# Building the Operating Intelligence Layer

Canonical HTML: https://www.praxis-agents.ai/articles/operating-intelligence-layer

## Humans were the original layer

A few years ago, humans were the Operating Intelligence Layer of a company - using their judgement to turn business context and processes into action in external tools (normally as emails, in shared files and drives, or other specialist platforms).

## Generative AI moved into the middle

Then Generative AI came along, and suddenly a layer appeared between humans and the external platforms that was doing some of the work for them. At first, this was as a separate platform that humans were copy/pasting from, but as AI became "agentic", Large Language Models started to operate the external platforms themselves. But this was happening in silos, with different AI platforms, models and connections being used across one company.

## Big tech wants to own it

Now, big tech companies and model providers are making a play to become the hollistic intelligence layer - tools like Claude CoWork, Microsoft CoWork, OpenAI's incoming "SuperApp" combining ChatGPT and Codex, Perplexity Computer, Google inserting Gemini into Mail, Drive, Docs, Sheets, etc.

## Companies should own their Operating Intelligence Layer

I believe companies should own their Operating Intelligence Layer, not outsource it. Because let's think about what an Operating Intelligence Layer actually is - it's that translation of business context into action that the humans were doing before, and in some cases (especially the big model providers), promising/threatening to do it without the humans. This is totally the wrong way to look at it - what makes any business great, especially service based businesses, are their people and how those people complete their work. By outsourcing this intelligence to external tech companies, many businesses risk losing what made them great in the first place.

## This does not mean ignoring AI

And this is not to say that businesses should stick their heads in the sand and not embrace AI, because they absolutely should - it's here now, it's here to stay and done right, can have a really positive impact on any business. But we are still so early in this new AI era, that the risk of outsourcing this crucial layer and being locked into one providers infrastructure, AI models, processes, and sharing your own processes with that provider is huge. What happens if your entire business runs on Provider A, and Provider B releases a new more powerful model, and all of your competitors can access greater intelligence much more cost-effectively? Or your company is bought by a larger business who operate on Tech Stack A, when you use Tech Stack B and then have to try and migrate everything over?

## What an Operating Intelligence Layer should be

Any Operating Intelligence Layer should be:

- **Flexible** - Not limited to any models, hosting, cloud provider or service, flexibility is key while the technology is changing so fast
- **Extensible** - Not static, allowing businesses to build custom capabilities as platforms and AI models change and evolve
- **Secure** - The risks of "shadow AI" where users are using personal AI tool accounts to complete work are huge and real - sensitive data must not leave company servers
- **Sovereign** - Any Operating Intelligence Layer should be able to function using locally hosted Open Source Large Language models, and not be totally reliant on external providers
- **Collaborative** - The Operating Intelligence Layer must not replace humans, it must increase their capacity to do valuable work that AI simply cannot

## Why I built Praxis Agents

That's why I built Praxis Agents, I had seen all these problems across various companies as I was working to implement AI solutions within them. Praxis Agents solves them and allows businesses to build their own Operating Intelligence Layer. One that they own and implement alongside their poeple, to empower them, not replace them.

## Want to know more?

If you'd like to know more, send me a message or find me at London Tech Week on Wednesday 10th June or Vercel Built In London on Friday 12th June.