A managed AI platform gives your own technicians a repeatable way to assess, govern and deliver AI across every client. Your own AI engineers (full-time hires sitting in your MSP) and a partner (who can deliver white-label under your brand, embed with your team as an AI Office, or both) cover the custom design and build work your team can't staff yet. If you plan to run AI service at scale yourself, start with a platform; if your clients need systems built and integrated, choose between hiring and partnering, and many MSPs will end up combining options.
By Sagar Pandya, Co-Founder and CEO, Spej. Exited MSP founder.
A note before we start: Spej is option 3 on this page. We offer white-label delivery to MSPs, an embedded AI Office for MSPs and holding companies, or both, and we have our own AI workspace product, Plooms. We have a stake in this question, so we've kept vendor statements attributed to the vendors' own published material, and we'll tell you where a platform or your own hires are the better fit.
Why MSPs are facing this choice now
Clients are asking and few MSPs are billing for it yet. In Kaseya's 2026 State of the MSP Report, 48% of MSPs ranked AI and automation as the top client need for 2026, and 13% were generating meaningful revenue from those services. Nerdio's survey of 153 MSP decision-makers managing Microsoft 365 and Azure found 32% had adopted AI internally, 12% reported measurable productivity gains, and 14% operated a revenue-generating AI service line (Nerdio). The same survey found 60% said management complexity was growing faster than their ability to scale, and 29% had lost technicians to workload-related burnout in the previous 12 months (Telecom Reseller).
So the question for most owners is how to add an AI service line without adding headcount they can't find.
What a managed AI platform does
Platforms built for this give an MSP's existing team a structured process they can repeat for every client. Lemhi is a current example. In its general availability announcement, Lemhi describes itself as "the managed AI services platform for MSPs," giving MSPs "one place to assess, plan, govern and deliver AI for each client," from the first readiness assessment to an ongoing managed AI service, natively inside the Microsoft ecosystem. It's a ConnectWise Invent Certified Integrator and pushes each client's AI tasks into ConnectWise PSA as service tickets, so technicians work them from the queue they already use. General availability started November 4, 2026 at IT Nation Connect, with a free trial at lemhi.com (Lemhi release).
Lemhi's CEO, John Harden, framed the problem this way: "Every MSP's clients are asking about AI, and most MSPs still don't have a service they can sell, deliver and bill for every month." His company also reports from its own assessment data that 64% of employees at companies already using AI personally use at least one AI tool their company hasn't adopted. That's the kind of finding a platform is good at surfacing across a client base.
Nerdio's report also found that MSPs running commercial management platforms were turning AI adoption into measurable productivity gains and revenue faster than those using native tools or community and open-source tooling. Nerdio sells such a platform, so weigh that accordingly, but the pattern makes sense: a shared process beats every technician improvising.
A platform fits well when:
- You want AI service to run through the same PSA, ticket queue and monthly rhythm as everything else.
- Your technicians will run assessments and governance work themselves.
- Most of what clients need is readiness, policy, permissions, adoption of tools like Copilot, and reporting.
- You want a consistent offer you can price per client and repeat across dozens of them.
What a partner does: white-label delivery, an embedded AI Office, or both
A partner supplies the people who design and build AI systems, and it can work with you in two ways.
White-label delivery. In Spej's white-label model, you lead the offer and own and manage the client relationship, and Spej delivers behind the scenes under your brand on agreed terms. Through the partner program Spej brings the security, architecture, governance and delivery work, plus ongoing support, and on Design & Build work the client owns the code and documentation.
Embedded AI Office. Spej also works directly with MSPs and holding companies as an embedded AI Office, sitting with your team and working alongside them. The benefit is a full Spej team that would cost far more to build in-house, without a full-time hire on your payroll. Spej does both modes, and some MSPs use them together.
The work that needs a partner tends to be specific to one client's systems and data. Two examples from our work: a commercial plumbing and drainage manufacturer, where every component price change meant 3 to 5 weeks of manual updates. Price updates went from 3 to 5 weeks to near real time (case study). A regional property and casualty insurer needed claim documents checked against policy terms and previous records, with inconsistencies flagged for a person to review (case study). Both were built into the client's own systems and data.
A partner fits well when:
- A client needs a system built or integrated across several of their applications.
- The client is regulated and needs security design and testing you can't staff.
- You need adoption training delivered at scale.
- You'd rather sell the outcome under your brand, or work with an embedded team, than hire an AI team first.
What hiring or embedding your own engineers looks like
This option means your own full-time AI engineers, hired by you and sitting inside your MSP. They work with your technicians, learn your clients and tools, and build and support AI work as part of your team.
The tradeoffs come down to cost, control and hiring:
- Control. Your engineers answer to you, know your stack and clients, and stay with the business. Whatever they learn builds up inside your company.
- Cost. You carry salary, benefits, tooling and management time whether or not client work fills their calendar. That's a fixed cost you take on before the AI service line has revenue behind it.
- Hiring. Experienced AI engineers are hard to find and hard to keep, and most MSPs have no one in-house who can judge candidates or set their direction. Plan for a long search and for ramp-up time after someone starts.
- Coverage. One or two hires have limited range. A single person is unlikely to cover security, architecture, governance, integration and training at the depth a regulated client expects.
This route fits when you expect steady AI work across your client base, you have a technical leader who can manage the team, and you want the skills to live inside your company for the long term.
Side by side
| Managed AI platform | Your own AI engineers | Partner (white-label or embedded AI Office) | |
|---|---|---|---|
| What you get | Software and a repeatable process your team runs | Full-time hires who build and support AI work inside your MSP | People who design, build and support AI work under your brand, or embedded with your team as an AI Office |
| Who does the work | Your technicians, guided by the platform | Your own engineers, with your technicians | The partner, with your team on the relationship and support you keep |
| Speed to start | Fast, since it's software your team can begin using once it's set up | Slowest, since you need to recruit, hire and onboard first | Fast, since the people already exist and the team is ready to start on agreed terms |
| Control | Your team controls delivery, inside the platform's process | Highest, since the team is yours | Shared, with the relationship and brand staying yours and delivery terms set in writing |
| Cost structure | Recurring software cost, passed through in a per-client service line | Fixed payroll and management cost, whether or not work is booked | Project fees and retainers, with no payroll commitment |
| Hiring burden | None beyond training your existing technicians | Heaviest, covering recruiting, retention and management | None, since the partner brings the team |
| Client relationship ownership | You own it | You own it | You own it in white-label; in an embedded AI Office you work alongside the team |
| Best for | Assessments, governance, permissions, adoption and reporting across many clients | Steady, ongoing AI work where you want the skills in-house long term | Custom builds, integrations, regulated security work, training at scale |
| When it fits | You have capable technicians and a repeatable offer to sell | You have steady demand and a leader who can manage engineers | You need specialist depth now and want to avoid a full-time hire |
| How it scales | Across your whole client base with the same process | With each hire, and limited by their range | Project by project, plus ongoing support where agreed |
| How you charge | A recurring managed AI line per client | Whatever mix of service fees and retainers covers your payroll | Project fees and retainers at your own pricing on agreed delivery terms |
| What to check | PSA fit, data handling, per-client pricing, exit terms | Hiring plan, who manages them, what happens if someone leaves | Who owns the client and support, who owns the code, how quality is checked |
Using both (or all three)
The options don't compete for the same job. A platform can carry the repeatable work across every client, the assessments, governance tasks and monthly reporting. A small in-house team, or a partner, takes the projects that need custom design. When a platform assessment turns up a high-value workflow that needs a real build, that's the handoff point.
Some MSPs bring in a partner first, white-label or embedded, to deliver early client work while they decide whether to hire. Others hire one or two engineers and use a partner for the security, architecture or training work those hires can't cover. Whichever mix you pick, agree in writing who supports the client after go-live, and keep the ticket in your PSA either way.
Questions to ask any of them
- How does this fit our PSA and the way our technicians already work?
- Where is client data processed, and what boundaries can we set per client?
- How are permissions reviewed before AI reads client files? Microsoft notes that Copilot only surfaces organizational data a user has at least view permission to (Microsoft Learn), so access cleanup comes first.
- Who supports the client after go-live, and how do escalations reach the right person?
- How is pricing structured per client, and what can we charge on top?
- What happens to our configurations, documentation and client data if we leave?
- Can you show a client result with a starting number and an outcome?
Where each option falls short
A platform gives your team a process, and your team still has to have the skills and time to run it. If technicians are already stretched, a new service line adds load before it adds revenue.
Your own AI engineers give you control and a team that stays, and they come with a hiring search, payroll before revenue, and limited range per person. If a hire leaves, the knowledge they built leaves too unless it's documented.
A partner gives you skilled people without a full-time hire, at project cost and with coordination overhead. If you white label, your brand carries the result, so vet delivery quality before you put your name on it. If you embed a partner team, the skills live with the partner, so agree up front how knowledge gets passed to your staff.
None of them fixes messy permissions or unreliable data by itself. Someone has to own that work, and it usually needs to happen before AI touches client files.
If you're weighing a partner, talk with our partner team. For the services to put on your menu first, see how MSPs can add AI services. Related: AI referral fees, AI implementation, the AI Office, and more client work.
FAQ
What is a managed AI services platform for MSPs?
It's software that gives an MSP a repeatable way to run AI work across many clients: readiness assessments, plans, governance and ongoing service, usually tied into the PSA and Microsoft tools the MSP already uses. Lemhi, for example, describes itself as the managed AI services platform for MSPs and launched general availability on November 4, 2026.
What does a white-label AI delivery partner do for an MSP?
A white-label delivery partner does the AI implementation work under the MSP's brand: security, architecture, governance, delivery and ongoing support. The MSP owns and manages the client relationship, and the partner works behind the scenes on agreed delivery terms. Some partners, Spej included, can also embed with the MSP's team as an AI Office.
Should an MSP buy an AI platform, hire AI engineers or partner with an AI firm?
Buy a platform if your team will run AI assessments, governance and recurring service for many clients and you want one repeatable process. Hire your own engineers if you expect steady AI work and want the skills in-house long term. Bring in a partner, white-label or embedded, when clients need custom builds, integrations or specialist security work and you'd rather not make a full-time hire. Many MSPs will combine options.
Should we hire AI engineers or partner?
Hire if you have steady demand, someone who can manage engineers and the time to recruit, since you carry payroll whether or not work is booked. Partner if you need specialist depth soon, want to test demand first, or don't have a technical leader for a new team. A partner can deliver white-label under your brand, embed with your team as an AI Office, or both, without a full-time hire. Some MSPs start with a partner and hire later once the work is proven.
Can an MSP combine a managed AI platform, its own engineers and a partner?
Yes. The platform can run the repeatable work across your client base, such as assessments, governance tasks and reporting, while engineers or a partner handle the projects that need custom design and build. Agree who owns support and the client relationship for each project before you start.
How do MSPs make money from AI services?
Through fixed-fee assessments, project fees for pilots and builds, and a monthly managed AI service. Few have done it yet: Nerdio's 2026 State of the MSP Business Report found only 14% of surveyed MSPs operated a revenue-generating AI service line.
What should an MSP ask before choosing any option?
Ask how it handles your PSA and ticketing, where client data is processed, how permissions are reviewed before AI reads client files, who supports the client after go-live, how pricing works per client, and what happens to your work and data if you leave.
If you'd like to see how Spej works with MSPs and holding companies, white-label or as an embedded AI Office, you can read more on our partners page.