MSP AI Services Pricing: What to Charge in 2026
Your client is going to ask for AI like it is a menu item.
"Can you help us with Copilot?" "Can you automate our support inbox?" "Can you build us an AI policy?" "Can you make sales faster?"
Those sound like technology questions. They are pricing traps.
MSP AI services pricing in 2026 needs to start with scope, not excitement. AI work can include license resale, data readiness, Copilot rollout, workflow automation, employee training, policy work, governance reviews, and ongoing support. Those are different services. They should not be shoved into one vague line item called "AI."
If you price AI like a license pass-through, you will undercharge. If you price it like open-ended consulting, clients will stall. The middle path is simple: package the work you control, put a margin target under it, and say no to outcomes you cannot own.
Quick answer: how should MSPs price AI services?
MSPs should price AI services by separating license costs from service work. For most SMB clients, start with a $2,500-$6,000 readiness or Copilot rollout project, a $750-$2,500 monthly governance retainer, or a $10-$30 per user monthly add-on only when the deliverables are clearly defined.
Those numbers are starting points, not market averages. The formula matters more than the range:
AI service price = delivery cost + tool cost + risk buffer + target margin
The AI part does not suspend normal MSP math. It makes the math more important.
What counts as an AI service for an MSP?
Do not sell "AI services" as one thing. That is how scope gets weird.
In an MSP context, AI services usually fall into six buckets:
- AI readiness assessment: Inventory approved and unapproved AI usage, data exposure, identity controls, policy gaps, and business use cases.
- Microsoft 365 Copilot rollout: Licensing guidance, data access review, pilot group setup, admin controls, training, and adoption support.
- AI governance package: Acceptable use policy, data handling rules, risk register, review cadence, and evidence collection.
- AI-assisted managed services: Using AI to improve ticket triage, QBR prep, documentation summaries, remediation planning, or client reporting.
- Workflow automation projects: Scoped automations for intake, reporting, summarization, internal knowledge, or client-specific processes.
- Ongoing AI support: Quarterly reviews, policy updates, user training refreshes, usage reporting, and new-tool evaluation.
Each bucket has different risk. Copilot rollout is partly technical. Governance is advisory. Automation projects are scope-heavy. Ongoing support can become an unlimited help desk for every weird prompt a client writes.
That is why the first pricing rule is boring and correct: name the deliverable.
Why the demand is real this time
The channel has heard plenty of fake AI urgency. This one is more practical.
AvePoint and Omdia surveyed 333 MSP executives across North America, Europe, and Asia-Pacific in April 2026. The headline number is useful: 51% of MSPs said data governance and compliance challenges are the main obstacle to client AI adoption. Not budget. Not generic skills. Governance and compliance.
The same research found 94% of MSPs are committed to automation around AI data readiness and compliance services, while only 43% report high maturity delivering AI-ready data environments. That gap is the business opportunity.
CRN's coverage of the study points to the same shift: MSPs are moving from one-time assessments toward recurring services that manage risk, enforce policies, and monitor data environments. Channel Dive reported a similar movement from AI-assisted ticketing into governance services and managed AI contracts.
Microsoft is pointing customers in the same direction. Its Secure and Governed Data Foundation for Microsoft 365 Copilot is built around three pillars: remediate oversharing, set up guardrails, and meet regulations. That is not a license conversation. That is paid readiness work.
NIST's AI Risk Management Framework Playbook gives the broader structure: Govern, Map, Measure, and Manage. You do not need to turn every SMB into a policy lab. But you do need a repeatable way to identify risk, price the remediation, and document what the client accepted.
Package 1: Copilot readiness and rollout
This is the easiest AI service to sell badly.
A client asks for Copilot. The MSP quotes licenses. Everyone feels productive for ten minutes. Then someone realizes Copilot can surface files users already had permission to access, including files they probably should not have had permission to access.
Microsoft's own guidance says Copilot value depends on a secure and governed data foundation. It specifically calls out oversharing, guardrails, and regulatory obligations. That is your service scope.
A sane Copilot rollout package includes:
- Current Microsoft 365 license and eligibility review
- SharePoint, OneDrive, Teams, and Exchange access risk review
- Pilot group selection and rollout plan
- Data handling and acceptable use policy
- Admin control review
- User training session
- 30-day adoption and issue review
Do not make license margin the business model. Microsoft pricing changes, distributor economics vary, and Copilot Chat may already be included for eligible Microsoft Entra account users with eligible Microsoft 365 subscriptions. Agent usage can also introduce metered cost. The margin is in readiness, rollout, training, policy, and recurring governance.
Example pricing for a 50-seat client
| Line item | Delivery cost example | Price to quote | Margin logic |
|---|---|---|---|
| Copilot readiness review | 8 hours at $95 burdened cost = $760 | $1,500-$2,500 | Small diagnostic with clear findings |
| Copilot pilot rollout | 16 hours at $95 plus $150 admin cost = $1,670 | $3,500-$5,000 | Includes policy, pilot, training, review |
| Ongoing Copilot admin and adoption | 3 hours/month at $95 = $285 | $750-$1,250/month | Covers review, changes, questions, reporting |
Use your own loaded cost. If your senior consultant costs $125/hour internally, change the math. The point is not to copy these numbers. The point is to stop treating Copilot as a resale SKU with a little free labor attached.
Package 2: AI-assisted managed services
This is where MSPs are most likely to overpromise.
If you use AI internally to summarize tickets, draft QBR notes, clean documentation, or speed up analysis, that does not automatically mean the client should pay an "AI fee." Clients do not care that your tool is shiny. They care whether the service is better, faster, clearer, or cheaper.
Price AI-assisted managed services only when the client gets a defined deliverable.
Good examples:
- Monthly risk summary built from tickets, alerts, and roadmap items
- Faster QBR prep with clearer budget decisions
- AI-assisted documentation cleanup project
- Ticket trend analysis with recommended fixes
- User training on safe AI workflows
- Executive summary of open IT risks and next actions
Bad examples:
- "AI-enhanced support" with no service difference
- "Productivity improvement" with no measurement plan
- Unlimited prompt help inside the managed services agreement
- Client-specific agent experiments with no owner, budget, or acceptance criteria
The pricing rule
If AI reduces your internal delivery cost, that savings belongs to your margin unless you choose to trade some of it for a better client promise.
Example: your team uses AI to cut monthly QBR prep from 6 hours to 3 hours. Your burdened cost is $85/hour. You saved $255/month in delivery cost.
Do not immediately discount the client by $255. You still own review, judgment, cleanup, relationship management, and the risk of bad output. A better move is to package the improved deliverable:
- Add a $500/month business review and roadmap reporting add-on for small clients.
- Add $10-$20/user/month for a premium managed services tier if the tier includes named deliverables.
- Keep base pricing flat and let AI improve gross margin when the client experience is unchanged.
The uncomfortable truth: AI is often a margin tool before it is a revenue line. That is fine. Margin is not a consolation prize.
Package 3: AI governance assessment and retainer
This is the strongest standalone AI service for many MSPs.
Governance sounds boring until a client realizes employees are pasting customer data into random tools, Copilot is blocked by messy permissions, and nobody has written down what AI is allowed to do.
The AvePoint/Omdia 51% number matters because it names the blocker. Clients do not need another AI pep talk. They need someone to turn messy questions into decisions:
- Which AI tools are approved?
- What data can employees put into them?
- Who reviews AI-generated client work?
- What happens when a tool exposes sensitive data?
- How often will permissions and usage be reviewed?
- Which risks are accepted by the client, not silently absorbed by the MSP?
A governance assessment should produce a practical output, not a 40-page fog machine.
Deliverables:
- AI tool inventory, including shadow AI findings where available
- Data exposure and access control review
- Acceptable use policy draft or update
- Risk register with owner, severity, and next action
- Roadmap of remediation projects
- Executive summary the client can approve
Example pricing
| Package | Typical scope | Delivery cost example | Price to quote |
|---|---|---|---|
| AI readiness assessment | 8-12 hours, one client environment, basic policy review | $900-$1,400 | $2,500-$4,500 |
| AI governance buildout | 18-30 hours, policy, risk register, roadmap, executive review | $1,900-$3,500 | $6,000-$10,000 |
| Quarterly governance retainer | Review meeting, policy changes, risk register update, new tool review | $350-$800/month | $750-$2,500/month |
| Compliance tier add-on | AI controls added to existing compliance service | Varies by users and review cadence | $15-$25/user/month |
CRN's coverage of the AvePoint/Omdia research makes the right point: the market is moving from one-time assessments to recurring services that manage risk, enforce policies, and monitor data environments. That is the packaging shift.
The audit finds the work. The retainer keeps the work alive.
The scoping mistake that kills AI margins
The fatal phrase is "we can help with AI."
That sentence has no boundary. It can mean Copilot licenses, a chatbot, policy work, workflow automation, prompt training, data cleanup, vendor selection, security review, or a monthly executive meeting. If it is not scoped, the client will define it later. They will not define it in your favor.
Use a four-part scope line:
- Tool scope: Which AI tools are included?
- Data scope: Which data sources, departments, tenants, or workflows are included?
- Decision scope: What decisions can the MSP recommend, and what must the client approve?
- Support scope: What happens after rollout, what is billable, and what is excluded?
Put exclusions in plain language.
Examples:
- Custom AI agents are not included unless quoted as a separate project.
- Productivity gains are not guaranteed.
- Client policy approval is required before rollout.
- Remediation work found during assessment is quoted separately.
- The MSP is not responsible for employee misuse after approved policy and training unless ongoing monitoring is contracted.
This is not legal advice. It is pricing hygiene. Your agreement still needs real contract language.
What to say when a client asks, "Can you do AI for us?"
Do not answer with a tool.
Say this:
Yes, but we do not start with the license. We start with what you want AI to do, what data it can see, what risk you are willing to accept, and who owns the decisions. We run a short readiness assessment first. Then we quote rollout, governance, and ongoing support separately so nobody is guessing.
That answer does three things.
First, it reframes AI as a business scope, not a product SKU.
Second, it creates a paid first step.
Third, it protects the MSP from inheriting every AI-adjacent problem after the client signs.
A good first paid step is simple:
- 60-minute stakeholder call
- M365 and data access review
- Current AI tool inventory
- 3-5 priority use cases
- Risk and readiness score
- Recommended package: rollout, governance, automation project, or no-go
Price that at $1,500-$3,000 for a small client and $3,000-$6,000 for a larger or regulated client. Credit part of it toward the implementation if you want. Do not make it free by default.
Where Scopable fits
AI service pricing breaks when discovery, risk, roadmap, budget, and quote scope live in different places.
Scopable helps MSPs turn assessment findings, client risks, roadmap items, budget assumptions, and approvals into priced work the client can understand. That matters because AI work is full of assumptions. If the client does not approve the assumptions, the MSP usually pays for the ambiguity.
This is the same reason assessment-first quoting matters for normal MSP projects. If the scope is vague, the margin gets donated.
For related thinking, read Scopable's guides to Microsoft Copilot for MSPs, MSP compliance pricing, scoping MSP projects, pricing, quoting, and margin protection, and AI quoting versus manual quoting.
A simple AI services menu for MSPs
Here is the menu I would start with.
| Offer | Buyer problem | Price range | Notes |
|---|---|---|---|
| AI readiness assessment | "Can we use AI safely?" | $2,500-$6,000 | Paid diagnostic, no free consulting spiral |
| Copilot rollout package | "Can you help us deploy Copilot?" | $3,500-$10,000 | License cost separate, governance included |
| AI governance retainer | "Who keeps this under control?" | $750-$2,500/month | Quarterly review, policy updates, risk register |
| AI-assisted reporting add-on | "Can you make reviews more useful?" | $500-$2,000/month | Must include clear deliverables |
| Workflow automation project | "Can AI speed up this process?" | $5,000-$25,000/project | Needs tight acceptance criteria |
| Compliance tier add-on | "Does this affect our compliance work?" | $15-$25/user/month | Best when layered onto an existing compliance package |
If you only sell one thing first, sell the readiness assessment. It is concrete, useful, and leads naturally to rollout, governance, and remediation work.
What not to sell
Do not sell "AI overhaul."
Do not sell guaranteed productivity gains unless the client agrees to a baseline, measurement plan, and owner.
Do not sell custom agents as a casual add-on to managed services.
Do not include unlimited AI support in a flat monthly agreement.
Do not let a license quote imply data readiness, policy, training, and governance are included.
Most MSPs do not get burned because AI is too complicated. They get burned because they let the word AI make them forget how scoping works.
Final rule: price the work you control
AI services are not mystical. They are MSP services with new inputs and more advisory risk.
You control the assessment. You control the rollout plan. You control the policy draft. You control the training. You control the review cadence. You control the quote assumptions.
You do not control whether the client's employees use AI well. You do not control whether Microsoft changes pricing. You do not control whether a department ignores policy. You do not control whether a custom automation changes a broken business process.
So price what you control. Exclude what you do not. Put the assumptions in the quote.
That is how MSPs should sell AI services in 2026: not as hype, not as a license skim, and not as free strategy hidden inside account management.
As paid, scoped work.
If you want to turn AI readiness into scoped work with budget and approvals attached, join Scopable early access.
Sources
- AvePoint and Omdia research announcement
- CRN coverage of AvePoint and Omdia MSP AI research
- Channel Dive on MSPs moving from AI-assisted ticketing to governance services
- Microsoft 365 Copilot plans and pricing
- Microsoft Secure and Governed Data Foundation for Microsoft 365 Copilot
- NIST AI RMF Playbook
