Skip to content
← All posts

Should your business run AI on its own computer? Privacy, cost and when it is worth it

Sooner or later, an owner who is careful about client information asks the right question: "Where does my data go when I paste it into an AI?" The answer leads to a menu of terms (cloud, open-source, open-weight, local) that is more confusing than it needs to be. Here is the plain version.

The terms, translated

An AI model is a very large file of learned numbers, called weights, plus the software that runs it. The difference between the options is who runs it and how much is shared.

  • Cloud AI. You use it through a website or app, such as ChatGPT or Claude. The model runs on the provider's computers, so your text travels to them.
  • Open-weight model. The finished model file is published, so anyone can download it and run it themselves. You get the loaf of bread, not the recipe.
  • Open-source model. Strictly speaking, you also get the recipe: the training code and often the data. In everyday conversation the two terms are mixed up, and most models called "open" are really open-weight.
  • Local AI. An open-weight model running on your own computer, using tools such as Ollama or LM Studio. Your text never leaves the machine.

Three questions that decide it

1. How sensitive is the information?

A draft social post and a client's medical or financial records are different things. For the second kind, local AI keeps everything on your computer. If you use cloud AI instead, read your plan's data terms: whether your content can be used to train models, how long it is kept, and what the business tiers change. I am not a lawyer, so treat this as a prompt to check, not as legal advice.

2. How much will you use it?

Cloud tools charge a subscription or per use. Local AI has no per-use fee, but you pay in hardware and setup time. For occasional use, cloud is almost always cheaper once your time is counted. For heavy, repetitive work, such as processing thousands of documents, local can start to pay off.

3. How hard is the task?

This is where owners overspend. A rocket engine is far more powerful than a car engine, and you still would not use one to get to the shops. The biggest reasoning models are slower and cost more, and they earn it on difficult, multi-step work. For drafting, summarising, tidying a spreadsheet or answering routine questions, a fast everyday model is usually the right choice.

A sensible way to start

  1. Use a good cloud model for everyday work. It is the easiest and the highest quality.
  2. Once you know exactly which tasks involve sensitive or very high-volume data, move those pieces to a local or privacy-focused setup.
  3. Reach for the heavyweight reasoning model only when the fast one clearly struggles.

What going local really costs

  • Local models are generally less capable than the best cloud models, although the gap keeps shifting.
  • You need a reasonably capable computer, and larger models need more of it.
  • You become the IT department. Setup, updates and troubleshooting are yours.

None of that rules it out. It means local AI should be chosen for a specific reason, such as a privacy requirement, and not because it sounds more serious.

AI tools and their features change quickly, so check the current details before you rely on them.

Keep reading

See all posts