AI promises to have a huge benefit for our market, but for many, early adoption is creating the opposite, more tools, pilots, disruption and distraction rather than any real return.
According to RAND, 80% of AI projects fail to produce business value—that's twice the rate of traditional IT projects. That raises a critical question for MSPs: What does it take to implement and adopt AI solutions without creating unnecessary cost, complexity and operational drag?
The answer isn’t to avoid AI or to chase the first shiny new tool. It’s to be more deliberate about where (and when) it’s used.
Drawing from his work with thousands of MSPs worldwide, Jason Kemsley, CRO, UpTime Global, sees a recurring pattern. Businesses find or build a new AI tool, get excited about its potential and then try to shoehorn it into their business. The project consumes time and money but does not deliver the expected return.
“I’m not saying don’t use AI. I’m saying that by rushing to deploy, you’re doing yourself a disservice to the issues you actually want to tackle,” Kemsley states. For MSPs, the goal shouldn’t be to be among the first businesses to deploy AI; it should be to be among the businesses that deploy it successfully.
First Movers Don’t Always Have the Advantage
When a new technology as significant as AI emerges, it can feel like a race to implement.
Competitors are announcing AI initiatives. Vendors are releasing new products every week. Customers are asking what you’re doing with AI. The natural response is: We need to move now. But moving first doesn’t necessarily mean gaining an advantage.
According to Kemsley, the MSPs that rush toward AI often end up compromising the foundations that determine whether a tool succeeds in the first place: Inconsistent data, undocumented processes, unclear roles and limited leadership capacity.
“The first 1,000 MSPs to deploy AI may also be the first 1,000 to deal with the backlash when it doesn’t work,” he explains. The goal isn’t to move first; it’s to be deliberate and move intentionally.
Why AI Projects Fail
According to Kemsley, failed AI implementations generally fall into three categories:
- Trying to be a first mover
- Scoping the ROI or project incorrectly
- Leading with the solution instead of the problem
The third is particularly common. They see a tool, get excited, find a problem it solves, force-fit it and wonder why it doesn’t deliver. MSPs that successfully implement AI reverse this process. They identify a problem, find the right solution, deploy deliberately and measure the impact.
Before considering an AI solution, Kemsley recommends asking three questions:
- Does it solve a current business problem?
- Does it improve or maintain customer experience?
- Does it have longevity and a roadmap that aligns with where the business is going?
If the answer to any of these questions isn’t clear, the initiative likely isn’t ready to deploy.
Build the Foundation First
Before asking “What can AI do for us?” Kemsley recommends asking a more fundamental question: “Are we actually ready to use it?” The answer depends on whether four core foundations are already in place:
- Clean data: AI inherits the quality of the data it is given. If your PSA is messy, AI inherits the mess.
- Documented processes: AI can’t reliably automate tribal knowledge. Processes need to be repeatable, documented and assigned a clear owner.
- Clear roles: Teams—and AI—need to know who is responsible for what.
- Leadership capacity: Someone has to own the transition, measure the results and make decisions about what happens next. Without leadership capacity, even a good AI implementation can become another initiative competing for attention.
Once those foundations are set, Kemsley suggests adopting a simple bet, park or kill framework.
- Bet: It’s in production, has an owner and is being measured. Keep it and continue.
- Park: It’s interesting, but structural gaps or prerequisites remain. Date it for later.
- Kill: There is no clear path to production, no owner and no measurable ROI. Cancel it.
The objective isn’t to accumulate AI experiments; it’s to build clear, AI-enabled processes that compound growth over time.
Where MSPs Can Start
Kemsley offers three areas where AI can offer meaningful opportunities for MSPs in the early phases of AI adoption:
- Triage and classification can reduce the time teams spend sorting and routing work.
- Documentation generation can improve consistency and reduce administrative effort – provided the underlying information is clean.
- Predictive alerting can help identify device health issues and anomalies, offering potential ROI when the necessary data foundations are in place.
“The key to implementing a new AI solution is to test the theory before investing heavily in building it yourself. Pilot an existing product to test the problem, measure the value and establish ROI. If it works, then consider how to improve or customize it,” explains Kemsley.
In other words, don’t spend heavily to prove a theory you can test cheaply.
The MSPs that gain the most from AI are not the ones that move fastest; they are the ones that move deliberately. They build the right foundations, start with a real business problem, measure what matters and walk away quickly when a solution does not deliver. In a market crowded with AI noise, that discipline is what turns experimentation into lasting advantage.
GTIA members, check out our library of business-building AI resources on the Member Portal.
Not a member yet? Join now!
