The Future of Company AI Systems

After a couple of years of rapid experimentation, the AI landscape is beginning to settle into a clearer structure. Companies, technology providers, and emerging standards are shaping a future where different AI systems play defined roles — and the value comes from how they connect. The signals that I see are coming from three places: how companies are using different AI systems, the technology providers are building, and different standardisation efforts currently underway.

In my view, four main pieces make up this AI puzzle: 

1. Task specific AIs
2. Company level AIs
3. Domain specific AIs
4. Connection protocols

Let’s take a look at each of these.

Task AIs

These are highly specialised AIs that excel at certain tasks — think coding assistants, writing assistants or search assistants. Sometimes they exist independently, other times they’re embedded into role-based AIs. Because they focus on execution, these systems can be viewed as tools — or as agents, depending on your perspective.

Company AIs

A company AI, in my mind, is any central system that allows connections to other AIs. It may not hold any data itself — it could just route connections to other systems or ingest raw data for retrieval-augmented generation (RAG) to answer questions. The AI can trigger actions, but these actions are limited, and rely on the role of task specific AIs to complete the actions.

Company AIs may not exist yet. But I do see systems with this vision beginning to emerge: Google Agentspace and enterprise search players like Glean are early examples.

Domain AIs

The domain AIs focus on narrower use-cases, for example, marketing, HR, or sales. You find them inside applications like HubSpot, Workday, or Happeo. Unlike company AIs, these systems typically own the data and feed it upwards through indexing APIs, tools, or agents. 

Connection protocols

For these systems to work together, they need to follow established protocols and processes, such as Model Context Protocol (MCP) and Agent2Agent Protocol (A2A). These are just now emerging so we’ll see if there are going to be the ones the world goes with or some variation. They do however signal the need for a common communication in this hierarchy of AIs.

What excites me the most is that these protocols make different AI systems interchangeable. If you find a new and better company AI, it is going to be somewhat “easy” to switch to another provider and still plug and play the role specific AIs into the new system.

Where Happeo fits

I see Happeo as being very good at managing and sharing trusted knowledge at both company and department level. We ensure that knowledge is available, detect what's missing, and give people the means to share it. 

For some customers, I believe we can evolve into the company-level AI system once we build more capabilities. But for most, we will fill the role of a Domain AI that provides company and department-level trusted knowledge. And because we have a strong engineering team, we’re already building impactful AI features that help our customers succeed.

How to utilise Happeo in your AI systems

Right now, we’re giving customers powerful ways to plug Happeo into their AI stack:

• Indexing API:  it allows copying Happeo data to the customers own RAG systems.
• Robust API: allows, for example, to utilize Happeo’s AI search to provide answers to connected systems.

An interesting capability that we are considering to provide is an MCP server, which exposes a lot of Happeo’s tools — from search to notifications to help center — and make it possible to build those capabilities directly into different AI agents.

What’s next

My goal is to keep Happeo future-proof and let our customers use it in whatever way fits their AI strategy best. If you’d like to try these functionalities, I encourage you to test out Happeo and reach out to our support team to get access to experimental functionalities like the MCP server.