Microsoft Patents an AI That Suggests Object Names Based on Context
Renaming things in software sounds simple, but in large applications it is one of those tasks that eats hours. Microsoft has filed a patent for an AI that reads what an object actually does before suggesting what to call it.
What Microsoft's AI object-naming feature actually does
You're deep in a project, looking at a list of objects with names like "Object1," "Button3," and "Panel_temp," and you have no idea which is which anymore. This is a real problem for anyone building software, designing an app, or even working in a tool like PowerPoint, where complex files can hide dozens of individually named elements.
Microsoft's patent describes a system that looks beyond the name itself. Instead of just offering a text-box where you type something new, it reads the object's attributes (what it looks like, what it does) and its place in the pecking order of other objects around it. It then sends all that context to an AI model, which generates a list of meaningful name suggestions you can pick from.
The idea is that the AI understands what the object is for, not just what it is called, so the suggested names actually reflect the object's role in the bigger picture. You pick the one that fits, and the application updates the name in memory everywhere that object is referenced.
How the AI reads object relationships to pick a name
The patent covers a method that works in three stages.
- Context gathering: The system pulls two things from memory: the current name of every named object in the application, and "context information" for each one. That context includes object attributes (size, type, behavior, visual properties) and an object hierarchy, meaning a map of how objects relate to each other, which is a parent, which is a child, which sits inside or alongside which.
- AI model input construction: When you want to rename a specific object, the system packages at least a portion of that context into a prompt for a generative AI model (think of it as a large language model similar to GPT). The prompt asks the model to suggest one or more names that make sense given the object's role and surroundings, not just a generic label.
- Suggestion and selection: The AI returns candidate names. The application presents them to you. You pick one, and the system replaces the old name stored in memory with your chosen suggestion, completing the rename.
The claim is intentionally broad: it covers any "application" that stores named objects, which means this could apply to code editors, UI design tools, spreadsheet apps, presentation software, or anything else where programmers or power users work with collections of named elements. The key technical novelty is that the AI input explicitly includes the hierarchical relationship data, so the model can infer context from structure, not just from the object in isolation.
What this means for developers working in large codebases
For a developer or a power user working in a complex file, bad object names are a constant, low-grade source of confusion. You open a project someone else built, or one you haven't touched in months, and the names tell you nothing useful. Fixing that by hand, across dozens of objects, is the kind of work people skip until it causes a real bug. A system that reads the structure and offers names that actually mean something could make that cleanup fast enough to actually happen.
The patent is also a signal about where Microsoft wants AI to sit inside its productivity and developer tools, embedded in the editing workflow rather than bolted on as a separate chat window. Microsoft's filing joins a growing body of software AI patents tracked across Big Tech patent news, where the clearest trend is AI moving from standalone assistants into the small, repetitive decisions that fill a working day.
This is the 20th Microsoft filing we've tracked in our AI vision coverage since May, adding to earlier applications like spotting people in video and reading photos aloud for blind users.
The person who feels this patent is the one staring at a screen full of labels like "Button3" and "TextBox7," knowing they should rename everything before handing the project to someone else, but not doing it because the effort is tedious enough to skip. This system makes that moment of discipline take thirty seconds instead of thirty minutes.
What keeps it from being a shallow convenience feature is that the AI sees the whole family tree of objects, not just the one you clicked on. A button nested inside a form inside a panel gets a name suggestion that reflects where it actually lives, which is how a human expert would name it.
The concrete payoff is fewer handoffs that break down because nobody can read the original author's naming logic. That is a real, recurring cost in any team building software together, and shortening it matters.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
10 drawing sheets from US 2026/0252314 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →