Co-founder Nick Frosst reiterates that return on investment is Cohere’s priority.
With the full release of its agentic artificial intelligence (AI) platform, Toronto-based Cohere is hoping to capture more security-minded enterprises.
First released through an early access program to select clients in January, Cohere’s North is a workspace platform that performs tasks when users create custom AI agents through text prompts. The company claims these agents can speed up and automate what it calls monotonous tasks, such as generating reports or summarizing documents.
In a virtual demo for media last week, Cohere co-founder Nick Frosst said North “is a tool for augmenting and automating all of the work you can do behind a computer that you would rather not do.” He added: “That’s really where we think AI is best suited.”
Founded in 2019, Cohere is Canada’s best-funded large-language model (LLM) developer. The company targets enterprise clients with a feature set that is privately deployable and compliant with several international security standards. For example, early client Royal Bank of Canada (RBC) deploys an on-premises finance-specific platform called North for Banking, hosting data in RBC facilities.
“I think there are not many jobs that this automates entirely, and almost every job it can augment to some percentage.”
Nick Frosst
Cohere
Other early clients include LG CNS, Dell Technologies, and telecommunications giant Bell, which is also providing Cohere with Canadian AI infrastructure access through a partnership.
Most of these clients are large-scale operations, Frosst acknowledged, as Cohere wanted to first target customers with the most stringent security requirements. “You can expect to see us moving down,” to smaller businesses with fewer security needs, he added.
Tools within North include chat, search, and document creation. It competes with enterprise offerings from tech giants, like Microsoft’s 365 Copilot and OpenAI’s ChatGPT Enterprise. North connects to popular workplace tools such as Google Drive, Slack, and Salesforce. Agents on North can access and use data from those sources, but Cohere claims it cannot “see” this data when deployed on-premises. For certain tasks, such as sending emails, a pop-up window will seek the user’s input to approve its actions first.
RELATED: Cohere to access Canadian AI infrastructure and new clients through partnership with Bell
Frosst said Cohere’s priority is helping customers see a return on their AI investments. His statements were mirrored by Cohere co-founder Ivan Zhang at Web Summit Vancouver in May, when he told BetaKit that some Cohere customers struggled to show the ROI necessary to justify their spending on AI. The next phase of the AI market, he told BetaKit, will be determining where the ROI is.
A recent report from Canadian venture capital firm Georgian indicated that half of Canadian companies found it difficult to connect AI rollout to ROI. Elliott Choi, Cohere’s director of product, said it’s hard to attribute ROI back to a specific software tool like North, because of how it’s embedded throughout different work tasks.
Recent reports indicate that Canadian AI deployment hasn’t yet resulted in productivity gains. Despite AI adoption doubling in the past year, new data from Statistics Canada shows Canada’s labour productivity increased just 0.2 percent last quarter, while the services industry, which includes technology companies, saw productivity slide by 0.5 percent.
Globally, some studies have indicated mixed productivity gains from AI tools. According to a report from consulting firm McKinsey, a majority of participating enterprises reported cost reductions within business units from using generative AI. The same report noted that more than 80 percent of respondents say their organizations aren’t seeing a tangible impact on enterprise-level earnings before interest and taxes.
RELATED: Amid AI proof-of-concept fatigue, Cohere co-founder urges potential customers to keep the faith and focus on ROI
Frosst added that Cohere’s enterprise customers have struggled to get past the demo phase to extract value from AI products, mostly due to the cost of deployment. Cohere seeks to facilitate this process with North: Frosst said the onboarding process for new clients has gone from months to weeks. He added that each custom model made by Cohere for an enterprise customer is “easier than the last.”
Frosst also addressed concerns about entry-level job displacement, arguing that technology like North is part of an ongoing AI-driven transformation in the job market. He said North is useful for every job behind a computer, but that it “augments” and “automates” a relatively small portion of an employee’s work.
“I think there are not many jobs that this automates entirely, and almost every job it can augment to some percentage,” Frosst said.
Tech hiring in North America has slowed over the past few years, as has the size of the average startup team, according to financial data firm Carta.
Cohere’s North generates revenue through a subscription fee, Frosst said, which is tied to the number of users, not usage. Cohere told investors it would hit $200 million USD in annualized revenue by the end of this year, sources told The Information last week. This would reportedly represent nearly three times its annualized revenue at the beginning of 2025. The Information’s report also said that the AI startup, which was last valued at $5.5 billion USD, is looking to raise between $300 million and $500 million at a $6.3-billion USD valuation. Cohere did not respond to requests for comment on its revenue growth or fundraising plans.
Feature image courtesy Cohere.