Only 5% of AI projects in practice yield a return on investment, according to an MIT study. Successful projects are often characterized by bottom-up experimentation with AI, where the most useful use cases are scaled up into larger projects. However, this kind of free experimentation is becoming increasingly difficult for organizations. Data security is a major concern because AI companies like OpenAI often train on users' chat data. This has already led to several data leaks, such as the recent incident at the municipality of Eindhoven. In addition, regulations like the European AI Act are imposing more restrictions on AI usage. For compliance, organizations often purchase enterprise subscriptions from AI providers such as OpenAI, Google, or Microsoft. However, this comes with high costs. Most subscriptions are around 25 euros per employee per month. There are also growing concerns about dependency on these large companies.
Profound Chats
That’s why we developed Profound Chats. This is an app that lets you chat with various AI models, generate images, record audio, upload files, and search the internet. It also enables collaboration and knowledge sharing within our organization. Instead of five separate subscriptions, we now have the same AI features in one place and maintain control over our company data.
A variety of models is essential
The AI world is highly competitive. Every week, new models are released that excel at specific tasks. No single company has the best models for every task. For example, Anthropic generally has the best models for programming, Elevenlabs for audio, and recently Google for video. These differences are likely to increase. It is therefore important to be able to use the latest models from different providers.
Vendor lock-in creates dependency
Most providers prefer that you remain their customer. As a result, switching is not easy. For example, you may lose your chats or images. With Profound Chats, we manage our own chats on our own server. This way, we can always switch models without losing our data.
Compliance is challenging with big tech
This also enables us to protect our data. We store our chat data securely within Europe and under our own management, with control over retention. No data is stored with the major AI providers, who are often based in the US. It is even possible to process everything within office walls using small local models. This ensures compliance with AI and privacy regulations.
Cost control: away from subscriptions
We are no longer tied to subscription-based licenses. Previously, we had to purchase five subscriptions to use the best models; now, we pay based on usage. This makes experimenting with models much more accessible. Organizations where people use AI less frequently can still use AI models safely at low cost.
A foundation for further innovation
Having our own system is not just an alternative to a subscription to services like ChatGPT. In addition to an external server, Profound Chats can also be deployed on a local server. This offers opportunities for custom integrations with your own systems. For example, automatically making company files available so you can search them with AI, or automatically summarizing meetings. The possibilities for integrations and automation are enormous.
