Enterprise Resource Planning system) connects business processes — stock, purchasing, finance, production — in one system.

In practice, the quality of this data determines how useful reporting and automation actually are. Messy fields and duplicate records produce unusable dashboards, no matter how good the tooling is.

Read more in our explainer on CRM and our explainer on ERP.

4. How to get started without breaking things

Start with a recurring problem that costs time, not with the technology itself. Test on a small scale, with success criteria agreed in advance, and decide upfront which data may and may not go into an AI tool. Only scale up once the pilot has demonstrably delivered something.

• Pick one concrete use case with a measurable outcome
• Agree who checks the output before it goes out
• Decide which data may be used (privacy rules still apply in full)
• Work out which risk category each application falls under

5. AI in marketing

Within marketing, AI mostly speeds up execution: drafting variants, segmenting audiences, summarising data, producing first drafts. The strategy underneath doesn't change because of it. Starting without clear positioning and messaging just means producing mediocre content faster.

Read more about content marketing.