HomeAI Agency AcademyLesson 38
Module 10 · Lesson 38

Manage margins, capacity, and usage

Understand labor, provider usage, support, and service limits before scaling delivery.

Last updated August 5, 202615–25 minutesFree AI agent course
What you will learn

Model margin and capacity

You will be able to model the service capacity and margin of an agent offer without hiding variable cost or assuming a pilot will support unlimited requests.

Why this matters

Good agent work is useful before it is impressive.

An agency can win projects and still create a loss-making service if it underestimates support, review, changes, model usage, channel charges, or the time required to keep quality high.

Core concepts

The language that keeps the work clear.

Gross marginWhat remains after the direct labor, provider, channel, and support costs required to deliver the service.
CapacityThe amount of client work a team can serve at an agreed quality level.
Usage limitA transparent boundary for variable model, message, voice, or integration consumption.
Service levelA realistic commitment about review, support, or response that the team can meet consistently.
The practical method

How to model margin and capacity

Track delivery effort

Measure the time spent in discovery, build, testing, client communication, review, support, and changes.

Track direct usage

Record the provider and channel cost drivers that vary with volume or complexity.

Set sensible limits

Define what usage, support, response, and change work are included before the client reaches a surprise bill.

Plan capacity by quality

Use review workload and exception volume—not sales desire alone—to decide how many clients a team can responsibly onboard.

Worked example

Worked case: manage margins, capacity, and usage

An agency offers a managed voice intake service. The setup fee covers design and testing; the monthly fee covers monitoring and review; voice minutes and message usage are tracked against an agreed allowance.

After two pilots, the agency learns that clients with complex appointment rules need more exception review. It updates its capacity model and creates a higher-touch tier rather than pretending every client fits the original price.

The team protects quality by making usage and support assumptions visible, reviewing them monthly, and proposing changes before limits are exceeded.

Build it in practice

Complete the working artifact

Setup labor: [hours/cost]. Recurring review/support: [hours/cost]. Variable usage drivers: [list]. Included limits: [list]. Quality capacity signal: [metric]. Change trigger: [condition].
Economics check

Calculate contribution margin before capacity.

For one client and period, start with collected revenue. Subtract model and platform usage, direct delivery labor, direct support, rework, and other variable service costs. Then test how the result changes when volume, exception rate, or provider price moves.

  • Use one definition of billable and productive utilization.
  • Reserve capacity for support, QA, documentation, sales, and improvement.
  • Track client concentration and the margin lost when one custom exception becomes permanent.

Sources used for this check

Practice

Before you move on

  • Model one offer with setup, ongoing labor, usage, and support.
  • Choose one limit that prevents hidden delivery cost.
  • Write the client notice that occurs before an overage or scope change.
  • Direct cost is visible.
  • Usage and support limits are understandable.
  • Capacity is based on quality workload.
  • Pricing can change when evidence changes.

Failure drill

Revenue grows while contribution margin disappears

The case

A client pays a fixed monthly fee. Usage doubles, the model price rises, support takes six extra hours, and senior staff redo failed work. The account still appears healthy because the report subtracts only software subscriptions.

Your call

  1. Which variable costs belong to the account?
  2. How will you value rework and senior review?
  3. Which threshold triggers repricing, redesign, or exit?
Reveal a defensible response

Response: Track revenue less direct model, tool, support, rework, and delivery labor costs. Review usage and exception volume by client, include capacity in the forecast, and agree on thresholds before the margin is gone. Growth that consumes unpriced labor is not scale.

Lesson progress

Finished this lesson?

Save your place on this device so it is easy to pick up where you left off.

Not marked complete yet.