Model the Practice

You do not need to become a global systems integrator to participate in this shift. But you do need to move beyond license resale and one-time adoption projects. The durable opportunity is a managed AI practice: packaged offers, measurable outcomes, reusable agent IP, recurring AgentCare, and a delivery model where humans and agents work together.

That is why the rest of this playbook does not stop at Copilot adoption. Adoption opens the door. Agent builds expand the account. AgentCare turns the account into a monthly operating relationship.

"How do we need to change so this does not become another project business?"

Move from product-led services to domain-specific AI offerings, reusable agent patterns, and internal accelerators that make every engagement faster and more defensible.

Lead with business outcomes, not feature checklists. Use the show-and-tell motion to open the door, then attach readiness, agent builds, outcome sprints, and managed AgentOps.

Standardize delivery around worklists, telemetry, reusable prompts, evaluation sets, and human-plus-agent delivery roles. The goal is not heroic delivery. The goal is repeatability.

Break the old practice silos. Copilot, security, data, Power Platform, Azure, and business process expertise need to show up as one coordinated motion for the customer.

“That’s interesting — but are you actually using this yourself?”

Personal agents are powerful — and they naturally want to grow. A well-used personal agent always reaches the same ceiling: someone wants it to connect to a live data source, trigger an action in another system, or run automatically. That ceiling is your Copilot Studio conversation. It doesn’t require a pitch — it surfaces naturally out of the agent the customer already loves.

This is also where personal agents become company-wide agents. The meeting prep agent your sales rep built? With Copilot Studio, it can connect to your CRM, pull live opportunity data, and trigger next-step reminders automatically. Same use case — enterprise scale, enterprise power.

The partner who builds a personal agent and disappears loses the Studio opportunity. Staying in the conversation means actively tracking three things:

This is the Agent Maturity Model playing out in real time. Personal agents are Level 1. When a customer hits the ceiling, they’re ready for Level 2 and 3. The Agent Maturity Model gives you the full framework. The Show & Tell motion you just ran is where every maturity journey begins. Your monitoring of their agents is what tells you when they’re ready to climb.

Security work isn’t a prerequisite burden — it’s what makes the AI investment actually pay off. The pivot is not a separate conversation. It is the natural conclusion of the demo you just ran.

This is how you sell the $15,000 data governance and security project. You aren’t selling “compliance” — you are selling the safe deployment of the AI they just fell in love with. And the partner who runs this three-step motion consistently is building a pipeline that funds itself: the security project funds the agent builds, the agent builds fund the managed retainer.

Contrast with SOW fatigue — every AI vendor promises, none measure. Outcome-based turns a cost line into an investment the CFO can actually defend.

The partner’s fee is tied to the champion’s goal too. That alignment is what kills the usual scope-creep fight and makes the champion the partner’s loudest advocate internally.

The partner brought the baseline and the measurement plan. IT gets to focus on tenant hygiene, security, and data access — not on building a business case they were never staffed to build.

• Customer can name one number that matters and is measurable today.
• Baseline data is accessible before you quote.
• Executive sponsor cares about the KPI personally and can clear friction.
• Pre-agreed MRR conversion path the customer will sign pre-sprint.

• No baseline and no time to establish one.
• KPI depends on variables you don’t control (seasonality, another vendor).
• Sponsor is a director or below — no authority to protect the KPI.
• It’s ongoing managed services — those are retainer, not outcome-priced.

AgentOps is the ongoing managed service practice of keeping your customers’ AI agents accurate, secure, and generating measurable value — month after month, without a defined end date.

Traditional MSP revenue is tied to seats and devices — static infrastructure that doesn’t grow on its own. AgentOps changes that equation. Every Copilot Studio agent your team builds and deploys creates a new, permanent managed service obligation: the agent must be tuned as usage patterns evolve, its grounding data audited as SharePoint sources change, its token consumption monitored for cost efficiency, and its security posture reviewed as Microsoft ships model and connector updates. That work never ends because the agent never stops running. That dependency is your retainer.

The AgentOps model structures this work into three monetizable components — a per-user governance base, a per-agent management fee (AgentCare), and an Azure consumption markup (Credit Wrap) — that stack together into a recurring revenue engine. A partner who has built and handed over six agents across three customers is already running an AgentOps practice. The question is whether that work is named, scoped, and priced — or whether it’s being given away as part of a flat managed services agreement that was written before agents existed.

The Stacked Default model invoices all four components as separate line items. The Flat-Tier Alternative bundles them into a single per-user rate — same economics, simpler invoice, lower visible margin.

Four dials. Every tier is a different mix of these four lines — and nothing else. Read them before the model cards below and the tier math on the next page lands in a single pass.

What it is: The full three-component model — Per-User Base + AgentCare + Credit Wrap — invoiced as separate line items. This is the recommended starting point for any partner building a durable AgentOps practice. It maximizes margin, creates explicit value visibility for the customer, and scales naturally as agent count grows: every new agent deployment adds revenue automatically without a contract renegotiation.

Use this when: your customer understands the value of each service layer, you have at least one deployed agent under management, and your team can track per-agent time. The transparent invoice is also a QBR asset — it makes value concrete and visible every month.

What it is: A single all-inclusive per-user rate that bundles the Per-User Base, AgentCare, and Credit Wrap into one invoice line. The customer sees one number; the complexity is absorbed internally. This is how most traditional MSPs already price — it removes friction from the sales conversation and makes renewals frictionless.

Use this when: you're earlier in your practice build and want to keep billing simple, when your customer base is cost-sensitive and detailed line items create friction, or when you're managing fewer than 3 agents per customer and the math still works at the bundled rate. The tradeoff is margin compression as agent count grows — at 6+ agents per customer, the Stacked model outperforms meaningfully.

To model flat-tier pricing: set agents/customer = 0 in the CPB Workbook Tab 2. The four-stage structure is identical.

What it is: AgentCare priced as a standalone offering — no per-user base, no Credit Wrap — just a monthly managed service fee per deployed agent. This is not the same as the AgentCare component within the Stacked model. As a standalone, it's designed for a specific scenario: a customer whose tenant is already under MSP management by someone else (or by their own IT team), but who has deployed Copilot Studio agents that need professional oversight.

Use this when: you inherit an agent someone else built, when a customer wants to separate AI services from their existing MSP contract, or when you're entering a new account through an agent build and want a recurring revenue bridge while you work toward full managed services. It's also the right model for partners who are pure Copilot Studio builders — strong on build, not yet running full managed services.

Critical rule: Agent builds are always scoped and billed separately as project engagements ($3,500–$8,000+ per agent). The AgentCare retainer covers ongoing management of deployed agents — never new construction. This boundary must be explicit in your SOW and renewal language.

What it is: A standalone or add-on offering where you purchase pooled Azure compute (Copilot Studio message credits, Azure OpenAI tokens, Power Automate flow runs) at CSP cost and bill the customer at a 25–35% markup. You're not just reselling credits — you're managing the consumption pool, optimizing credit allocation across agents, monitoring for unexpected spikes, and ensuring no agent runs out of budget mid-month. That management overhead justifies the markup.

Use this when: any customer is running Copilot Studio agents at meaningful volume. Credit Wrap is always additive — it stacks on top of either the Stacked Default or Flat-Tier model. The only question is whether you surface it as a visible line item (Stacked) or absorb it into your flat rate (Flat-Tier). At Tier 2 the dollar amounts are modest; at Tier 3 with 6+ agents on Azure AI Foundry, it becomes a meaningful margin layer.

GM: ~85% — essentially pure margin once you've set up the pooling mechanism. Small on its own — but across a 50-customer base at Tier 3, Credit Wrap adds $75K–$90K annually with no incremental labor. Never leave it on the table.

You’re likely ready to claim if you meet all four of these:

A working Excel model. Enter your actual customer counts, seats, rates, headcount, and overhead, and a live P&L builds as you type — scenario-planned against the tiers and monetization models covered in the Revenue Runway above.

During practice planning, before you quote your first Copilot engagement. Revisit Tabs 1–5 whenever you consider adding a new tier, adjusting your monetization model, or scoping a new outcome-based sprint against a fresh KPI.