Reviewing AI Usage

The Usage page in Settings shows a full picture of how your organization has consumed AI features — credits spent, number of calls, and a per-call activity log — so you can understand where your balance is going and adjust accordingly.

Prerequisites#

Switching between periods#

At the top-right of the page two tabs let you switch between This month and Last month. All cards — the summary tile, the breakdowns, and the activity log — update to reflect the selected period.

When you are viewing This month, the summary tile shows a trend pill (+N% vs last month, -N% vs last month, or flat vs last month) that compares this month's credit spend against the previous month. The pill does not appear when you switch to Last month.

Summary tile#

The summary tile at the top contains two metrics side by side:

• Credits charged — the total credits consumed during the period.
• Calls — the total number of AI calls made, followed by the count of unique feature types and model tiers used.

Breakdowns#

Three breakdown cards give you different views of the same spend.

By reason#

The By reason card shows how credits (or estimated cost in BYOK mode) split across feature types. The three reasons you will see today are:

Each pill shows the label, the credit or cost value, and the call count for that feature.

By user#

The By user card lists each team member's share of the total. A colour-coded bar at the top gives a proportional overview; the list below breaks down individual totals.

By default only the top 5 contributors are shown. Click Show all (N) to expand the full list, and Show top 5 to collapse it again.

If a member has since been removed from the organization, their calls remain in the log and appear as Unknown user.

By model#

The By model card shows a horizontal bar chart comparing credit (or cost) consumption across each AI model tier used during the period. Each bar is labelled with the provider name and model tier. The longest bar is always the highest-consuming model; other bars are scaled proportionally.

Activity log#

Below the breakdowns, the Recent activity table shows every individual AI call during the period, newest first. Columns visible on desktop:

When your organization is on Intervy credits, the model that ran each call is an internal implementation detail and is not shown. The Model column appears only when your organization is configured for BYOK — see BYOK mode below.

The table loads up to 50 calls at a time. When more exist, a Load more button appears at the bottom. Once all calls for the period have loaded, the table shows "End of activity for this period" instead.

BYOK mode#

When your organization has BYOK active, Settings → Usage shows a notice that usage tracking is not available. Because Ivy routes calls directly to your provider, Intervy does not receive the data needed to build a usage log. The page shows a link to your provider's own usage dashboard — Anthropic Console for Anthropic keys, OpenAI Platform for OpenAI keys — where the full token and cost breakdown is available.

Tips for reducing credit spend#

• Chat is the largest driver for most teams. If the By reason card shows Chat at the top, consider asking Ivy more targeted questions with more context, which reduces the number of back-and-forth rounds.
• Check the By user breakdown. If one team member accounts for a disproportionate share, review what tasks they are using Ivy for — they may be running workflows that could be automated differently.
• Use a lighter model tier where quality permits. The By model breakdown shows which tiers are in use. If Haiku or a faster model meets the task requirements, switching the default in Settings → AI Provider reduces cost.
• BYOK removes credit spend entirely. If your team's usage is growing, connecting your own Anthropic or OpenAI API key means Intervy credits are no longer consumed. See the BYOK section in the Billing and Credits doc.

Permissions reference#