Building an AI Agency Health Dashboard That Drives Decisions

Your quarterly board call is in two hours and you are scrambling through spreadsheets trying to answer basic questions about your agency's health. Revenue is in QuickBooks. Utilization data is in your time tracking tool. Project status is scattered across Linear boards. Pipeline data is in your CRM. And employee satisfaction data from last month's survey is in a Google Form that nobody has analyzed yet. You spend the next ninety minutes pulling numbers from five different systems, pasting them into a slide deck, and hoping the numbers are consistent. They are not — your revenue number from QuickBooks does not match the sum of your client billings because of timing differences, and now you are not sure which number to present.

This is what running an agency without a health dashboard looks like. You have data everywhere and insight nowhere. Decisions are made on gut feel because assembling the data required for informed decisions takes longer than the decision-making window allows. And by the time you notice a trend — declining utilization, increasing client concentration, growing accounts receivable — it has already become a problem.

An agency health dashboard is not a reporting tool. It is a decision-making tool. Its purpose is to give your leadership team a real-time view of the agency's vital signs so that problems are caught early, opportunities are seized quickly, and strategic decisions are grounded in data rather than intuition.

What to Measure: The Essential Metrics

The hardest part of building a dashboard is deciding what to include. The temptation is to track everything. Resist it. A dashboard with forty metrics provides no more clarity than no dashboard at all.

Organize your metrics into five categories. Each category represents a dimension of agency health, and together they provide a comprehensive view.

Financial Health

These metrics tell you whether the agency is financially sustainable and profitable.

Monthly Recurring Revenue (MRR). The total value of recurring client engagements per month. This includes retainers, ongoing maintenance contracts, and long-term project engagements with predictable monthly billing. MRR is your stability metric — it tells you how much revenue you can count on before closing any new work.

Revenue growth rate. Month-over-month and year-over-year revenue growth. Track total revenue and MRR growth separately. A spike in total revenue from a one-time project is different from MRR growth from new retainers.

Gross margin. Revenue minus the direct cost of delivering the work (primarily salaries and contractor costs for the people doing the work). Healthy AI agencies target 50-65% gross margin. If gross margin is below 45%, you are either pricing too low or staffing too expensively.

Net margin (EBITDA margin). Revenue minus all costs — direct costs, overhead, sales, and administration. Healthy AI agencies target 15-25% net margin. This metric tells you how much of every dollar you earn actually stays in the business.

Cash position and runway. How much cash you have on hand and how many months of operating expenses it covers. Agencies should maintain three to six months of runway. This metric prevents cash flow crises.

Revenue concentration. The percentage of revenue from your top client and your top three clients. If any single client exceeds 30% of revenue, you have a dangerous concentration risk. If your top three clients exceed 60%, diversification should be a strategic priority.

Accounts receivable aging. How much money is owed to you and how long it has been outstanding. Group AR by age: current (0-30 days), 31-60 days, 61-90 days, and 90+ days. Growing AR aging indicates collection problems or client financial issues.

Delivery Performance

These metrics tell you whether you are delivering quality work on time and within budget.

Project health status. A simple red-yellow-green indicator for every active engagement. Green means on track. Yellow means at risk (timeline, budget, or quality concerns). Red means off track (missed milestones, budget overruns, or client escalation). The ratio of green to yellow and red is a key health indicator.

Milestone completion rate. The percentage of planned milestones that are delivered on or before the scheduled date. Track this over time — a declining completion rate signals capacity or quality issues.

Budget variance by engagement. For each engagement, compare the budget (planned hours times rate) against actual cost (actual hours times cost). Negative variance means the engagement is going over budget. Track the trend — a single month of negative variance is normal, three consecutive months is a problem.

Quality metrics. These vary by the type of AI work you do. For ML model development, track model performance against agreed targets. For data engineering, track pipeline reliability and data quality scores. For deployed systems, track uptime and error rates.

Client satisfaction. Measure this formally through regular surveys — quarterly NPS or CSAT scores — and informally through engagement manager assessments. A declining satisfaction trend for any client should trigger immediate investigation.

Team Health

These metrics tell you whether your people are productive, engaged, and likely to stay.

Utilization rate. The percentage of available hours that are spent on billable client work. Target 65-80% for individual contributors. Below 65% suggests you have more capacity than demand. Above 80% suggests your team is overworked and has no time for internal development, training, or recovery.

Billable utilization by role. Break utilization down by role — ML engineers, data engineers, project managers, designers. This reveals imbalances. If your ML engineers are at 90% utilization while your data engineers are at 50%, you have a staffing mismatch.

Employee satisfaction (eNPS). Employee Net Promoter Score measures how likely your team is to recommend the agency as a place to work. Track this monthly or quarterly. A declining eNPS is a leading indicator of turnover.

Voluntary turnover rate. The percentage of employees who leave voluntarily over a trailing twelve-month period. Healthy agencies target below 15% annual voluntary turnover. Above 20% indicates systemic issues with compensation, culture, or growth opportunities.

Open roles and time to fill. How many roles are currently open and how long the average role takes to fill. Growing open roles combined with lengthening time to fill indicates a hiring challenge that will eventually constrain growth.

Overtime and overwork indicators. Track hours worked above standard workweek across the team. Sustained overtime is a leading indicator of burnout and turnover.

Pipeline and Growth

These metrics tell you whether you are building a healthy business for the future.

Pipeline value by stage. The total dollar value of opportunities in each stage of your sales pipeline — initial contact, qualified lead, proposal submitted, negotiation, closed-won, closed-lost. A healthy pipeline has three to four times your target revenue in early stages, tapering to actual target revenue in late stages.

Win rate. The percentage of proposals that result in signed engagements. Track overall win rate and win rate by service line, client size, and lead source. A declining win rate may indicate pricing issues, competitive pressure, or poor qualification.

Average deal size. The average revenue per signed engagement. Track this over time — growing deal size indicates you are moving upmarket. Declining deal size may indicate pricing pressure or a shift toward smaller clients.

Sales cycle length. The average time from first contact to signed engagement. For AI agency engagements, 30-90 days is typical. Lengthening sales cycles may indicate increased competition, more complex decision-making processes, or poor qualification.

Lead source effectiveness. Where are your opportunities coming from — referrals, content marketing, outbound sales, partnerships? Track the volume, conversion rate, and average deal size by source to allocate your marketing and sales effort effectively.

Operational Efficiency

These metrics tell you whether your internal operations are running efficiently.

Revenue per employee. Total revenue divided by headcount. This is a proxy for operational efficiency. Healthy AI agencies generate $150,000-300,000 per employee per year, depending on billing rates and utilization.

Overhead ratio. Non-billable costs (rent, admin, tools, management) as a percentage of revenue. Target 20-30%. Above 35% suggests your overhead is too high relative to your revenue.

Project setup time. How long it takes to set up a new client engagement — from signed contract to first billable work. Reducing this metric directly improves cash flow and client satisfaction.

Internal project completion rate. What percentage of planned internal improvements (tech debt, tool upgrades, process improvements) are completed on schedule? This metric reveals whether your operational improvement goals are achievable or chronically deferred.

Building the Dashboard

Choose Your Dashboard Platform

Your dashboard needs to pull data from multiple sources and display it in a single, accessible view.

For small agencies (under 20 people), a well-structured spreadsheet or a simple dashboard tool like Google Looker Studio works. The data volume is manageable, and the setup cost is minimal.

For mid-sized agencies (20-60 people), a dedicated business intelligence tool like Metabase, Preset, or Mode provides better visualization, automated data refresh, and more sophisticated analysis. These tools can connect to your data sources and update automatically.

For larger agencies (60+ people), consider investing in a full data warehouse approach with tools like dbt for data transformation and Tableau, Looker, or Power BI for visualization. This provides the most flexibility and scalability but requires dedicated data engineering effort to set up and maintain.

Regardless of the tool, the dashboard must be:

• Automatically updated (no manual data entry for primary metrics)
• Accessible to all leadership team members
• Simple enough to understand at a glance
• Drill-down capable for deeper analysis

Data Sources and Integration

Map each metric to its data source and define how data flows into the dashboard.

Common data source mappings for AI agencies:

• Financial metrics: QuickBooks, Xero, or your accounting system
• Utilization and time data: Harvest, Toggl, or your time tracking tool
• Project status: Linear, Jira, or your project management tool
• Pipeline data: HubSpot, Pipedrive, or your CRM
• Employee satisfaction: Lattice, Culture Amp, or survey tools
• HR data: Gusto, Rippling, or your HRIS

Automate data collection wherever possible. Manual data entry is unsustainable and error-prone. Most modern SaaS tools have APIs that can push data to your dashboard automatically. If a tool does not have an API, look for Zapier or Make integrations as a bridge.

Define data refresh frequency. Financial data should refresh daily. Project status should refresh daily or real-time. Pipeline data should refresh daily. Employee satisfaction data refreshes when surveys are conducted (monthly or quarterly). Match the refresh frequency to how quickly the data changes and how quickly you need to respond.

Dashboard Layout

Design the dashboard layout for the leadership team's workflow.

The executive summary view. This is the first thing leadership sees — a single screen with the five to eight most important metrics, each showing current value and trend. Think of it as the agency's vital signs. If everything is green, you can stop here. If something is yellow or red, drill down.

Category detail views. Each of the five metric categories should have its own detail view with the full set of metrics, historical trends, and drill-down capability. When the executive summary flags a financial concern, the finance detail view provides the context.

Engagement-level view. For each active engagement, show revenue, margin, budget status, timeline status, client satisfaction, and team utilization. This view helps delivery managers and account managers monitor their engagements.

Team-level view. For each team member, show utilization rate, current assignment, PTO balance, and any relevant development metrics. This view helps with staffing and capacity planning.

Setting Thresholds and Alerts

Thresholds turn metrics into signals. Define green, yellow, and red ranges for each metric, and set up alerts for yellow and red conditions.

Example thresholds:

• Gross margin: Green above 55%, yellow 45-55%, red below 45%
• Client concentration (top client): Green below 20%, yellow 20-30%, red above 30%
• Utilization: Green 65-80%, yellow 60-65% or 80-85%, red below 60% or above 85%
• AR aging (90+ days): Green below 5% of AR, yellow 5-10%, red above 10%
• eNPS: Green above 30, yellow 10-30, red below 10

Configure automated alerts for red conditions so that leadership is notified immediately rather than waiting for the next dashboard review. A Slack notification when any metric hits red ensures timely response.

Using the Dashboard Effectively

A beautiful dashboard that nobody looks at is a waste of time. Build the dashboard into your management rhythm.

Weekly leadership check-in (15 minutes). Every week, the leadership team reviews the executive summary. Are any metrics yellow or red? Do any trends concern us? Are there decisions we need to make based on what the data shows?

Monthly deep dive (60 minutes). Once a month, review each category detail view. Analyze trends, discuss root causes for any concerning metrics, and set actions to address issues.

Quarterly strategic review. During your quarterly business review, use the dashboard data as the foundation for strategic discussion. The dashboard provides the facts; the QBR provides the analysis and decision-making.

Ad-hoc decision support. When a decision comes up — should we hire another ML engineer? Should we raise our rates? Should we pursue this new client? — check the dashboard for relevant data. Decisions informed by data are consistently better than decisions based on gut feel.

Share relevant data with the broader team. Your team should have access to metrics that affect their work — utilization targets, project status, and team health indicators. Transparency about agency performance builds trust and helps people understand how their work contributes to the whole.

Evolving Your Dashboard Over Time

Your dashboard should evolve as your agency grows and your understanding of what drives success deepens.

Start simple. For your first dashboard, track the five or six metrics that are most critical for your current stage. Do not try to build the perfect dashboard on day one.

Add metrics based on questions. When a leadership discussion surfaces a question that the dashboard cannot answer — "What is our average revenue per engagement by service line?" — add the metric. The dashboard should grow in response to real information needs, not theoretical completeness.

Remove metrics that nobody looks at. If a metric has been on the dashboard for six months and nobody has ever referenced it in a decision, remove it. Dashboard clutter reduces the signal-to-noise ratio.

Iterate on thresholds. Your initial thresholds are estimates. After six months of data, review whether the thresholds are triggering at the right levels. If you never see red alerts, your thresholds may be too lenient. If everything is always red, they may be too tight.

Validate metric accuracy regularly. At least quarterly, spot-check the dashboard data against the source systems. Data integration errors accumulate over time, and a dashboard with wrong numbers is worse than no dashboard at all.

Your agency health dashboard is the lens through which you see your business clearly. It replaces scattered data and gut instinct with structured, timely, and actionable insight. Build it intentionally, use it consistently, and evolve it continuously. The agencies that operate with this level of data-driven awareness make better decisions, respond faster to changing conditions, and ultimately build more sustainable businesses than those flying blind.