What is generative AI?

Most business software follows rules someone wrote. Earlier AI mostly classified and predicted: is this transaction fraud, how many will we sell next month. Generative AI writes. Ask it for a product description, a reply to a customer, a summary of a contract or a first draft of some code, and it produces one. For text, the technology behind it is the large language model, the engine in ChatGPT, Claude, Gemini and Microsoft Copilot.

An example. A wholesaler adds a few hundred new lines from a supplier, each with a spreadsheet row of specifications and nothing else. Generative AI can draft a description, the key attributes and a translation for every line from the data in the PIM, in the wholesaler's own tone, for the team to check before anything goes live.

Where it earns its place

Generative AI is at its best on reading and writing at scale: drafting product content, answering questions from your documents with RAG, reading emailed orders and specifications, summarising accounts and cases, and powering assistants on storefronts and in customer portals. Put it to work in steps, with access to your systems, and it becomes an AI agent.

For numbers and rules, such as demand forecasting, pricing and anomaly detection, specialised models and plain code are the right tools, and the strongest systems use each where it fits.

What it needs to do well

• Your own information. A model knows what it learned in training. To write about your products and customers, it needs to be given them.
• A check before it matters. Generative AI writes what sounds right, which is usually, but not always, what is right. That's called an AI hallucination, and good systems ground every answer in your sources and check it before a customer sees it.
• The right terms. Business versions of the main tools can be set up so your data isn't used to train anyone's model.

How we do it

We use generative AI where it pays: product content drafted from your PIM and checked before it publishes, and assistants whose replies are grounded in your content or the customer's own account and checked before the customer sees them.

Why letting staff use generative AI tools on the right terms is safer than blocking them: Why blocking ChatGPT at work makes the data risk worse, from Enceladus AI, the PureNet Group's AI strategy company.