|
As AI search startup Glean moves into competition with rivals like Anthropic in automating white-collar tasks, its executives are zeroing in on two topics keeping corporate chief information officers awake at night: Token costs and data security.
Glean last week shared internal evaluations showing its AI assistant is much cheaper to use than Anthropic’s Claude Cowork app in performing more than 180 business tasks. Glean says that’s because its tools help customers such as Intuit, Booking.com, and Zillow to develop a so-called context layer that helps agents quickly find data they need to complete tasks, including an “enterprise graph” that helps the AI understand connections between corporate data.
For instance, Glean’s AI assistant, powered by models such as Anthropic’s Claude Opus 4.8, used 70% fewer tokens to accomplish the same tasks as Anthropic’s Claude Cowork application, which was running Claude Sonnet 5 with high reasoning, according to a Glean spokesperson. (Anthropic, on its website, says Sonnet 5’s performance is “close to that of Opus 4.8” but costs less.)
At any rate, Glean says its assistant charged customers an average of 81% less than Claude Cowork on a per-task basis, in part because the assistant also used Glean’s built-in software that routes tasks to different models, depending on the level of complexity.
An Anthropic spokesperson didn’t have a comment.
Glean CEO Arvind Jain, whose company was last valued at $7.2 billion after its last financing in June last year, says the context layer can help customers avoid the cost of AI conducting lengthy searches to find the right data.
“AI was not that expensive last year because it was not doing complex things. But now, [AI] is doing complex things for you, and that's why it runs for a long time and costs a lot,” he said. The company announced in May that it had surpassed $300 million in annual recurring revenue, the value of its contracts over the next 12 months.
Glean’s focus on AI prices shows how the seven-year-old startup plans to compete in the crowded market for AI agents for businesses, where it’s going up against major cloud providers as well as Anthropic and OpenAI.
At the same time, Glean says its agents can operate in business settings with the same level of security and data protection that companies typically use for their employees.
In June, Glean launched agents that serve as virtual coworkers, including a Slack-based agent that does the work of an on-call support engineer whose job is to ensure a company’s applications stay up and running. When an application does go down, the agent gathers details to troubleshoot what went wrong and figure out options for fixing it.
Glean is also taking steps to make its agents safer for companies to use, like giving each agent its own identity and authentication credentials to ensure that they’re only accessing data they are authorized to work with, according to Jain.
“If you have these agents that are running for a long time, and they have a lot of control and capability to take actions in your systems and do things, they can easily go haywire,” Jain said.
|
Gift Subscription
