OpenAI, Anthropic, Google and Microsoft build the AI. We make it pay in your business.
We pick the model, wire it into your systems and put the checks in. Then we stay on to look after it.
| Platform | What it is | What we do with it |
|---|---|---|
| OpenAIGPT models, ChatGPT for business and the OpenAI API. | GPT models, ChatGPT for business and the OpenAI API. | Assistants that draft, summarise and answer from your own documents, and agents that take on routine work with a person signing off. |
| AnthropicClaude models and Claude Code. | Claude models and Claude Code. | Careful reasoning over long documents, customer-facing assistants and automated work that has to be right. Our own enquiry replies run on Claude. |
| GoogleGemini and Google Cloud’s Vertex AI. | Gemini and Google Cloud’s Vertex AI. | AI inside Google Workspace, and models run on Google Cloud next to data you already keep there. |
| MicrosoftMicrosoft 365 Copilot, Copilot Studio and Azure OpenAI. | Microsoft 365 Copilot, Copilot Studio and Azure OpenAI. | Copilot set up properly across Outlook, Teams and your files, and custom Copilot agents connected to your business systems. |
| Amazon Web ServicesAmazon Bedrock and AWS infrastructure. | Amazon Bedrock and AWS infrastructure. | AI and the systems around it hosted in your own AWS account, with the security and cost controls set up by Ricky, the AWS specialist on our team. You own the account and the data. |
| Open-weight modelsLlama, Qwen, Mistral, Gemma and DeepSeek. | Llama, Qwen, Mistral, Gemma and DeepSeek. | Good models you can run on a server you own or in a private cloud. There are no per-use fees and the data stays with you. |
| Model routingOne gateway in front of many models, such as OpenRouter or our own router. | One gateway in front of many models, such as OpenRouter or our own router. | Each job goes to the cheapest model that does it well, falls back automatically if a provider is down, and every call is logged and costed. |
| Your own kitWorkstations and servers with NVIDIA GPUs or Apple silicon. | Workstations and servers with NVIDIA GPUs or Apple silicon. | For data that can’t leave the building. Nothing goes to an outside provider, and we benchmark it on your machine before you commit. We run the same set-up for our own work. |
Product names are trademarks of their owners. Revitt is an independent company and is not affiliated with or endorsed by these providers.
How we choose
We don’t resell any of them
We pick the model that does your job best for the money. When a better or cheaper one comes out, and they do every few months, we can move you over.
The AI is the easy part
The vendors sell the model. Making it pay means fitting it to how your business actually runs and wiring it into your systems, then looking after it when the model changes underneath you.
Controls before anything goes live
Personal data is kept out of the model where it should be. Anything that sends, changes or deletes something waits for a person, and every step is logged.