Glean: choose what changes near your team’s AI limits

Decide what should change as usage grows

If your team uses Glean Assistant for analysis and planning, ask who should decide how it balances answer quality with usage. Its latest release gives administrators that choice by department, with an option to change settings as people approach their monthly allowance.

Glean’s November 3 release notes describe Auto mode policies called Economy, Balanced and Frontier. Auto mode chooses the AI model and the amount of reasoning for each chat request. In familiar terms, it decides which AI to use and how much effort to spend.

The administrator can assign a policy across the organization or to individual departments. These controls apply where Glean supplies the model access, called Glean Universal Model Key. They are not currently available when the customer supplies their own model account.

Read the announcement: Glean: November 3 release notes and Auto policy changes.

Understand the 75% and 90% thresholds

The optional automatic setting moves people on Frontier to Balanced at 75% of their monthly usage limit. At 90%, it moves all users to Economy. The automatic-change setting applies across the organization.

These are percentages of a person’s usage allowance, not a promise that the company’s total cash bill will stay below a particular amount. Check the allowance and prices in your own agreement before making a budget forecast.

Glean describes Balanced as the recommended default. Frontier prioritizes its most capable models and costs more. Economy prioritizes lower usage, with lower expected quality on complex work such as data analysis and slide creation.

Protect important team work

Our perspective: allowing a quieter setting to take over can be reasonable for routine questions, but it should be a conscious choice for work that informs a customer commitment or business decision. A lower usage cost is useful only if the result remains good enough.

Ask the team lead which tasks need more careful reasoning and what review they require. A person near their limit may need a clearer route to request a different setting, rather than repeatedly trying a difficult question without understanding why results changed.

Glean displays the models used in responses and notes switches during intermediate work. Use that information when investigating a disappointing answer. We have not tested comparative answer quality or measured savings.

Keep the policy in scope

The configuration guide says these Auto policies cover Glean Assistant chat. They do not change the models used by automated agents or by software making requests directly. Those need a separate review.

For teams building business tools around Glean, that distinction matters: changing the chat policy is not evidence that an automatic customer workflow now has the same cost controls.

1. This week, ask the administrator to confirm the account uses Glean Universal Model Key and review departmental policies.
2. Have each team lead compare representative routine and complex tasks under the available policies, using approved information.
3. Decide whether to enable the organization-wide automatic changes and tell staff what the 75% and 90% thresholds mean.
4. Review automated workflows separately, then revisit the policy after the first monthly usage cycle.