Kief Studio

Your Clients Want AI Services. Most MSPs Can't Deliver Them Yet.

Support

  • Published
  • Updated
  • 10 min read
Soft violet-magenta and gold curtains of light ripple across the upper frame like an aurora above a dark empty horizon.

TL;DR: Clients are asking their IT providers for AI services, and most providers are not ready to sell them yet. Census data puts business AI use near 20%, and only about a quarter of North American IT providers say they earn AI revenue. The gap is governance and delivery, not interest. Partnering often beats building alone.

Update, October 2026: an earlier version of this note led with survey figures from an industry software vendor and added salary math, market size projections and anecdotes we could not trace to a primary source. We rebuilt it around government data, an industry association's survey, academic research and analyst statements we could check.

If you run a managed service provider, you have heard the question by now: "Can you do AI for us?" Clients want AI services from the people who already run their email, their laptops and their backups. The demand is real. What most providers are missing is a way to deliver AI services that hold up after the demo, with an owner, a policy and a support plan behind them.

Here is what the public data says about that gap, why hiring your way across it is hard, and what a workable path looks like for a small or mid-sized provider.

How many businesses actually use AI services today?

Fewer than the headlines suggest, and that is good news for a provider who wants to get AI services right instead of first.

The U.S. Census Bureau asks businesses every two weeks whether they used AI in a business function. For the period ending May 3, 2026, the national rate was 19.8%, and from December 2025 to May 2026 "overall AI usage hovered between 17% and 20%." Looking ahead, "between 20% and 23% of businesses expected to be using it in the next six months." (U.S. Census Bureau (opens in a new tab))

Size matters. Census found "37% of firms with at least 250 employees reported using AI in their business operations," while "Less than 20% of firms with four or fewer employees reported using AI." Some sectors run well ahead of the national rate: Information at 39.7% and Finance and Insurance at 33.9%. (U.S. Census Bureau (opens in a new tab))

Put those numbers next to each other. Most of your small clients are not using AI in their operations yet, but a meaningful share expect to start soon, and your larger clients are already there. That is a client base asking for help with AI services, not one that has already hired someone else.

What clients mean when they ask for AI services

In my experience the request is rarely "build us a model." It is closer to one of these:

  • "Our staff are pasting customer data into chat tools. Is that okay?"
  • "Can something draft replies to the routine tickets and emails?"
  • "Can we search our own documents and get a straight answer?"
  • "Our software vendor turned on an AI feature. Who checks what it does?"

Two of those four are governance questions, not engineering ones. That matters for how you package AI services, and we will come back to it.

Why can't most MSPs deliver AI services yet?

The channel's own industry association has been measuring this. GTIA, the Global Technology Industry Association, surveyed 1,042 channel company executives and owners in December 2024. It found that "half of North American channel firms anticipate significant AI-driven sales growth in the next two years," up from one-third the year before, with AI "surpassing cybersecurity as the predicted leader of revenue growth." (GTIA via PR Newswire (opens in a new tab))

Expectation is not revenue. In its 2026 State of the Channel summary, GTIA reported that North America leads in AI maturity, "with one-quarter of ITSPs self-identifying as 'AI-driven' organizations actively generating AI revenue." (GTIA (opens in a new tab)) So roughly three in four providers in the most mature region are not there yet.

There are three reasons I see over and over, and the public data backs each of them.

The skills gap is real, but it is not the whole story

GTIA's same survey found skills shortages are "now cited by nearly four in 10 channel companies as an inhibitor to success," and that "firms are facing increased strain on capacity and resources." (GTIA via PR Newswire (opens in a new tab))

But hiring one AI engineer does not give you AI services you can sell. Someone still has to decide which client data the system may touch, write the acceptable use policy, test outputs before they reach a customer and own the thing when it misbehaves at 2am. Those are operations and governance jobs. Most MSPs already do that kind of work for email and endpoints. They have not built it for AI yet.

Many AI projects never make it to production

Gartner predicted in June 2025 that "over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls." (Gartner (opens in a new tab)) Gartner analyst Anushree Verma described many current efforts as "early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied." (MarTech (opens in a new tab))

Gartner also flagged "agent washing," vendors relabeling chatbots and older automation as agents, and estimated that among thousands of vendors claiming agentic products, "only around 130 offer real agentic features." (MarTech (opens in a new tab)) If you resell AI services built on one of the others, your client's disappointment lands on you, not the vendor.

Packaging AI services is a commercial problem

An MSP can use AI internally for ticket summaries or patch notes and still have nothing to sell. Selling AI services means a scope, a price, a service level, a support path and a clear line on what the system will not do. Most providers have that for backup and security. Few have written it for AI, which is why "we use AI" so rarely turns into "we earn AI revenue."

Is it better to build AI services in-house or partner?

The best public evidence I know of points toward partnering for most small and mid-sized providers, at least to start.

MIT's NANDA initiative studied generative AI deployments for its 2025 report, The GenAI Divide. Fortune summarized one of its clearest findings: "Purchasing AI tools from specialized vendors and building partnerships succeed about 67% of the time," while internal builds succeed about one-third as often. The same research found "the biggest ROI in back-office automation." (Fortune (opens in a new tab))

That second point is useful for an MSP. The back office is exactly where you already have trust and access: ticket queues, onboarding paperwork, document search, invoice matching. Those are better first AI services than a customer-facing chatbot with your name on it.

When building in-house makes sense

  • You already have engineers who write and maintain production code, not only scripts.
  • You have several clients asking for the same AI services, so one build serves many.
  • You can staff on-call for it, the same way you staff on-call for servers.

When partnering makes sense

  • Clients are asking now and your bench is already full.
  • Each request is different and you cannot predict which will repeat.
  • You want to learn which AI services sell before you hire for them.

A partner arrangement is also how most providers already handle specialties like penetration testing or custom development. White-label development works the same way for AI services: a build partner does the engineering, and you keep the client relationship and the brand. I wrote more about where white-label arrangements go wrong for MSPs (opens in a new tab), and it is worth reading before you sign anything.

What governance do AI services need before you sell them?

This is the part most providers skip, and it is the part clients will judge you on later.

Start from a public framework

You do not need to invent a governance model. NIST's AI Risk Management Framework, released January 26, 2023, "is intended for voluntary use and to improve the ability to incorporate trustworthiness considerations" into AI systems, and NIST added a Generative AI Profile (NIST AI 600-1) on July 26, 2024. (NIST (opens in a new tab)) Its core "is composed of four functions: govern, map, measure, and manage," and govern "is a cross-cutting function that is infused throughout AI risk management and enables the other functions." (NIST AI Resource Center (opens in a new tab))

For an MSP, that maps cleanly to a service you already know how to run:

  1. Govern: a written AI use policy per client, with a named owner on their side and yours.
  2. Map: an inventory of where AI already runs in the client's tools, including features vendors switched on.
  3. Measure: checks on output quality before anything reaches a customer, with a log of what failed.
  4. Manage: a way to turn a feature off, a person to escalate to and a review on a set schedule.

Amelia, my wife and Kief Studio's co-founder and CEO, wrote about governing AI agents by sequence instead of approving each click (opens in a new tab), which is a practical way to set that policy without slowing every task down. Our knowledge base also covers AI agent governance for a small business and AI quality gates.

Be careful what you claim

Marketing AI services carries its own risk. In September 2024 the Federal Trade Commission announced Operation AI Comply, a sweep against deceptive AI claims. Lina M. Khan, the FTC's chair at the time, said, "Using AI tools to trick, mislead, or defraud people is illegal," and "there is no AI exemption from the laws on the books." (Federal Trade Commission (opens in a new tab))

Practical translation: do not promise a client that AI will replace staff, cut costs by a set percentage or never make mistakes. Describe what the system does, what a person checks and what happens when it is wrong.

What should an MSP do this quarter?

You do not need a new division. You need a short list of AI services you can actually deliver and support.

  1. Ask your top ten clients one question. "Where is AI already running in your business, and who approved it?" The answers become your first assessment offer.
  2. Write one AI use policy template. Base it on the NIST functions above. Customize it per client.
  3. Pick one back-office workflow. Ticket summaries, document search or onboarding forms. Run it internally first.
  4. Define the service. Scope, what a person reviews, response times, how to switch it off. If you cannot write that page, you are not ready to sell it.
  5. Decide build or partner per workflow. Use the lists above. Revisit after two or three clients.
  6. Measure for 30 days. What it handled, what it sent to a person, what failed. That report is also your renewal conversation.

Notice that four of the six steps are things an MSP already does for every other service. That is the point. AI services are a managed service with a model inside, and the providers who treat them that way will be the ones still selling them in two years.

Frequently asked questions

What are AI services for an MSP?

AI services are managed offerings with AI inside: an AI use policy, an inventory of AI features already running, automation for back-office work like ticket summaries or document search, and monitoring of what that automation does.

How many businesses use AI right now?

About one in five. The Census Bureau put the national rate at 19.8% for the period ending May 3, 2026, with 37% of firms with at least 250 employees using AI and fewer than 20% of the smallest firms.

Do MSPs need to hire AI engineers to offer AI services?

Not to start. Many providers partner for the engineering and keep the client relationship, policy and support. MIT research summarized by Fortune found purchased tools and partnerships succeeded about twice as often as internal builds.

Why do so many AI projects fail?

Gartner points to escalating costs, unclear business value and inadequate risk controls. In practice that means projects started without a defined workflow, an owner or a way to check output before it reaches a customer.

Is there a standard for governing AI services?

The NIST AI Risk Management Framework is a free, voluntary framework built on four functions: govern, map, measure and manage. NIST also publishes a Generative AI Profile that applies it to tools like chat assistants.

Can an MSP white-label AI services?

Yes. A build partner can do the engineering while the work ships under your brand and contracts, the same way many providers already handle custom development or penetration testing.

This article is general information, not legal advice. Kief Studio is not a law firm, and reading it does not create an attorney-client relationship. Laws and regulations change, so check the current text and talk with a licensed Massachusetts attorney about your situation.

Where we fit

Kief Studio does white-label development for agencies and studios, and for partners who need dependable engineering behind their brand. We build it, your name stays on it, and we stay out of the client relationship unless you bring us in. When the engagement includes it, we host and maintain the work under your name.

On the AI side, our AI implementation work takes an idea out of the demo and into daily work: retries, error handling, quality gates on every output and a human path for anything the model is unsure about. We run it on LTFI (opens in a new tab), so nothing is installed on your client's side. If your clients are asking for AI services and your bench is full, that is the conversation to have.

First conversation is free. No commitment.

Tell us what your clients are asking for

Related articles

  1. Build

    Kaseya Promised Managed Service Providers 5x Capacity Without Hiring. Here Is What That Exposes.

    Kaseya pitched managed service providers 5x capacity with agentic AI. What was announced, what BLS, CISA and Gartner say, and where to point the freed hours.

    Brian Gagne

  2. Support

    Your Clients Want AI Services. AI Service Delivery Is the Support Ticket After the Demo.

    AI service delivery fails after the demo, not during it. What Klarna, Air Canada and research show, and support terms to write before you sell an AI agent.

    Amelia Gagne

  3. Support

    What Managed IT Services Actually Include, With Real Examples

    Managed IT services examples for Worcester and Central MA: patching, help desk, access, backups, monitoring, vendors and reviews, plus how to read a scope.

    Brian Gagne