The AI big question: make or buy

Here’s an interesting decision more and more businesses face: whether to acquire AI technology ready-made, or create it in house.

We are still very much in the midst of a growth phase in terms of the adoption of AI in corporate environments: after an initial trial period that attracted huge take-up when algorithms were released by (in order), Dall-E, Midjourney, Stable Diffusion, ChatGPT, Perplexity, Claude, Gemini, etc., more and more businesses have seen the significant gain in productivity derived from largely administrative tasks (writing a letter, using images, creating a spreadsheet or a presentation, etc.), and extend their use when they look at applying them to other more strategic or more relevant areas.

Which is when security issues arise: open models — the vast majority — pose a danger when it comes to introducing corporate information into them, since it may be used later in replies to other users. In response, some companies limit or even prohibit the use of generative algorithms, which carries the risk of restricting innovation. However, there are other options, such as closed installations of bought-in algorithms, or using ones created in-house by companies with secure environments, as Microsoft does — which is logical, given its experience — with Copilot.