Getting your Trinity Audio player ready...

Why Your PSA Is A Hidden Goldmine

Most MSPs treat their PSA as a utility. It is where tickets get logged, time gets tracked, and workflows stay organized. What often goes unnoticed is the deeper value hiding inside. Beneath the mountain of closed tickets and archived boards is a dataset with the potential to reshape how your business operates entirely. The problem for most is simply that it is trapped inside with no way to extract the true value. That is, until AI has entered the chat. 

I first started thinking seriously about this when I came across a LinkedIn post from Lee Silverstone, the Founder and CEO of Zofiq. In the post, he made a strong case for the actual value (in numeric terms) sitting inside the average MSP’s PSA. It was not vague or theoretical, but calculated by real-world data. I immediately reached out to better understand just how much value is really there, and how many MSPs are actually making use of it today.

Here’s what Lee had to say as we dug deeper on the topic: 

The Untapped Value of MSP Ticket Data

Most MSPs are sitting on thousands of tickets, each one a breadcrumb that tells the story of how their business (and their client’s business) operates. But according to Silverstone, the real opportunity isn’t just in volume. It’s in recognizing the value that those tickets represent when properly labeled and used as training data. “By my estimates, every MSP is sitting on $30 to $500 worth of value for every 1,000 tickets in their PSA. There are two prongs here: the quantity of tickets and the quality of those tickets. Together, those factors create the value that often goes untapped.”

The math behind those numbers might surprise you. “Those numbers come from looking at the real-world cost of labeling data. If you hired a human to label tickets, you would be paying anywhere from three cents to fifty cents each depending on complexity. That is the baseline value sitting in your system before you even start thinking about what AI can do with it.”

And that baseline is only part of the story. Once AI enters the picture, the upside increases dramatically. “I think we know where the floor is, but we do not know where the ceiling is. The floor is zero or even negative if the data is bad. But the ceiling could be unlimited, because a single ticket could prevent churn on a massive customer, and the impact compounds every time the AI applies that learning again.”

Minimum Tickets Required To Train AI 

One of the most common myths around AI in the MSP space is that it only works at a massive scale. The belief is that unless you’ve got a million well-labeled tickets, it’s not even worth trying. But according to Lee Silverstone, that kind of thinking is flat out wrong. “We’ve seen PSAs with tens of millions of tickets, and we’ve seen PSAs with just tens of thousands. The bad advice out there is that you need a million perfect tickets before you can get value out of AI. That is an old school mentality, the reality is you can get a lot of value without going to that extreme.”

There is, of course, a lower bound. You still need enough data for a pattern to emerge. “You need as few as 100 tickets, but that can go up to 3,000 per use case to build something impactful. It really depends on the task or workflow. That is really the sweet spot. Once you get into that range, you start unlocking significant value.”

The bigger issue, Lee says, is that MSPs are underestimating what they already have. Most are sitting on enough ticket volume to power at least a few focused use cases, they just don’t realize it. “I think the reality is most MSPs don’t understand how valuable their data actually is. They are sitting on something that could make their business stronger, but they are not leveraging it the right way. And that is the main message we are trying to get across.”

Recent & Relevant Data Matters Most

Not all ticket data holds the same value. Just because data exists doesn’t mean it’s useful, especially when environments, tools, and processes have changed. Lee Silverstone is quick to call this out, noting that outdated data can do more harm than good. “There is a point of diminishing returns when you are looking too far back in the past and things have changed. If a customer used to be on SonicWall and now they are on Cisco, you are just confusing the model if you train on that old data. What matters is the recent activity that actually reflects how your MSP operates today.”

This is why the default instinct to use everything in the archive can backfire. Too much irrelevant context muddies the waters instead of adding clarity. “A lot of MSPs are sitting on archived boards full of tickets that no longer matter. The labels are outdated, the customers may not even be active anymore, and the information is irrelevant. The way we think about it is that six months to a year of data is usually enough to get the job done.”

Lee likens this to how you would train a human, which puts the point into perspective. “If you were teaching a person about your processes, you would never hand them tickets from three years ago where the environment is totally different. The same principle applies here. If you would not show it to a human, then you should not show it to the AI.”

The Net-Negative of Bad Data

Training AI is not all that different from onboarding a new support technician. If the inputs are unclear, inconsistent, or outdated, the results will reflect that confusion. In short, poor training leads to poor performance. The same applies to AI systems that rely on historic ticket data to make decisions. “When your data is super messy, it just becomes less valuable. If you train an AI model on that type of data, you end up needing to set wide confidence intervals. That means the agent will often skip tickets because it cannot be certain enough to take action.”

According to Lee Silverstone, this is where some vendors go wrong. “Bad AI companies will take all the data, train an agent, and let it run on everything. What happens is it mislabels, miscategorizes, or even solves the issue wrong. That not only kills ROI, it can actually create a net negative impact if you are feeding really poor data into the system.”

The inverse is also true. There are vendors that will take the last few tickets, or even worse, the current ticket alone, and drop it into an LLM hoping for the best. Lee points out how dangerous that mindset can be. “This is where things quickly go off the rails, and why it’s important to work with a vendor that actually understands the data science behind what they are selling. Too many people just think throwing an LLM at the problem is training, and it isn’t.”

And even well-built systems will start to underperform when the input is flawed. “When you have poor data, properly built AI agents will not make decisions because the guardrails stop them from acting without certainty. The result is that fewer tickets are handled and the overall ROI goes down. In other words, bad data directly reduces the value of the system.”

The Compounding Value of Agents

Once an AI agent learns how to handle a specific type of issue, the payoff doesn’t just happen once. It builds over time. Silverstone explains how even a single ticket can lead to significant long-term value. “You could have a single ticket with telemetry data in it that allows an AI agent to proactively deal with that issue in the future. If that prevents a $200,000 ARR customer from churning, then that one ticket is worth however much you value not losing that customer.”

The impact becomes even more powerful as the same problem surfaces again. Each time it happens, the system is already equipped to respond. “The beauty of AI is that once it learns from an event, it does not just stop there. The third, fourth, and fifth time the same issue comes up, it is already productive. That compounding effect is where the value really starts to grow.”

In high-stakes environments, this compounding benefit can act as a safety net. It may even catch problems that would have otherwise led to a massive blowout. “MSPs sometimes experience catastrophic events that cause them to lose customers, and those losses are enormous. If agents can prevent even one of those outcomes, the value of the tickets that informed those actions is essentially unlimited. That is why we do not yet know where the ceiling of ticket value really is.”

Guardrails, Humans, and the Path to ROI

AI might make a great engine, but it still needs a driver. Without proper boundaries, even the best-trained agent can cause damage. That’s why Lee Silverstone emphasizes the importance of clear constraints when deploying AI in real MSP environments. “At Zofiq we are very big on saying, if it is not one hundred percent certain, do not let the AI make a decision. You have to put the right guardrails in place. Otherwise you risk mislabels, wrong assignments, or broken SLAs, which can do more harm than good.”

Of course, mistakes will happen, which is why humans still play a critical role in the loop. “When an AI agent assigns a ticket to the wrong person, a human can step in and correct it. The AI then learns from that feedback and updates its model of the world. It is the same way you would teach a new employee, by sitting next to them and helping them learn as they go.”

AI on its own isn’t enough to deliver consistent results. The real power comes from combining automation with human oversight. As the system learns and improves, your team is still there to guide and course-correct when needed. That dynamic is what allows AI to become a reliable extension of your service operation without putting customer experience at risk.

How Ticket Data Is Driving M&A

While many MSPs are still wrapping their heads around how to use ticket data internally, some buyers are already a few steps ahead. According to Lee, private equity firms and strategic acquirers have started baking data value into their investment strategy. “I can tell you right now that private equity firms and roll up plays are already factoring this into their thesis. It is literally in their deck. They see the ability to turn MSP data into a flywheel effect, and that is a huge opportunity.”

The potential to create scale through data is an entirely new phenomenon in Managed Services.. Historically, growth often came at the cost of service quality. But with shared data and AI agents that can learn across environments, every new acquisition has the potential to improve the overall experience (not the opposite). “For the most part MSPs have not had a flywheel effect in their business. Services usually get worse as they grow. But with AI and shared data, every acquisition makes the next customer experience better, which flips the model on its head.”

And this shift isn’t theoretical. It’s already happening in the market. “There are deals happening today where the value of the ticket data is part of the play. When one MSP buys another, now they share data, and the AI agents can learn from both environments. I have personally seen this at least three times in the last two months.”

Data Sharing and Ecosystem Value 

When you think about it, the value of ticket data doesn’t have to stop at the boundaries of a single MSP. Lee sees potential for broader collaboration, especially among peers serving different markets. “There is absolutely a world where MSPs in different markets choose to share training data with each other in a structured and secure way. Instead of the ad hoc support you see today in community forums or Reddit threads, you could have an opt in model where the benefits of working together scale in a proactive manner.”

Rather than eroding the sense of community, AI may actually deepen it. “This industry is built on community, and I do not think AI diminishes that. In fact, I think AI amplifies it. The ability to share insights safely and securely across peer groups could make the whole ecosystem stronger.”

And the opportunity extends far beyond the PSA. “It is myopic to think only about ticket data in the PSA. Logs from SaaS tools, RMM platforms, and other integrations all contain valuable signals. Everyone is sitting on more data than they realize, and it is not limited to one system.”

Conclusion 

Ticket data has been hiding in plain sight for years. It sits quietly in PSAs, accumulating with every issue resolved and every technician note logged. But as Silverstone makes clear, that data is no longer just administrative overhead. It is raw material for smarter systems, more proactive service, and a flywheel effect that was not achievable prior.

Most MSPs won’t need to look far to get started. If you have six to twelve months of support tickets, you already have enough to train meaningful AI use cases. The real unlock is understanding what that data is worth, how to clean it up, and where to apply it. 

The good news is that this isn’t reserved for the biggest players. The path is open to any MSP that’s willing to rethink how they use what they already have. The question now isn’t whether your data has value. It’s whether you’re going to do anything with it.