AI Coding Upends The MSP Software Model: How MSPs Are Taking Control Of Their Own Tech Stacks

“We're not going out looking for a software vendor first anymore,” says Sandy McGrath, president of MIPGlobal. “The first question is, ‘What's possible with agents and automation?’ If we can't build it or it doesn't make sense, then we'll start looking for a software partner.”

**W**hen Sandy McGrath wanted to cut costs, he didn’t start renegotiating software contracts. He started canceling them.

Over the past six months, the president of Calgary, Alberta-based managed intelligence provider MIPGlobal has been sitting down with his leadership team, looking at the functionality of every tool in the software stack and asking if an in-house-built AI agent can do it instead.

In many cases, the team has found, the answer is yes. So far MIPGlobal has moved off one vendor’s tools and is now saving about $1,200 per month. Those savings, McGrath believes, could reach at least $5,000 per month by the end of the year as internally built agents replace other tools in the company’s software stack.

“We’re not going out looking for a software vendor first anymore,” he said. “The first question is, ‘What’s possible with agents and automation?’ If we can’t build it or it doesn’t make sense, then we’ll start looking for a software partner.”

The first tools he eliminated were ScalePad’s Lifecycle Manager and some reporting products.

“The AI replaces it by working directly with our systems,” he said. “It can do what we want immediately instead of making someone log into another platform.”

He also eliminated Microsoft Power BI licenses by connecting AI agents directly to the company’s operational systems.

Going forward, he’s approaching tools with a build-over-buy mentality.

And he’s not alone. Across the MSP industry, AI coding tools have sparked one of the biggest changes to MSP operations. With the advent of vibe coding—where people can describe in natural language the functionality they want software to have and AI can build it from there—MSPs are using GenAI to improve the functionality of their software stacks, fill in gaps in functionality or, in some cases, replace software they are paying for with tools they have built themselves.

As a result, several MSPs told CRN they are now scrutinizing the tools they use to run their business—including stand-alone billing, invoicing and documentation tools from MSP-focused vendors as well as MSP platforms from companies including ConnectWise, Halo, Kaseya and N-able—with an eye toward determining which to keep paying for and which can be improved or replaced.

Executives from the software vendors themselves told CRN they see opportunities for solution providers to use AI-generated tools to complement their own offerings but, unsurprisingly, cautioned against swapping them out altogether.

But there’s no question that the burst in vibe coding is changing how MSPs think about their software stacks.

For Kevin Damghani, CEO of Grand Rapids, Mich.- based ITPartners+, vibe coding has changed how his MSP buys software.

“We’re in a SaaS buying freeze,” said Damghani. “The first lens is, ‘Can we build this ourselves?’ before we do anything else.”

Using AI, his MSP built a CRM tool around its professional services automation (PSA) platform and project management applications around its own workflows.

“You can track upsells, renewals, account expansion, see when customers are growing in head count, analyze ticket sentiment with AI, identify churn risk from surveys and even process renewals automatically with account manager oversight,” he said. “That’s just the CRM feature, not even the project management side. What we’re doing there is AI-generated project buildouts from scopes of work, automatically integrating PDFs, analyzing project comments, risk registers and due dates, then providing a clear picture of what’s happening across the project. It’s mind-bending what can be done.”

He said the tools took about two weeks to build “and we’re expanding [the use of AI-built tools] across the company.”

And while he is saving on some costs, that wasn’t the original motivation.

“It started with saying we could do it better,” he said. “Then, once we did it better, it evolved into asking what else we could improve while saving costs at the same time.”

Antwine Jackson, who like Damghani and McGrath is using AI to build agents to improve his company’s software stack, said the majority of MSPs have been “at the mercy of SaaS vendors” for too long.

“If you wanted a feature or new functionality, you had to wait for them to innovate,” said Jackson, president and founder of Raleigh, N.C.-based Enitech. “In this new agentic era, we don’t have to wait anymore. We can innovate on our own, and the barrier to entry isn’t nearly what it used to be.”

He added that his MSP has replaced one vendor tool, declining to name the tool, and said he expects to jettison several others.

“It started as a functionality play,” he said. “Then we realized we’re probably going to save $15,000 to $20,000 a year by continuing down this path. The cost savings are really the byproduct. The functionality is what we’re after.”

Jackson currently has no plans to replace core operational platforms like PSA, remote monitoring and management (RMM) or cybersecurity tools with AI-built alternatives. But customer success platforms, business intelligence tools, reporting applications and other specialized products are all on the line.

“Those are the low-hanging fruit,” he said. “But we’re going to stay in our lane. If you’re wrong in security, the cost can be catastrophic. You’ve assumed the liability, and now you also have to maintain and secure the platform against things like supply chain attacks.”

‘We Built It To Fit Our Workflow’

Some MSPs are building AI agents to complement their existing tools or fill in feature gaps rather than outright replacing them.

Bill Blum, whose MSP has spent the past year building AI agents internally, said his team realized the biggest opportunity wasn’t ripping out mature software but extracting more value from it.

“We really haven’t replaced any tools. In fact, we’ve actually purchased more of them,” Blum, CEO of Bound Brook, N.J.-based Alpine Business Systems, told CRN. “What we’re doing is creating agents that interface with those tools, mine the data, automate workflows and give us information those platforms don’t easily provide on their own.”

One such agent reviews tickets and timesheets, another audits backup environments and others generate cybersecurity proposals, executive reports and Microsoft Teams meeting agendas automatically, he said.

“That takes 15 or 20 minutes off somebody’s desk every week,” he said. “Multiply that across all of our managers and meetings, and it’s incredible.”

Another MSP is taking a similar approach, arguing that AI agents are giving companies the ability to shape software around their businesses, not the other way around.

“The big problem with vendors, especially PSA [providers], is MSPs have to work their business based on the software,” said Bill Campbell, CEO of Waldorf, Md.-based Balancelogic. “That doesn’t work because every MSP has a thousand little workflow differences. You buy an expensive platform, and then you have to adjust your business and your processes to fit the software instead of the other way around. That’s backwards.”

So he spent the last year building AI applications around those differences. One project brings on-boarding, invoicing, ticket visibility, documentation and self-service into a single customer portal. Another automatically generates branded proposals and managed services agreements by pulling customer information from internal systems.

“What used to take us a week or even a week and a half now happens automatically,” he said. “I wouldn’t say we built it better; I’d say we built it to fit our workflow.”

For other MSPs, building their own AI tools is about eliminating the friction that comes from waiting for tool vendors to fulfill their feature requests.

As the CEO of Denver-based New Charter Technologies, Peter Melby’s AI agent strategy is to build whatever his MSP can’t buy. That philosophy drove the company’s acquisition of Orchestra AI in 2025. Melby said the goal wasn’t to compete with software vendors but rather to control the company’s own AI road map.

“It’s not because the vendors are falling short,” he said. “It’s a natural challenge with innovation. Software vendors can only build things they can effectively sell. If it isn’t commercially viable, they can’t justify building it. The result is they’re often limited to creating better versions of what we’re already doing today.

“We’re building systems not to replace vendors, but because vendors aren’t necessarily aligned to an entirely new way of thinking about future service models,” he added.

And even as AI lowers the barriers to software development, Melby expects many tools to eventually be overtaken by other offerings.

“There will always be reasons to build things that don’t exist yet,” he said. “Right now, the gap between what’s available in the market and what we can use to create a better customer experience is probably bigger than it will ever be again. That’s not because vendors are behind. It’s simply where we are in the innovation cycle.”

The allure of using GenAI to build your own tools is certainly one that Chris Day, CEO of ScalePad, the Vancouver, B.C.-based maker or the life-cycle management software removed by MIP Global, understands.

His team has gone through the same thought exercises, he said, at one point considering building internal replacements for major systems.

“We’d look at HubSpot and say, ‘Maybe we should just build our own CRM. We could strip out everything we don’t use and make it exactly how we want it,’” he said. “Then we realized we’d be signing ourselves up to maintain that forever.”

The complexity comes after the agent is built, as it then has to be maintained, managed and secured, he cautioned.

“You can build something in a few days that looks great,” Day told CRN. “People log in, they’re impressed and it feels like you’ve built a better version. Then reality starts showing up.

“Everyone sees the shiny part above the water,” he added. “What they don’t see is everything underneath: the audit logs, change history, integrations, permissions, notifications, APIs, reporting, scalability, all the things that actually make software work. Then feature requests start coming in. Then the bugs. Then somebody buys another security product and suddenly they want an integration. Every request becomes another project.”

Instead, he believes MSPs should focus on building on top of existing platforms.

“If your PSA doesn’t have the report you want, build the report,” he said. “If two products don’t integrate, connect them through their APIs.”

Other MSP software vendors also warned that building the first version of an AI agent is the easy part. The real challenge is maintaining it.

“MSPs have always been builders,” Manny Rivelo, CEO of Tampa, Fla.-based ConnectWise, told CRN in an email. “That is part of what makes this industry so resilient, entrepreneurial and fast-moving. But the question is not whether an MSP can build an AI agent for a specific task. The more important question is whether that agent has the right foundation behind it: trusted data, workflow context, security, governance, integrations, auditability and human oversight.”

Mike Adler, executive vice president, CTO and chief product officer of Burlington, Mass.-based N-able, said agents should sit on top of existing platforms rather than replace the infrastructure MSPs depend on.

“We’re the presence on the endpoint,” Adler told CRN. “We’re the platform that backs up the data, deploys software, monitors devices, runs scripts, detects threats and enforces security. MSPs should connect AI agents to those capabilities through APIs and MCP servers so they can automate their own processes.”

‘It’s Not About Removing The Human’

Many executives at other MSP vendors CRN spoke with supported the idea that MSPs should build on top of their platforms but warned of downsides for those who try to outright replace the platforms themselves.

Tim Barton-Wines, executive at Ipswich, England-based vendor Halo, said the industry’s future depends on making it easier for customers to use software in the ways they want.

“I spend a lot of time working with enterprise CTOs,” he said. “A lot of people just want the ability to yank data out of our platform, push it through their own proprietary models and calculations and then stuff it back into Halo to power things like stopping customer churn. Our attitude can’t be, ‘No, you can’t do that.’ That’s just not the mindset to have.”

Instead, he said vendors should embrace openness.

“Focus on your products. Make your APIs better. Fully document them and make them public,” he said. “We are an API company. It’s only going to become more important. Focus on the product first, and everything else comes later.”

That same message was echoed by Pratik Wadher, CTO of Miami-based Kaseya, who said that the next MSP software battleground won’t be whose AI model is best, but whose platform is easiest for partners build on and AI agents to operate across.

“I think having an open platform is the right way,” Wadher told CRN. “MSPs can contribute workflows that others can use, vendors can integrate into the platform and agents can interact through APIs and MCP servers. That’s where the industry is going.”

Rather than discouraging MSPs from experimenting with agents, he believes they should build enough to understand what AI can—and can’t—do.

“We generate roughly 180 million tickets a year and have billions of historical tickets,” he said. “You’re simply not going to beat that dataset. We can build higher-level AI functions that benefit everyone, while MSPs focus on the custom rules that make their businesses unique.”

Barton-Wines believes that the industry is in the first phase of AI, which he describes as when it’s added to just about everything to see what sticks.

“The first phase is basically, whatever the question is, the answer is AI,” the Halo executive said. “Everyone’s stuffing AI into everything and hoping for the best. The reality is it isn’t as good as people think it is.”

The next phase is making technicians dramatically more productive.

“It’s not about removing the human,” he said. “It’s about using AI for what it’s genuinely good at—pattern recognition—and keeping the human there to review and make the decision.”

The Hard Part Comes After The Prototype

While Matt Rose is using AI for internal operations, he’s concerned that some MSPs are mistaking AI-generated code for production-ready software.

“As much as vibe coding is cool, if you are a vibe-coding person that has a development background, that’s fantastic. But I think a lot of MSPs don’t have any idea what they’re really looking at when they look at code,” said Rose, co-founder and chief experience officer of Orlando, Fla.-based Tech Rage IT. “I like the idea of eventually maybe building a report or something that our PSA doesn’t have using AI. I think that’s a great use case.”

He also questioned whether MSPs’ estimates of cost savings are accounting for the price of development.

“It’s one thing to say, ‘I’m not paying a vendor,’” Rose said. “But are you really taking into account how much time did it take you to build it? If you’re fiddling around for eight hours a day, that’s eight hours a day that you could be selling.”

Karl Fulljames, CTO at Vancouver, B.C.-based Nucleus Networks, agreed. The CTO has built a customer-facing project management portal that integrates with software vendor ConnectWise, generates AI-powered project summaries and automates ticket routing.

“I get nervous when people say, ‘Hey, I’m saving $8,000 a month,’” he said. “That’s great, but who’s maintaining this? Who’s doing your security checks? Who’s validating that it’s actually secure? There are a whole lot of people jumping into this with no plan.”

From the vendor perspective, the opportunity is shifting from controlling every workflow to becoming the foundation MSPs build on.

Halo’s Barton-Wines said open APIs will become more and more important as MSPs create their own AI-driven workflows. And in the future, MSPs will both build and buy tools, “I think it’ll absolutely be a combination,” he said.

ScalePad’s Day said he’s not fearful of the company’s software being replaced. He simply believes AI just raises expectations for vendors. “It’s a threat to software vendors that don’t innovate,” he said.

Going forward, he expects many MSPs will eventually rethink their approach around what they’re building agents for.

“I think everyone should be building,” he said. “Just don’t build commodities. If you’re sitting in your office rebuilding a ticketing system, another MSP is already in front of your customer helping them with other opportunities.”

Michael Cervino,co-founder and CEO of Radnor, Pa.-based Circle Square Consulting, agreed that MSPs should focus on “improving broken processes, connecting systems and creating capabilities that vendors do not provide.”

“I’m not going to build my own [governance, risk and compliance] platform. There’s no ROI there,” Cervino told CRN. “But if there’s a process that’s clearly broken or something that doesn’t exist, that’s where AI becomes incredibly valuable.”

Paul Vedder, co-founder and CEO of West Palm Beach, Fla.-based VXIT, agreed that MSPs should experiment with AI but cautioned against trying to replace their entire technology stack.

“We aren’t going to replace everything,” Vedder told CRN. “We’ll replace some low-level tools. But the complicated tools? We’re going to let the professionals do those.”

Instead, Vedder’s MSP is using AI to increase employee capacity.

“We’re building augmentation to our current team,” he said. “We’re not replacing people. We’re not getting rid of our staff. We’re just going to be able to add more capacity.”

He also hired a full-time employee to explore AI opportunities.

“I think in 18 to 24 months you’re going to look at an org chart and you’ll have virtual employees that are AI agents doing work,” he said. “But they still need to be maintained. They need to be managed.”

Agents Can Lead To Competitive Advantage

For MSPs, the biggest agentic AI opportunity is not just about cutting costs but how they can build, deliver and scale.

“If you’re implementing AI just to save money, you’re doing it for the wrong reason,” Balancelogic’s Campbell said. “Sure, you might eliminate a monthly software subscription, but now all the support falls on you. Are you really saving money? We’re implementing AI to give us additional capabilities. That leads to additional revenue, greater efficiency and more productivity. The cost savings come later.”

And as AI continues to change the relationship between MSPs and vendors, many MSPs are learning what to build, what to buy and where to draw the line.

“If a pain point comes up in a management meeting, we don’t have to wait six months for a vendor road map anymore,” MIPGlobal’s McGrath said. “We can assign somebody internally to build it. That ability to pivot quickly can save me thousands of dollars a year while giving me new functionality exactly when I need it.”

That flexibility, he added, may give him a bigger competitive advantage.

“This is a fundamental shift,” he said. “These agents are staff augmentations. One employee might manage four, five or 10 agents and scale themselves tenfold. That’s what’s going to allow companies like ours to keep doubling year over year without hiring at the rate we used to. It’s just not financially viable to keep scaling the old way anymore.”