Service Economics
← All Resources

The AI Cost Integration Playbook

For when your team uses AI tools but you have no idea what they're costing per client.

When to Use This Playbook

  • Your team uses AI tools in delivery but costs are buried in "software" or "overhead"
  • You can't answer what AI costs you per engagement
  • AI API bills are growing and you don't know what's driving the growth
  • You want to build AI costs into engagement pricing going forward

What You'll Need

  • A list of AI tools and subscriptions used in delivery
  • API billing data or subscription invoices for the last 3 months
  • Rough understanding of which delivery activities use AI
  • 2–3 hours for the full audit; ongoing tracking takes 15 minutes per week

The Process

Step 1: Audit Current AI Tool Usage

List every AI tool your teams use in client delivery. Include both paid subscriptions (ChatGPT, Claude, Midjourney) and API usage (OpenAI API, Anthropic API, custom models). Don't forget embedded AI in existing tools — Notion AI, GitHub Copilot, Grammarly Business.

Step 2: Map Tools to Delivery Activities

For each tool, identify which delivery activities it supports. Content drafting? Code generation? Data analysis? Quality review? This mapping is essential for attribution — you need to know which engagements benefit from which AI tools.

Step 3: Choose an Allocation Model

Three models, pick the one that fits your current data:

  • Direct Attribution: AI costs are tracked to specific engagements via API keys, project tags, or usage logs. Most accurate but requires tooling.
  • Pro-Rata: Total AI cost divided across engagements by revenue share, headcount, or hours. Simple but imprecise.
  • Absorption: AI costs treated as overhead and factored into fully-loaded rates. Simplest but hides engagement-level economics.

Step 4: Implement Minimum-Viable Tracking

Start simple. A spreadsheet that captures monthly AI costs by tool, mapped to engagement or engagement type via your chosen allocation model. Automate later; understand the numbers first.

Step 5: Build Into Engagement Pricing

Once you understand your AI cost per engagement, factor it into new engagement pricing. For existing engagements, use the data to assess whether current pricing still works — or whether you need the Margin Recovery Playbook.

Common Mistakes

  • Only counting the API bill.The AI Cost Iceberg shows that visible API costs are a fraction of true AI delivery cost. Don't forget prompt engineering time, quality review, and rework.
  • Waiting for perfect data. Rough allocation is infinitely better than no allocation. Start with what you have and refine.
  • Treating AI cost as pure overhead. If AI contributes directly to delivery, it should be attributed to engagements — not hidden in a general overhead bucket.

Related Frameworks

DigitalCore automates this playbook — see how it works →