Cohere North 2: Enterprise AI Agents That Remember — and Stay on Budget

Cohere's North 2 isn't a new model — it's an enterprise agent platform upgrade built around the two things that stall real deployments: cost governance and memory. Here's what's in it, what's missing, and the questions to ask before production.

Key takeaways

• North 2 (launched Oct 5): redesigned orchestration harness, cross-session memory, reusable Skills and Libraries, plus Applications and Automations builders.
• North Admin is the standout: cost controls, rate limits, user quotas, org-wide caps, per-model/user/department spend views, and alert thresholds before limits hit.
• Model-agnostic (bring your own model), deploys in cloud, on-prem, or fully air-gapped. SOC 2 Type 2, ISO 27001, ISO 42001.
• What's missing: no public pricing, no independent benchmarks, no stated memory duration — and the financial-data connectors are planned, not shipping.
• Bottom line: the architecture direction is right (spend limits + persistent context + deployment choice), but build the business case only against a vendor quote and validated performance.

On October 5, Cohere launched North 2 — calling it the biggest upgrade yet to its enterprise agentic AI platform. It is not a new model. It's an upgrade centered on a redesigned orchestration harness, cross-session memory, and stronger administrative controls. (VentureBeat; Unite.AI)

The release is about controlling how agents operate: what they spend, what context they use, what actions they can take, and where they run.

Cross-session memory and reusable workflow logic

The core of North 2 is an advanced agent orchestration system for complex, multi-step processes — with people in the loop where decisions need a human. (Unite.AI) Three additions stand out:

• Memory. Agents keep context across sessions instead of starting fresh. Cohere says context persists longer across sub-sessions and agents — though it gave no figure for how much longer. (VentureBeat; subagentic.ai)
• Skills. Reusable capabilities agents call instead of rebuilding the same logic per workflow — a way to standardize workflow logic instead of every employee rebuilding it independently.
• Libraries. Shared knowledge and assets that agents and users across the organization draw from.

Cohere also bundles Applications (prototyping decks, dashboards, documents, and apps from natural language inside chat) and Automations (pre-built templates, drag-and-drop workflow builder, real-time monitoring).

North Admin: cost governance before the bill arrives

The most relevant feature for operations teams is cost governance. North Admin gives administrators cost controls, rate limits, user quotas, and organization-wide caps. Token spend is viewable per model, user, department, or agent — with alert thresholds that warn teams before limits are reached — and flow control defines consumption tiers by request and token rate. (VentureBeat)

VentureBeat's trackers supply the context: 21% of surveyed enterprises had only reactive monitoring of agent spend, and 68% had traced a confidently wrong agent answer to missing or inconsistent business context in the prior six months. Two different operational risks — weak cost controls and insufficient business context — addressed by separate governance and knowledge features.

One important gap, noted by VentureBeat: Cohere publishes no price for North — pricing depends on deployment scale, usage, support, and customization — and no performance benchmarks for North 2. The caps can govern usage, but buyers still need a vendor quote to model total cost.

Model-agnostic, and it runs where your data lives

North 2 is explicitly model-agnostic: run Cohere's models or bring your own, with model, harness, search, and embedding components interchangeable — positioned as reducing lock-in. It deploys in the cloud, on-premises, or fully air-gapped. (VentureBeat)

On compliance: SOC 2 Type 2, ISO 27001, and ISO 42001 certifications. North Admin enforces guardrails at the individual agent level — filtering content, validating model responses — while autonomy policies restrict agents to authorized actions and route critical decisions to a person. Cohere says it runs continual security testing, including red-teaming and third-party vulnerability scanning. (Unite.AI; MobileSyrup)

New connectors: Slack, SharePoint, OneDrive, Outlook, Exchange, Jira, Linear, Notion, GitHub. Financial data connectors (PitchBook, Crunchbase, Daloopa, FiscalAI, S&P Global, FactSet) are planned, not shipping — verify before designing a workflow around them.

On the same day, PwC and Cohere announced a global alliance pairing North's platform, models, and retrieval tech with PwC's risk, regulatory, and transformation practices — aimed at moving clients from AI pilots to measurable production outcomes. (Unite.AI) It underlines the work outside the platform itself: workflow redesign, governance, risk assessment, systems integration.

What the launch doesn't tell you

• No price. Nothing public says what a deployment costs. Evaluate ROI claims only against a quoted price.
• No benchmarks. No independent performance figures for North 2. Treat capability claims as vendor descriptions until third-party evaluations exist.
• No stated memory duration. "Context persists longer" is a direction, not a specification.
• Planned vs. shipping. The financial data connectors are planned, not available.

There's enough public information to evaluate North 2's architecture and governance direction — but not to build a business case. Pricing, workload-specific performance, and several implementation details still require vendor validation.

The questions to ask before production

Memory governance. What gets persisted, how long is it retained, who can inspect or delete it, and how do source-system permissions carry through? Answer these before using persistent memory with sensitive workflows.

Visibility, alerting, or enforcement? A cap is only as real as its enforcement. Does crossing an org, user, or agent limit stop execution, throttle it, switch models — or merely alert an administrator?

Usage governance is not total cost of ownership. A cap controls consumption. It doesn't set implementation, infrastructure, integration, support, administration, migration, or human-review costs. Buyers need the full model to calculate returns.

Portability of the workflow. "Bring your own model" and "low switching cost" aren't synonymous. How portable are agents, skills, libraries, prompts, policies, and observability data when the model changes?

Connector depth. A connector is not a production integration. Validate permissions, write access, schema coverage, error behavior, and auditability before designing a workflow around one.

Approval mechanics. Which actions can require approval, who receives the request, what happens on timeout, are approvals recorded?

Auditability. Can you reconstruct what an agent attempted, which tool it called, which data it accessed, what it cost, and where a human intervened?

What this means for your business

Three lessons apply to any AI agent project, with or without North 2:

Put enforceable limits on agent consumption. Someone must approve agent spend — and see it per user, per department, per agent, in real time. Without enforceable limits, usage can exceed the assumptions behind the business case.

Solve consistency before scaling. Standardize the tasks first: document the steps, fix the data sources, then give the agent the script. VentureBeat's tracker reinforces it: 68% of surveyed enterprises traced a confidently wrong agent answer to missing or inconsistent business context in the prior six months.

Make deployment location an explicit architecture decision. Require a documented data-flow and deployment architecture before production use.

The bottom line

North 2 pairs operational controls with agent capability: spend limits, security policies, persistent context, deployment choices, administrative visibility. The direction is right.

But direction isn't a business case. Ask to see the limits, the logs, the data boundaries, the approval controls — and full pricing — before moving an agent workflow into production.

Want agents that stay on budget? Take the free AI audit — a short assessment mapping where AI agents could create value in your business and what the governance, cost, and integration requirements would look like.

FAQ

What is Cohere North 2?

The October 5, 2026 upgrade to Cohere's enterprise agentic AI platform — not a new model, but a redesigned orchestration harness with cross-session memory, reusable Skills and Libraries, Applications/Automations builders, and the North Admin governance layer (cost controls, quotas, guardrails).

How does North 2 control AI agent costs?

North Admin provides cost controls, rate limits, user quotas, and org-wide caps, with spend visible per model/user/department/agent and alert thresholds before limits hit. But Cohere publishes no pricing — you need a vendor quote to model total cost.

Can North 2 run on-premises?

Yes — cloud, on-premises, or fully air-gapped, and it's model-agnostic (Cohere's models or your own). Certifications listed: SOC 2 Type 2, ISO 27001, ISO 42001.

What's missing from the launch?

Public pricing, independent benchmarks, a specified memory-retention duration — and the financial data connectors (PitchBook, FactSet, etc.) are planned, not shipping. Enough to evaluate the architecture direction; not enough to build a business case without vendor validation.

Sources

• venturebeat.com — Cohere's North 2 Puts AI Agents on a Budget and Gives Them a Memory
• unite.ai — Cohere Launches North 2 With Redesigned Agent Harness and Memory
• subagentic.ai — Cohere North 2 Agent Platform
• mobilesyrup.com — Cohere Rolls Out Enterprise AI Platform North 2
• unite.ai — PwC and Cohere Form Global AI Alliance