For small and medium-sized businesses, data confidentiality is often a top priority.
Customer documents, contracts, internal procedures, commercial data, and technical information cannot always be sent to an external cloud platform.
A local AI such as Bonsai can therefore serve as an interesting technological foundation.
It can be installed on a company-owned server and connected to business tools designed around the company’s actual needs.
For example, it could be used to create:
- an internal assistant capable of searching through company documentation;
- a technical support assistant;
- a sales assistant connected to the CRM;
- a system capable of summarizing customer interactions;
- a search interface for internal procedures;
- a copilot for administrative or technical teams;
- an agent capable of preparing certain actions without sending data to an external service.
The goal is not simply to install a local chatbot.
The real objective is to build a solution that fits the company’s processes, access rules, documents, and business requirements.
A small business could therefore have its own private artificial intelligence, hosted on its own infrastructure and connected only to approved tools.
Bonsai may not be able to handle every request perfectly today. Its performance can be limited when dealing with highly complex tasks, advanced reasoning, software development, or workflows involving multiple tools.
However, this is not a permanent obstacle.
Compressed AI models will continue to improve. Their quality, speed, and ability to interact with business applications will increase with each new generation.
For a company, starting today can therefore also be a way to prepare for this evolution.
Bonsai illustrates an important shift in artificial intelligence. The future will not depend solely on increasingly large models, but also on models capable of running with fewer resources.
An AI with 27 billion parameters that can fit into just a few gigabytes is already impressive, even if some of its capabilities are reduced during compression.
Its limitations are real, particularly when it comes to complex reasoning, advanced coding, and autonomous agents. But these limitations should not hide the most important point: Bonsai makes a category of models accessible that was previously reserved for very powerful machines.
For a small or medium-sized business with strong confidentiality requirements, this approach can represent a genuine alternative.
It can also be useful for repetitive tasks that do not require advanced reasoning but can still improve working conditions and productivity.
A local AI can be installed on the company’s own infrastructure and connected to business applications developed specifically around its activities.
The company can therefore retain greater control over its data, users, and processes.
Bonsai may not yet be the perfect local AI.
But it clearly demonstrates that compact AI is possible and that it could become a serious alternative to centralized cloud services.
To discover the technologies needed to build a local AI environment, our Local AI series gradually introduces the models, tools, and methods required to create your own AI infrastructure, from the first experiments to more advanced business applications.
