AI Advisory: PE IT Services & MSP Portcos

Why AI moves margin in IT services.

IT services and managed service providers (MSPs) are professional services businesses that sell ongoing technology operations to other companies, usually under a fixed monthly fee plus project work. The economics turn on three numbers: bench utilization, gross margin per ticket, and renewal protection. AI is finally cheap enough and reliable enough to move all three. The category has become the most actively acquired services vertical in private equity, with 466 MSP transactions closing in 2025 according to Omdia, and AI-first delivery is the thesis on almost every new platform.

Walk into a service desk at any MSP between 9am and 11am on a Monday. The Tier 1 queue is full of password resets, MFA enrollment, printer mappings, and Microsoft 365 license toggles. Each ticket eats six to fifteen minutes of a technician's time. The technician costs $55K to $75K loaded. The customer pays a flat monthly fee, so every minute spent on a $5 ticket comes straight out of gross margin. ConnectWise's January 2026 acquisition of zofiQ was an admission of where the floor is: agentic AI can resolve a meaningful chunk of these tickets without a human in the loop. Vendor decks claim 70 to 80 percent. Real-world numbers from Atera, ConnectWise Sidekick, and the in-house Copilot Studio builds land at 15 to 35 percent on routine categories after a six to ten week shadow-mode pilot. That's still a $400K to $1.1M annual saving on a 40-technician service desk.

Project staffing is the second leak. The standard pattern is a spreadsheet maintained by a delivery manager who's also billing 60 percent of her time. The bench shows up at 65 percent utilization when the industry benchmark is 72 to 78. Nobody can explain the gap because the data sits across the PSA, the CRM, the HR system, and a Google Sheet. A staffing copilot that reads from all four and flags conflicts before they happen recovers five to ten utilization points. On a $30M services revenue base, that's $1.5M to $3M of pure margin that wasn't there before.

Then there's the renewal. Half the MSP contracts out there have scope ambiguity baked in: "endpoint management" without a defined endpoint count, "advisory hours" without a cap, "security tooling" without a named SKU list. The portco's account managers don't catch the drift because they're chasing new logos. By the time the renewal lands, the customer's team has grown 22 percent and the MSP is delivering 22 percent more work for the same fee. A renewal-prep agent that walks the contract against actual ticket volume and surfaces the deltas eight weeks before signature recovers five to ten percentage points of renewal margin. That's the cleanest AI win in the category and nobody's pricing it as such yet.

Five AI use cases moving margin right now.

Pulled from current retainer engagements with MSP and IT services portcos in the $25M to $300M revenue band. Vendor names appear where the category has converged on a credible build-on-top option. All five are in production at multiple PE-backed platforms as of Q2 2026, including Lyra Technology Group, Coretelligent, and several Evergreen Services Group subsidiaries that have shipped AI service-desk pilots.

Tier 1 service desk agent for routine ticket categories.

Password resets, MFA enrollment, license toggles, printer mappings, and known software-install requests make up 35 to 50 percent of inbound volume at most MSPs. Each takes a technician six to fifteen minutes. An AI agent with safe write access to Active Directory, Entra ID, and the documentation library can resolve a real fraction of these autonomously, with a human approval step on anything touching production identity.

ConnectWise Sidekick (post-zofiQ acquisition) is the obvious build-on for ConnectWise shops. Atera's Robin handles the same patterns inside an Atera-native PSA + RMM. HaloPSA shops typically build on Microsoft Copilot Studio plus Power Automate plus a custom connector. Real deflection lands at 15 to 35 percent on these categories, not the 70 to 80 percent vendor decks claim.

Sized ROI $400K to $1.1M per year on a 40-technician service desk

Implementation 6 to 10 weeks shadow mode, then graduated write access over a further 8 weeks.

Project staffing optimizer with skills + bench awareness.

Delivery managers staff projects from memory and a spreadsheet that's two weeks stale. The result is over-allocation on the senior bench (the people who said yes first) and under-allocation on the mid-tier (the ones who don't volunteer). Bench utilization sits five to ten points below industry benchmark and nobody can fully explain why.

A staffing copilot that reads the PSA project pipeline, the HR skills matrix, current allocation, and pending PTO recommends staffing assignments with a "why this person" rationale. The delivery manager keeps the decision. The agent does the cognitive work of holding 80 people and 30 projects in working memory at once.

Sized ROI 5 to 10 points of utilization recovery, worth $1.5M to $3M on a $30M services base

Implementation 8 to 12 weeks, mostly spent cleaning the skills matrix and PSA data.

Renewal-prep agent that walks contracts against actual delivery.

Most MSP contracts have scope ambiguity that the customer's growth quietly turns into uncompensated work. Endpoint counts grow. User counts grow. The MSP keeps delivering against the original fee. By renewal time, the gap is 10 to 25 percent and the account manager has no clean way to surface it.

A renewal-prep agent ingests the signed contract, ticket volume by category for the trailing 12 months, endpoint and user growth from the RMM, and produces a one-page brief: here's the scope, here's what you've actually delivered, here's the gap, here's the proposed renewal pricing. The account manager walks into the conversation prepared instead of negotiating from a deficit.

Sized ROI 5 to 10 points of renewal margin recovery, typically $300K to $1.4M annualised

Implementation 4 to 8 weeks. The hard part is contract structure standardisation, not the model.

Sales engineering scoping copilot.

Sales engineers spend 30 to 50 percent of their week on scoping calls and follow-up SOWs for deals that have a 25 percent close rate. The math doesn't work. The bottleneck isn't SE talent. It's the time between the customer's question and a quotable answer.

A scoping copilot trained on the portco's historical SOWs, win/loss data, and standard scope library drafts a first-pass SOW from a discovery call transcript. The SE reviews and corrects in 20 minutes instead of three hours. Closed-won SOWs feed back into the training data. Win rate lifts because response time drops; SE capacity doubles because the cognitive load on each opportunity drops by an order of magnitude.

Sized ROI 2x SE capacity, plus 3 to 6 points of close rate lift on responded opportunities

Implementation 6 to 10 weeks. The SOW library quality determines everything.

Bench utilization root-cause attribution.

Bench underutilization gets blamed on "the pipeline" every Monday in every services firm in America. The actual causes are more specific: a particular skills gap blocks a specific kind of project, two senior people are functionally a single bottleneck, the staffing process favors recency over fit. The data exists in the PSA, the HR system, and the CRM. Nobody has the time to triangulate it.

A weekly model that scores the previous 30 days of staffing decisions against project outcomes and surfaces the top three root causes is enough to drive a real conversation in the Monday leadership meeting. Modest model, massive operational lever. This is the use case the CFO usually champions first.

Sized ROI Recovers 2 to 4 points of utilization within 90 days of being trusted

Implementation 4 to 6 weeks for the model. Change management at the delivery-manager level takes longer.

Sources we monitor for this sector

What the advisory reads weekly.

Channel Futures · MSP industry coverage, MSP 501 rankings, channel-program shifts.
ChannelE2E · M&A activity, PE platform consolidation, vendor partner-program changes.
Omdia MSP coverage · Quarterly deal-count and disclosed-value tracking across the MSP M&A market.
CRN · Vendor product launches, partner program economics, channel-channel relationships.
The MSP 501 list · The annual operator-economics benchmark for utilization, gross margin, and revenue-per-tech.

Five questions to ask before approving an AI purchase at an IT services portco.

The MSP AI vendor pitch has gotten polished fast in the last 18 months, helped along by ConnectWise's zofiQ acquisition and Atera's aggressive product cycle. The questions below are the ones the polish doesn't survive. Ask any one of them on a vendor call and the honest answers separate real solutions from deck-only ones.

Question 01

"What does your integration with ConnectWise, Autotask, or HaloPSA actually look like, three layers deep?"

The PSA isn't optional in MSP land. If the AI vendor can't read tickets, time entries, contracts, and asset data out of ConnectWise Manage, Autotask, or HaloPSA in close-to-real time, the deployment will stall in integration nine months in. Most decks show a logo grid implying full integration. The honest answer is usually a connector built on the API plus a roadmap item for everything else.

Why most vendors get this wrong: they have a clean integration with one PSA (usually Autotask via Kaseya's ecosystem, or HaloPSA via its open API) and use that screenshot to imply parity with ConnectWise Manage. The ConnectWise integration is often the partner-built one, not the vendor-built one, and its data coverage is much thinner than the deck suggests.

Right answer pattern: a working list of named MSP customers running on the same PSA version as your portco, plus the specific named integration partner if the work is done by a third party. If the vendor can't name two customers within a phone call, the integration story isn't real yet.

Question 02

"What's the real Tier 1 deflection rate, measured against a controlled baseline, not a vendor case study?"

Deflection numbers in vendor decks usually mean "tickets the AI touched" not "tickets the AI fully resolved without a human." Real Tier 1 deflection in MSP service desks lands at 15 to 35 percent on routine categories after a six to ten week shadow-mode pilot. Anything above 50 percent in a vendor pitch is a definition problem.

Why most vendors get this wrong: they have a commercial reason to inflate the number. The case-study deflection rates are computed against the customer's worst-performing baseline, with the AI counted as resolving tickets it merely classified. Push for the methodology and the picture changes quickly.

Right answer pattern: the vendor agrees to a 30-day baseline measurement before the AI goes live, a clean definition of "resolved without human touch," and a public deflection number measured against the baseline, by ticket category. If they push back, the number isn't defensible.

Question 03

"Whose data trains the model, and what's the contractual line on shared learning across your customer base?"

Multi-tenant AI service desk vendors get smarter the more MSPs they have. Your portco's ticket history, runbooks, and client environment data is a competitive asset. If it flows into a shared training set, the portco is paying for the privilege of educating its future competitors.

Why most vendors get this wrong: they conflate "your data is private" with "your data doesn't train the model." Those are different statements. The first is about access; the second is about model weights. Many SaaS contracts permit the second under "aggregated and anonymized" clauses that aren't really anonymous at MSP scale.

Right answer pattern: a clean contractual line saying model weights derived from your data stay in your instance and don't propagate to the shared base model. For a PE-backed MSP, this matters double because the platform's data is part of the exit asset.

Question 04

"What's the all-in TCO, including the implementation partner, change management, and the 'AI seats' that creep up after year one?"

The sticker price on an MSP AI deal is rarely the real price. The portco runs a stack of PSA, RMM, documentation, and ITSM that each need integration. The vendor's preferred implementation partner often shows up in month two with a six-figure scope. AI-seat pricing tends to escalate as agent count grows. None of this is in the deck.

Why most vendors get this wrong: the SaaS line item is the only one with their name on it. They have no commercial incentive to surface the implementation partner's scope until you've signed the SaaS contract. By that point, your negotiating room is gone.

Right answer pattern: a single TCO worksheet covering the SaaS line, the implementation partner line, the internal change-management line, and the realistic ramp curve to full deployment. Ask the vendor to put their name on a 24-month all-in number. If they won't, you don't have a TCO.

Question 05

"Why are we buying this instead of building it on Microsoft Copilot Studio or Glean plus a small in-house team?"

A lot of MSPs already pay for Microsoft 365 E5 or Glean and have a small platform team. For Tier 1 ticket triage, knowledge retrieval, and renewal prep, the in-house build on Copilot Studio plus Power Automate plus a connector to the PSA can match a meaningful slice of the vendor's capability at a fraction of the cost. The vendor sale here is time-to-value, not capability.

Why most vendors get this wrong: they pitch "AI is hard, you need us" when the honest answer is the platform team you already pay for can do most of this in 90 days. The vendor's job is to compress that timeline; that's a real value, but it's not the same value the pitch claims.

Right answer pattern: a build-vs-buy worksheet comparing 24-month TCO of the SaaS path against a named in-house build, including the opportunity cost of the platform team's time. For Tier 1 deflection and SE scoping, buy usually wins on time-to-value. For renewal prep and bench analytics, build is increasingly the right call at scale.

AI Advisory for Private Equity Portfolios in IT Services and MSPs.

Why AI moves margin in IT services.

Five AI use cases moving margin right now.

Tier 1 service desk agent for routine ticket categories.

Project staffing optimizer with skills + bench awareness.

Renewal-prep agent that walks contracts against actual delivery.

Sales engineering scoping copilot.

Bench utilization root-cause attribution.

What the advisory reads weekly.

Five questions to ask before approving an AI purchase at an IT services portco.

"What does your integration with ConnectWise, Autotask, or HaloPSA actually look like, three layers deep?"

"What's the real Tier 1 deflection rate, measured against a controlled baseline, not a vendor case study?"

"Whose data trains the model, and what's the contractual line on shared learning across your customer base?"

"What's the all-in TCO, including the implementation partner, change management, and the 'AI seats' that creep up after year one?"

"Why are we buying this instead of building it on Microsoft Copilot Studio or Glean plus a small in-house team?"

Bring an AI advisor into your next MSP diligence call.

Why AI moves margin in IT services.

Five AI use cases moving margin right now.

Tier 1 service desk agent for routine ticket categories.

Project staffing optimizer with skills + bench awareness.

Renewal-prep agent that walks contracts against actual delivery.

Sales engineering scoping copilot.

Bench utilization root-cause attribution.

What the advisory reads weekly.

Five questions to ask before approving an AI purchase at an IT services portco.

"What does your integration with ConnectWise, Autotask, or HaloPSA actually look like, three layers deep?"

"What's the real Tier 1 deflection rate, measured against a controlled baseline, not a vendor case study?"

"Whose data trains the model, and what's the contractual line on shared learning across your customer base?"

"What's the all-in TCO, including the implementation partner, change management, and the 'AI seats' that creep up after year one?"

"Why are we buying this instead of building it on Microsoft Copilot Studio or Glean plus a small in-house team?"

The advisory across other PE sectors.

Bring an AI advisor into your next MSP diligence call.