On-premise, EU cloud or public AI: what fits an accounting firm?

There are three ways an accounting firm can put AI to work on client files. Each is a reasonable choice for some firms. Here is what each means in practice.

Public AI tools

General assistants such as ChatGPT or Copilot, used through a browser or an office suite.

• Strengths: cheap or free to start, nothing to install, good at general writing and questions.
• Weaknesses: client data goes to an outside company, often outside the EU; you depend on its terms for storage and training; it does not know your firm's clients, deadlines or rules; you cannot show afterwards what it did with which data.
• Fits: questions that contain no client information.

AI in an EU cloud

Software from a provider that processes data in European data centres, often built into accounting packages.

• Strengths: data stays in the EU; the provider looks after the infrastructure; often integrated with one package.
• Weaknesses: client data still leaves the office and sits with a processor you must vet and contract; confidentiality depends on the provider; usually limited to that provider's own software.
• Fits: firms comfortable with a cloud processor and working mainly in one package.

On-premise AI

The AI runs on a machine installed in the firm's own office.

• Strengths: client data never leaves the building; no outside processor; every action can be recorded and shown; it can work across all the software the firm uses.
• Weaknesses: a machine in the office, which someone has to maintain; a higher starting cost than a subscription to a public tool.
• Fits: firms for whom confidentiality is a professional duty, not a preference, and who want the routine work done, not only questions answered.

Where Leyoda fits

Leyoda takes the third route and removes its usual weakness: we install the machine, maintain it and support it, for one monthly rate. A person approves every step Bertha prepares. See what it costs, or see if it fits your firm.