Copilot & Foundry: The Microsoft AI Stack for Small Business · Lesson 6 of 10
Pricing, Licensing, and TCO
Time: ~25 min reading + ~15 min worksheet Goal: Build honest 3-year TCO models for Microsoft-stack AI engagements. Know the licensing traps. Price your own work correctly.
---
Most consultants can't cleanly explain what a Copilot or Foundry engagement will cost the client over 3 years. They quote build fees, ignore consumption costs, gloss over licensing prerequisites, and get ambushed by Azure spend surprises in month 3.
Doing this module well is how you stop losing deals to more polished competitors — and how you avoid selling unhappy engagements that miss their stated ROI because the math was sloppy.
All dollar figures here are 2026 directional. Always verify current pricing with Microsoft before quoting a client.
Here's what the client needs *before* you can deploy each tier.
To use M365 Copilot
- A qualifying M365 license per user (Business Standard, Business Premium, E3, or E5) — ~$12–$57/user/month
- M365 Copilot add-on license per user — ~$30/user/month, annual commitment
- Total per user: ~$42–$87/month, depending on base license tier
To use Copilot Chat (free tier)
- Any qualifying M365 license — that's it
- No add-on required
To use Copilot Studio
Several paths, depending on usage:
- Bundled capacity — M365 Copilot licenses include a message allowance for Copilot Studio
- Copilot Studio standalone license — ~$200/month per tenant, includes a monthly message allowance
- Copilot Studio Pay-As-You-Go — consumption-based, billed through Azure subscription
- Power Platform premium connectors — extra cost per user or per app for premium connectors (SAP, Salesforce, etc.) — ~$15/user/month for premium connector access
To use Foundry
- Azure subscription with appropriate billing account
- No per-seat license — everything is consumption
- Azure AI Search (separate service, tiered pricing)
- Azure OpenAI access (may require application-to-use for certain models in some regions)
- Azure networking (VNet, private endpoints — add real cost at enterprise scale)
Governance / Purview
- Microsoft Purview requires its own licensing (often bundled in E5, sold separately otherwise)
- DSPM for AI, sensitivity labels, DLP — various components, various licensing
- Budget $3–$12/user/month for Purview functionality relevant to AI governance, depending on what's included in base licenses
For any real engagement, there are seven places money flows. Budget for all seven or you'll miss something.
1. User licenses
M365 base, M365 Copilot add-on, premium connectors.
2. Platform capacity
Copilot Studio capacity, standalone license, or PAYG.
3. Azure consumption
If Foundry is involved: model tokens, Azure AI Search, Foundry Agent Service, compute, storage, networking.
4. Governance / security licensing
Purview functionality at the right tier.
5. One-time build cost (your consulting fee)
Discovery, design, build, test, launch.
6. Training and change management
Rolling out AI requires training. Budget real dollars.
7. Ongoing tuning and managed service
10–25% of build cost annually is typical. Can be sold as a managed service.
Let's walk through three realistic scenarios. Use these as templates.
Example A: Small SMB, M365 Copilot + one Copilot Studio agent
Client: 50-person accounting firm on M365 Business Standard
Year 1: | Item | Cost | |---|---| | M365 base (already paying) | — | | M365 Copilot add-on, 50 users × $30 × 12 | $18,000 | | Copilot Studio capacity (covered by Copilot licenses) | $0 | | Azure consumption (minimal, mostly nothing) | $500 | | Your consulting build | $30,000 | | Training budget | $3,000 | | Ongoing tuning (half-year) | $2,000 | | Year 1 total | $53,500 |
Year 2+: | Item | Cost | |---|---| | M365 Copilot licenses | $18,000 | | Azure consumption | $500 | | Ongoing tuning | $5,000 | | Year 2 total | $23,500 |
3-year TCO: ~$101,000. Per-user cost of AI: ~$670/year. Totally reasonable for a firm averaging $100K+/person in billings.
Example B: Mid-market SMB, Pattern 2 (Copilot Studio + Foundry)
Client: 150-person manufacturer on M365 E3 with Purview bundled
Year 1: | Item | Cost | |---|---| | M365 Copilot add-on, 150 × $30 × 12 | $54,000 | | Copilot Studio standalone license ($200/mo) | $2,400 | | Premium connectors, 30 users × $15 × 12 | $5,400 | | Azure consumption: Foundry + Azure AI Search + storage + networking | $24,000 | | Your consulting build (Pattern 2 engagement, 5 months) | $95,000 | | Training | $8,000 | | Ongoing tuning (half-year) | $9,000 | | Year 1 total | $197,800 |
Year 2+: | Item | Cost | |---|---| | M365 Copilot | $54,000 | | Copilot Studio | $2,400 | | Premium connectors | $5,400 | | Azure consumption | $30,000 | | Ongoing tuning / managed service | $22,000 | | Year 2 total | $113,800 |
3-year TCO: ~$425,000. Per-user cost: ~$945/year. Real money, real commitment.
Example C: SMB with a customer-facing Foundry product
Client: 25-person logistics startup. Custom customer-facing chatbot. 50K conversations/month expected.
Year 1: | Item | Cost | |---|---| | M365 base (maybe a few Copilot licenses for internal) | $9,000 | | Azure consumption at 50K conversations/month | $45,000 | | Azure AI Search tier | $6,000 | | Foundry Agent Service, storage, networking | $8,000 | | Your consulting build (Pattern 3, 7 months) | $150,000 | | Design / UX | $15,000 | | Training (minimal — internal users few) | $2,000 | | Ongoing engineering (partial year) | $12,000 | | Year 1 total | $247,000 |
Year 2+: | Item | Cost | |---|---| | M365 | $9,000 | | Azure total | $70,000 (higher if conversation volume grows) | | Ongoing engineering / managed service | $36,000 | | Year 2 total | $115,000 |
3-year TCO: ~$477,000. This is a real investment. Only justifiable if it's a core part of the customer experience driving retention or revenue.
As of 2026, Microsoft introduced the Agent Pre-Purchase Plan — a unified SKU that lets organizations buy capacity across Copilot Studio and Foundry agents without choosing up front. Useful for:
- Large rollouts with mixed architecture
- Organizations where you don't know year-2 mix yet
- Procurement that prefers one line item
For SMB engagements, it's rarely the right fit — smaller customers are better off with specific licenses until they know their pattern. But know it exists; larger clients will ask.
For the build portion, price based on value, not hours. Ranges from Modules 2–4:
- M365 Copilot rollout (50-person company): $15K–$40K
- Simple Copilot Studio agent: $5K–$10K
- Departmental Copilot Studio agent with integrations: $10K–$30K
- Multi-agent Copilot Studio suite: $30K–$80K
- Pattern 2 engagement (Copilot Studio + Foundry): $40K–$150K
- Foundry product (Pattern 3): $100K–$500K+
Plus ongoing managed service. A good default:
- Retainer for tuning and support: 15–25% of build cost annually
- Or month-to-month: $2K–$15K/month depending on complexity
This is where real consulting income lives. Build fees are a one-time event; managed service is recurring.
Six places clients get blindsided. Name them up front.
1. Azure egress fees
Pulling data out of Azure costs money. Easy to forget. Can run $500–$5K/month at scale.
2. Premium connectors
If your Copilot Studio agent uses a Salesforce, SAP, or other premium connector, every user who interacts with it may need a premium connector license. Add it up.
3. Dev environment costs
You'll want dev, test, and prod environments. Each has its own Azure spend.
4. Copilot Studio message overages
Hit your capacity, and you're in PAYG rates — which are more expensive per message. Monitor carefully.
5. Purview at the right tier
If the client is on E3, they may not have DSPM for AI, which costs extra. Surprise line item.
6. Third-party integrations
The LOB system that's "just an API" away often needs middleware, custom connector builds, or licensing from its vendor.
When presenting TCO to an owner, follow this structure:
- Here's what you're paying Microsoft (licensing + Azure).
- Here's what you're paying me (one-time build + ongoing).
- Here's the value I'm projecting (time savings, revenue impact, cost avoidance — from Foundations Module 9).
- Here are the assumptions (adoption, ramp, hourly rates).
- Here are the risks (pricing drift, scaling surprises, change management).
- Payback period and 3-year cumulative net value.
Do this on one page. If you can't do it on one page, you haven't done the math cleanly enough.
---
- Seven cost categories: user licenses, platform capacity, Azure consumption, governance, build, training, ongoing.
- M365 Copilot: ~$30/user/month add-on, on top of qualifying M365 base license.
- Copilot Studio: message-based, bundled / standalone / PAYG options.
- Foundry: pure consumption, across multiple Azure services.
- Three example TCOs: $100K, $425K, $477K over 3 years for increasingly ambitious engagements.
- Hidden costs: egress, premium connectors, dev environments, message overages, Purview tier, third-party licensing.
- Your consulting pricing: value-based, with a 15–25% managed-service retainer for the real income.
---
Next up: Module 7 — Meet Northside HVAC. The worked-example project that takes us through Modules 7–10.
---
Time: ~15 min Goal: Practice building a clean 3-year TCO so you never get ambushed on a quote.
---
Client: 80-person regional law firm. Already on M365 Business Premium. Scope:
- M365 Copilot rolled out to all 80 users
- Two Copilot Studio agents: one for HR (internal) and one for client intake (low volume, Teams-embedded)
- No Foundry (yet)
- Your build for the Copilot rollout + both agents: $65K one-time
- Ongoing managed service: 18% of build/year
Build a 3-year TCO. Fill in:
Annual recurring (Year 1+): | Line | Annual | Notes | |---|---|---| | M365 Copilot licenses (80 × $30 × 12) | $______ | | | Copilot Studio capacity | $______ | (assume covered by Copilot bundle) | | Azure consumption | $______ | (minimal for Pattern 1) | | Premium connectors (est. 20 users × $15 × 12) | $______ | (for intake agent connector) | | Purview at appropriate tier | $______ | (80 users × ~$6 × 12) | | Annual recurring subtotal | $______ | |
One-time Year 1: | Line | Amount | |---|---| | Your consulting build | $______ | | Training | $______ (estimate ~$4K for 80 users) | | One-time subtotal | $______ |
Ongoing managed service:
- Year 1 partial (6 months): $______
- Year 2 full: $______
- Year 3 full: $______
3-year cumulative TCO: $______
In the scenario above, which hidden costs should you flag to the owner before quoting? List at least three.
- ______________________________________________________
- ______________________________________________________
- ______________________________________________________
The owner sees the 3-year TCO and flinches at the number. Write a 60-second response that contextualizes it against business value. (Hint: use the Foundations Module 9 sizing framework.)
Your response: ```
```
---
Exercise 1 — sample TCO:
Annual recurring:
- M365 Copilot licenses: $28,800 (80 × $30 × 12)
- Copilot Studio capacity: $0 (covered by bundled capacity from Copilot licenses, likely fine at this scale)
- Azure consumption: $600 (truly minimal for Pattern 1)
- Premium connectors: $3,600 (20 users × $15 × 12)
- Purview AI governance: $5,760 (80 × $6 × 12 estimated)
- Annual recurring subtotal: ~$38,760
One-time Year 1:
- Consulting build: $65,000
- Training: $4,000
- One-time subtotal: $69,000
Ongoing managed service (18% of $65K = $11,700/year):
- Year 1 partial (6 months): $5,850
- Year 2 full: $11,700
- Year 3 full: $11,700
3-year cumulative TCO:
- Year 1: $38,760 + $69,000 + $5,850 = $113,610
- Year 2: $38,760 + $11,700 = $50,460
- Year 3: $38,760 + $11,700 = $50,460
- 3-year TCO: ~$214,530
Per-user cost: ~$894/year. Reasonable for a firm with knowledge workers billing $150–$400/hour.
Exercise 2 — hidden costs to flag:
- Premium connector licensing — if the intake agent uses a premium connector (Salesforce, HubSpot, DocuSign, etc.), that adds meaningful per-user cost that scales with user count. Easy to miss in initial quote.
- Purview tier — law firm is on Business Premium which includes some Purview capabilities but may need additional licensing for advanced DSPM for AI or Insider Risk Management. Worth verifying.
- Dev/test environments — you'll want separate environments for development, testing, and production. Each has its own Copilot Studio capacity and Azure spend. Bake this into the build cost or surface as a separate line.
- Microsoft licensing adjustments — M365 Copilot pricing and terms change; annual commitment means 12-month lock-in. If Microsoft changes pricing mid-year, you need to renew at new rates.
- Capacity overages — if either agent gets unexpectedly popular, message consumption could exceed the bundled Copilot Studio capacity and start drawing on PAYG rates. Monitoring required.
- Client-specific integration work — if the intake agent needs to talk to an on-prem legal practice management system, that may require data gateway setup, custom connector build, or licensing from that vendor.
Exercise 3 — sample response:
> "Fair reaction — let me contextualize it. Over three years you're investing about $215K. Here's what I'm projecting on the value side: M365 Copilot alone typically delivers 10–15% time savings on email, document, and meeting work. For 80 knowledge workers at your average billing rate, even 10% savings is well over $1M/year in recoverable capacity. The two agents layer on top of that — the HR agent alone should save your HR manager 6–8 hours a week of policy questions. I'm not selling you a product; I'm selling you the ability to redeploy several hundred hours a year of senior staff time into billable work. On those assumptions, payback is under 6 months and year-2 cumulative net value is well into the low six figures. The big-looking TCO number is the investment. The return is the capacity you free up. Want to walk through the math line by line?"
Why this works: validates the reaction ("fair"), reframes from cost to investment, sizes return with Foundations-course rigor, ends with an invitation to scrutinize. Owners trust transparency.
---
Check yourself: Can you build a 3-year TCO from scratch in 20 minutes, flag the hidden costs, and frame the ROI without stumbling? That's the last check before the worked example starts in Module 7.
Previous: How They Fit Together: Architecture Patterns · Next: Project Kickoff: Northside HVAC Internal Helpdesk