Guild Insights Dashboard

• Guild Insights is the admin dashboard for tracking LLM usage, token consumption, and financial spend across an entire organization.
• It surfaces spend by workspace, agent, provider, model, and user, with a KPI strip for high-level metrics and drill-down for the detail underneath.
• Insights is admin-only. Access is restricted to users with the Admin role in the organization.
• The point of Insights is not just observability. It is the first move in an optimization loop: once you can see what an agent is spending, you know where to focus improvements.

Guild Insights is a dashboard built for organization administrators who need to understand how AI is being consumed across their teams. It combines usage data (tokens processed, sessions completed, retries), spend data (estimated USD at list prices), and behavioral signals (cache hit rate, output ratio, per-workspace and per-agent breakdowns) into a single view. Full documentation is at docs.guild.ai/insights/usage.

Insights lives under the Insights tab in the organization dashboard, above Settings. Access is admin-only because the data crosses workspace boundaries and includes cost information most orgs treat as internal.

The philosophy underneath the product is simple. Visibility is not a substitute for optimization, but it is the necessary precondition for it. You cannot cut what you cannot see. Insights makes cost visible in the shapes that let you act on it.

KPI strip

The top of the dashboard displays high-level metrics for the selected time window: total spend, total tokens (prompt + output + cache write), distinct sessions, retries, cache hit rate, output ratio, and change vs. the previous period.

Usage over time

A daily bar chart toggles between tokens and spend. Hovering shows the exact value and date. It is the single fastest way to see whether an organization's AI usage is climbing, flat, or falling.

Drill-down by dimension

From the top-level view, admins can slice usage and spend by workspace, agent, provider, model, and user. Every dimension answers a different operational question: which team is driving spend, which agent is disproportionately expensive, which provider is winning the vendor split, which user leads consumption.

Real-time updates

Data updates in real time as sessions complete. There is no batch delay, so an admin investigating a spike sees the same picture the runtime does.

Finance needs cost attribution

An organization-level total is almost useless for planning. Finance and platform leaders need to know which team spent the money, on which agents, using which providers. Insights answers those questions without requiring anyone to reconcile provider invoices by hand.

Engineering needs optimization signal

A dashboard that only shows total spend does not tell engineering what to fix. Insights breaks spend down to the agent and the session level, so the top cost drivers become obvious targets for optimization. Knowing that a specific agent is 40% of the workspace's spend is what starts the conversation about cheaper models, prompt trimming, and tool-call reduction.

Governance needs visibility across providers

Organizations run agents against multiple LLM providers. Insights presents that across-provider view in one place, so no one has to log into Anthropic, OpenAI, and Cursor consoles separately to answer basic questions about who is using what.

Optimization compounds when you can measure it

The most valuable thing Insights does over time is create a feedback loop. Ship an optimization. See whether it landed. Adjust. Repeat. Every organization that has gotten AI spend under control did it by making the loop tight, not by making the initial guess brilliant.

Access the dashboard

From the organization dashboard, open the left sidebar and click Insights (above Settings). Admin role is required.

Investigate a spend spike

Start on the Usage Over Time chart, toggle to Spend, and scan for the day the spike started. Drill into that day by workspace, then by agent, then by session to find the exact source. Compare the KPI strip's "VS Previous Period" number to a healthy baseline before assuming something is wrong.

Compare provider mix over time

Break spend down by provider and view the trend. Providers whose share is climbing without a corresponding rise in output are the ones worth investigating first.

Export for finance

Downloadable reports let admins hand finance a clean view of AI spend for the month or quarter. Same underlying data, different consumer.

Spend is estimated at list prices

Insights reports spend in USD based on standard list prices. Actual invoices may differ if your organization has negotiated pricing or committed discount agreements with providers. Use Insights as the trend signal and reconcile against provider invoices for the exact number.

Cache Rate is a lever, not just a number

The Cache Rate KPI (percent of prompt tokens served from provider caches like Anthropic Prompt Caching or OpenAI Cached Input) is one of the best signals for how well your prompts are structured. A low cache rate on high-volume agents is usually money left on the table.

Output Ratio catches over-generation

Output Ratio (output tokens as a share of billable input plus output) surfaces agents that generate more than they need to. An unusually high ratio on a summarization agent, for example, suggests the prompt is not constraining response length.

Admin-only access is deliberate

Insights crosses workspaces and includes cost data. Restricting access to admins keeps the visibility appropriate to the role. Individual workspaces still have their own per-workspace views for team-level insight.

Guild is a control plane for AI agents, a place to build, deploy, and govern the agents your teams run in production. Insights is where organizations actually see, understand, and act on the AI usage happening across their teams.

• The full Insights documentation covers every metric, filter, and export.
• AI Spend Explorer is the free, ungated version of the Insights dashboard for teams that want a snapshot before committing.
• Explore the full Guild platform at guild.ai.