A Canadian platform for building AI agents at work 🇨🇦

Another interesting announcement today on the Canadian AI front.

Cohere is launching North 2.

Cohere is a Canadian artificial intelligence company based in Toronto.

With North 2, it essentially wants to let companies build their own AI agents to work with their information and their tools.

🤖 What does an AI agent at work actually do?

Imagine your company has:

• email
• documents
• tasks
• projects
• internal procedures
• and several applications

Today, you usually have to open all those different tools yourself to get your work done.

An AI agent can instead receive a task like:

“Get me ready for my meeting with this client tomorrow.”

It could then look at the information you gave it access to, find the relevant documents, and prepare the result.

That is exactly the kind of work North 2 is meant to make easier.

🧰 Agents that use your tools

North 2 can connect to several tools already used in companies.

For example:

• Outlook
• SharePoint
• OneDrive
• Slack
• Jira
• Notion
• GitHub

So the idea is no longer just chatting with an AI.

The AI can start to work with your work environment.

🧠 Agents can also have memory

One especially interesting new piece is memory.

An agent can keep context across different sessions instead of always starting from zero.

Cohere is also adding libraries, where an organization can store shared knowledge and resources.

And Skills—reusable capabilities that different agents can use.

So we are starting to get something that looks more like:

Memory + knowledge + tools + agents + automations

rather than a simple chatbot.

🔄 You can even build automations

North 2 also lets you build processes with several steps.

For example:

1. A new document arrives
2. The agent analyzes it
3. It finds the relevant information
4. It prepares a summary
5. It asks for human approval if needed

So a human can stay in the loop for important decisions.

🔐 Why companies might find this interesting

Giving an AI access to a company’s documents and tools obviously raises a big question:

Where does our data go?

That is one of the areas Cohere emphasizes heavily.

North 2 can be deployed in different environments—including in the cloud, in the company’s private environment, or even in a setup completely isolated from the Internet.

The company can also define what an agent is allowed to do and require a human step for certain actions.

That is especially relevant for governments and regulated industries.

💰 And you need to control the agents

Another problem shows up once you start deploying lots of AI agents.

They consume resources and tokens.

So North 2 lets administrators set limits per user, group, or agent and track usage.

In other words:

You don’t just hand agents the keys and hope everything goes well.

You can control their access, their actions, and their consumption.

🇨🇦 And again: it’s Canadian

That is what makes this announcement especially worth watching here.

A Canadian company is building a platform that lets organizations create AI agents while putting a lot of emphasis on data control, security, and sovereignty.

North 2 can use Cohere’s models, but the platform does not necessarily force a company to use only those—it is designed to work with other models too.

🍳 The simple version

ChatGPT and other assistants have trained us to:

“Ask me a question.”

Platforms like North 2 are gradually moving us toward:

“Give me something to do.”

The agent can have access to memory, knowledge, and tools to complete several steps.

And above all, the company keeps ways to decide:

• what information it can access
• which tools it can use
• which actions it can take
• and when a human must step in

That is an important shift for AI at work.

And it is coming from a company right here. 🇨🇦

Read more

• Official North 2 announcement (Cohere)