Artificial intelligence (AI) can make your service desk faster while leaving your business less profitable. The tools cost money immediately. The freed hours become valuable when you turn them into lower costs, more clients or a broader service.
That is the commercial challenge behind the Managed AI chasm. What can a managed service provider (MSP) sell once more of the technical work takes care of itself?
We read Part II of Top Down Ventures’ The New Era of MSP, published with the Global Technology Industry Association (GTIA) in October 2026, and agree with its central argument: the opportunity is to take responsibility for more of the client’s business operations. The report’s most useful ideas connect that broader offer to the economics, the client relationship and the evidence needed to charge for results.
The client is bringing AI to you
Top Down opens with a reversal. MSPs introduced clients to managed information technology (IT), cloud and security. This time, clients are arriving with tools they already use.
Lemhi, a Top Down portfolio company, interviewed more than 100 MSPs. The most common first approach was a support ticket asking for Claude to access the client’s systems, according to the report (p. 25). A permissions question can quickly become a business question: which work should this tool do, and who will take responsibility when it does?
The report describes three layers of service. Managed IT and security keeps the technology running, with AI improving delivery behind the scenes. Managed AI adds governance, readiness assessments, adoption support and initial automations. Managed business process outcomes (Managed BPO) takes responsibility for a defined piece of the client’s work.
That last layer changes the buying conversation. A software deployment has a completion date. A process that needs to run correctly every week creates an ongoing service obligation.
Top Down sees three forms of Managed BPO (pp. 27 to 29):
- Operate the software. Configure and run the client’s business applications, adapting them as its needs change.
- Operate a workflow. Take on a repeatable process in administration, sales, customer service or another function, including the handoffs between people and agents.
- Operate a function. Deliver a back-office service such as managed accounting, with the specialist expertise it requires.
The workflow opportunity includes work a small business currently leaves undone. Follow-ups slip and reports arrive late because the records behind them never get reconciled. Improving that work can create value without replacing an employee.
For an MSP, the first commercial decision is therefore which client problem to own. Buying an agent comes later.
The strongest position is between the platforms
An MSP selling generic ticket automation is competing with features its core vendors are adding. Its stronger position is work that crosses systems and requires knowledge of the client.
Top Down draws that boundary explicitly: incumbents are well placed to automate jobs inside their platforms; specialists have more room in workflows spanning several vendors and the customer’s operations (pp. 44 to 45).
Consider an invoice workflow. Reading the invoice is one step. Completing the job can involve the accounting system, purchasing records, approval rules and a person who knows how to handle a disputed charge. Someone must own those connections and exceptions. That is the service the client can keep buying as the underlying tools change.
This also explains Top Down’s view of Microsoft. It expects pressure on commodity license resale while seeing continued demand for partners who handle adoption, operations and governance across Microsoft and other providers (pp. 48 to 50). The lesson is to make that work a visible, separately priced service.
The gap between a demo and a service
Top Down’s stack map has three tiers: infrastructure, orchestration, and agents and delivery. Its 2026 assessment identifies twelve gaps, eight with no incumbent activity (pp. 31 to 35).
The report highlights four that matter when you are choosing a vendor:
- Model routing: can you choose models for quality, speed and cost, or are you tied to one provider?
- Agent evaluation: can you see whether work was correct, including errors and human intervention?
- Portable governance: can permissions and policies cover agents from different vendors?
- The owner’s view: can the client understand what AI is doing, what it costs and where it is useful?
The map is a buying checklist. A claim that a platform supports agents says little about whether you can run a dependable client service on it.
Top Down’s build-or-buy advice is equally practical: retain established core systems and build client-specific workflows around them. Replacing the software that holds your contracts, billing or security records creates a permanent maintenance obligation. Connecting those records to a well-defined workflow can be a manageable project.
Before your next renewal, ask for a demonstration using the actions your service needs. Check access through the Model Context Protocol (MCP), which connects AI tools to existing software, alongside permissions, recovery, data export and agent pricing. A line on a feature list tells you the endpoint exists. The demonstration tells you whether it does the work.
Inside the $10 million MSP example
Top Down estimates that agents could perform around 40% of an average technician’s workflow by 2028. Its illustration of a $10 million MSP shows why that forecast matters, and where an owner needs to challenge the arithmetic (pp. 51 to 52).
| The report’s assumptions | Annual amount |
|---|---|
| Revenue, with roughly 40 employees | $10 million |
| Total payroll | $4.2 million |
| Technician payroll within that total | $2.5 million |
| Technician capacity released, valued at 40% of payroll | $1 million |
| Additional agent and tool cost | $300,000 |
There is a $1 million capacity opportunity and a $300,000 bill. How you use the capacity determines what happens next.
The report illustrates a margin route: adjusted earnings before interest, taxes, depreciation and amortization (EBITDA) rising from about $1.2 million to $1.9 million. That requires the modeled $700,000 net benefit to become a real economic saving. Keeping revenue, headcount and payroll unchanged while adding tools does not produce that improvement.
The growth route is different. At the report’s average of $2.32 in service revenue per dollar of service wages, the released capacity could support about $2.3 million of additional services. Those services still need customers, pricing and successful delivery. Treat the figure as delivery capacity. Turning it into revenue is a sales job.
An MSP could also avoid its next hire as demand grows. That improves future economics without cutting the existing team. Across all these routes, supervision, rework and implementation consume some of the available hours.
The management task is to track the conversion: hours released, hours spent supervising, costs actually removed and additional profitable work delivered. Give someone responsibility for agent quality and exceptions. A utilization estimate alone cannot tell you whether the investment pays.
The tool bill is already testing margins
The report’s historical argument is encouraging. Earlier reductions in delivery work gave MSPs room to add networking, cloud, security and compliance. Broader scope helped providers maintain and increase revenue even as existing tasks became cheaper.
But the recent figures are a better guide to the next purchasing decision. Service Leadership’s 2021 to 2025 data, reproduced by Top Down, shows rising tool costs alongside falling service gross margin (pp. 54 to 56).
Set a ceiling for tool spending as a share of revenue. Test whether a new product replaces an existing cost, improves delivery or enables a service customers will buy. Include supervision and consumption charges in that test. It is possible to add capability and lose margin in the same purchase.
Price the work you can prove
Outcome pricing is attractive because a client’s willingness to pay can exceed the cost of the hours needed to deliver. It also creates a new obligation: agreeing what counts as a successful result.
Top Down describes pricing by agent, action, workflow and outcome. It expects hybrid contracts to become prominent: a fixed retainer with a variable fee linked to results (p. 53). That lets recurring revenue continue while the unit being sold changes.
We would use the fixed component to cover ongoing operation, monitoring and support. Any performance component should attach to a result the MSP can influence and both parties can verify. Charging for revenue growth while controlling only one small part of the sales process creates an avoidable dispute.
An invoice-processing pilot makes this concrete. Agree the starting processing time, volume, exception rate and human review required. Then define what a completed invoice means, which exceptions stay with the client and how the service will be measured. This is a pilot design; the numbers come from the pilot once it runs.
The commercial product is the whole operating arrangement: setup, routine execution, oversight and exception handling. Its price needs to account for all four. A successful pilot should tell you both whether the client gets a worthwhile result and whether you can deliver it at a sustainable margin.
Your next growth engine may be the clients you already have
Managed BPO gives an MSP more scope to sell within an existing relationship. Customer success is how that scope becomes visible and credible.
Drawing on Chris Day’s MSP to MVP, Top Down argues that the industry has concentrated on winning clients while underdeveloping the work after the sale: onboarding, demonstrating impact and expanding the account (pp. 56 to 60).
The useful measure is net dollar retention (NDR): recurring revenue retained from a starting client group after expansion, contraction and churn, excluding new customers. A cohort moving from $10 million to $11 million produces 110% NDR. Top Down proposes 120% as an ambition for MSPs, equivalent to 20% net growth from the existing cohort, and calls it a target to aim for; no benchmark puts the industry there today (p. 57).
That is a more demanding test than keeping logos. It asks whether clients find enough additional value to deepen the relationship.
It also requires judgment about where to invest. A client eager to redesign operations may justify advisory work and pilots. One that wants dependable email may be a good account for a simpler service. Selling the same AI program to both makes delivery expensive and the second client uncomfortable.
Keep the context that makes the next service better
Top Down’s final argument is that an MSP’s valuable knowledge extends beyond tickets and device data. It includes the owner’s goals, past decisions, approval rules, business reviews and accepted results (pp. 60 to 62).
That context explains why a workflow works for one client and fails for another. It also tells you where to propose the next service. A model provider can supply intelligence; the MSP still has to connect it to the way this business operates.
Build a governed, exportable record of that knowledge. Record the exceptions and the lessons from each completed workflow, then use them to improve the next run. Keep client permissions and confidentiality intact.
This is where delivery, retention and marketing meet. An agreed baseline and a result the client accepts support the renewal conversation. With permission, the same evidence supports a case study that helps the next buyer understand your offer.
Five decisions for the next quarter
Top Down closes with an operator playbook. We would turn it into these commitments:
- Choose a workflow and the right clients. Start with a small group that trusts you and wants the change. Agree a baseline, a result and a paid pilot scope.
- Name the person responsible for agent quality. Give them authority over testing, permissions, human review and recovery when work fails.
- Decide how released capacity will earn its return. Identify the cost you can remove, the hire you can avoid or the additional service you can sell. Track what actually happens.
- Challenge the stack before adding to it. Test cross-vendor access, governance, accuracy and full cost. Set a spending ceiling and remove duplication.
- Make client outcomes part of the commercial review. Keep the agreed evidence, track expansion and contraction, and offer the next service when the results justify it.
Then make the offer clear on your website. Name the workflow, the client it suits, what you operate and how success is measured. A business owner needs enough detail to decide whether to have a conversation.
Crossing the chasm starts with one service that runs reliably, earns a margin and has results the client has accepted. Once you have that, you have an account worth expanding and a specific story to tell the market.
If you want to assess how clearly buyers can find and evaluate your MSP’s offer, book a call.
Sources and scope
This article interprets Part II, “Crossing the Managed AI Chasm,” printed pages 24 to 63 of Top Down Ventures’ The New Era of MSP: From Managing Technology to Delivering Outcomes, 2026 edition, powered by GTIA. Page references throughout refer to the report’s printed pagination.
The workforce forecast and $10 million illustration are Top Down’s estimates. The margin, cost and service-revenue-to-wage figures are cited through its reporting of Service Leadership. Lemhi is a Top Down portfolio company. The pricing recommendations, the invoice example and the practical reading are our own; no client result sits behind them.