AI-native

AI-native software is built around AI for the main work the product exists to do. A firm may use it to prepare an annual review from client records. A back-office system that works as usual and adds an AI summary button is different: AI helps with one task, but the product does not depend on it for its main job.

Imagine a firm trialling a review-preparation product. For Elena Shah's annual review, it reads the current fact-find, the last review letter and the latest provider statements. It drafts a list of changes and questions, with a link from each point to the document it used. The adviser checks the list and chooses what to raise with Elena. Preparing that list is the product's main purpose, so AI is central to it. The product could still show Elena's documents without AI, but it would no longer do the work the firm bought it to help with.

The list can help the firm prepare for the meeting, but the adviser checks the linked documents before treating any point as a change in Elena's circumstances. Calling the product AI-native says nothing about which records its AI may open, whether it can change them or whose approval it needs. Those are permissions to check in the product itself.

Not to be confused with

• AI-enabled software adds AI to particular tasks in a product whose main work can still be done without those features. Vendors use both labels loosely, so a firm should ask what the AI actually does.
• A company may call itself AI-native because AI is central to its business. This entry describes a product, not its vendor.

Overheard

Where it sits

AI-native

• works withAI-enabledHarnessPermission