Cohere North 2 Tackles Enterprise AI Agent Cost Control With Memory & Token Spending Caps

Cohere North 2 Targets Enterprise Cost Control and Context Gaps

Cohere released North 2, an overhaul of its enterprise AI agent platform designed to solve two critical problems: runaway agent spending and inconsistent business context1

1

2

2

. The platform introduces shared agent memory, granular token spending caps, and a rebuilt orchestration system that handles multi-step workflows while keeping humans in the loop for critical decisions1

1

. North 2 runs in the cloud, on-premises, or in fully air-gapped deployments, positioning Cohere to compete directly with Microsoft Copilot and Google Vertex AI1

1

.

According to VB Pulse data from July, 21% of enterprises exercise only reactive monitoring of agent spend with no real-time way to stop an agent before costs escalate1

1

. The same research found that 68% of enterprises traced confidently wrong agent answers to missing or inconsistent business context in the past six months1

1

. North 2 directly addresses both issues through its cost control features and contextual memory capabilities.

Rebuilt Orchestration System and Shared Agent Memory

The platform's redesigned orchestration system runs complex multi-step processes autonomously, with humans kept in the loop for designated decisions1

1

. Users build agents and automations from simple prompts, then share them across the organization to prevent duplicate work2

2

. Paul Teyssier, VP of Product AI at Cohere, explained that shared agent memory lets context persist longer across sub-sessions and agents, though he did not provide specific duration figures1

1

.

Reusable agent skills package logic that agents call repeatedly, eliminating the need for employees to rebuild the same capabilities1

1

2

2

. Shared libraries hold knowledge and assets accessible to anyone in the organization1

1

. From a natural-language request, North 2 can produce slide decks, dashboards, or lightweight apps that users prototype inside the chat window2

2

.

Granular Token Spending Caps and Admin Controls

North Admin, the platform's management console, now tracks token use down to the individual user and agent level1

1

2

2

. Administrators can define consumption tiers for users and groups based on request and token rates, set organization-wide caps, and configure alert thresholds that warn teams before limits apply1

1

2

2

. Usage can be broken down by model, user, and department, then exported into a customer's own analytics systems1

1

.

The model-agnostic architecture lets administrators assign models to each team, including ones customers bring from outside Cohere, reducing vendor lock-in1

1

2

2

. Teyssier emphasized that North 2's model, harness, search, and embedding components are interchangeable1

1

. North Admin integrates with existing identity and access management systems and sets roles and permissions across the platform1

1

.

Security Controls and PII Screening

North 2 applies security controls at the individual agent level through North Admin1

1

2

2

. Guardrails scan prompts and responses for PII screening and detect prompt injection protection attempts1

1

2

2

. Autonomy policies limit agents to authorized actions and route critical decisions to humans1

1

2

2

.

Cohere maintains continual security testing including red-teaming and third-party vulnerability scanning, plus detailed logs of all changes and tools for monitoring service performance1

1

. The platform holds SOC 2 Type 2, ISO 27001, and ISO 42001 certifications1

1

. Jacob Rideout, CTO of HiddenLayer, said his team's agent workflows pull untrusted data from marketing, engineering, and security operations, adding they "only turned those workflows on once we could see and control what the agents were doing"1

1

.

Market Position and Early Adopters

In a July agentic orchestration survey of 107 respondents, Microsoft led at 41%, followed by Anthropic's Claude platform at 28%, LangChain/LangGraph at 10%, and OpenAI at 7%1

1

. Microsoft, Anthropic, and OpenAI have added their own spend controls, though these govern usage only within each provider's own platform1

1

.

Bell Cyber, the security operations unit of Bell Canada, uses North inside its own operations2

2

. Jawed Ahmad, Bell Cyber's chief technology and AI officer, said North lets the unit run agentic AI with control and sovereignty requirements "built in from the start"2

2

. LG CNS runs North internally and brings it to customers across South Korea2

2

. Radical Ventures built an automation on North that prepares briefings its team reads before founder calls, replacing hours of manual context gathering2

2

.

Cohere is leveraging its partnership with Nvidia to reduce North's operational costs, saying its models produce more tokens per second per node on Nvidia's Blackwell and Hopper GPUs2

2

. Kari Briski, vice president of generative AI at Nvidia, said this helps North 2 "deliver more tokens for less"2

2

. The platform update arrives less than three weeks after Cohere signed a merger agreement with German AI developer Aleph Alpha GmbH2

2

.