Skip to content

Is Locking Down AI Enough for MSP Clients?

Locking down AI is not enough, because it protects your MSP's clients without making their businesses any better, and Gary Boyle argues the MSP that helps clients build with AI is the one that wins. In Gary's account, a small MSP owner who isn't a developer rebuilt a client's broken legacy app in about five hours of supervised work with an AI coding tool. The governance you already run for file sharing and email covers most of the risk. For more context, see how artificial intelligence is transforming MSP software for IT support and how to prepare your MSP for the next wave of technological change.

In this BMK Vision Roundtable, Josh Peterson, CEO of Bering McKinley, and Gary Boyle, COO and Management Consultant, talk through sharing AI-built apps with clients safely, borrowing the governance you already have, and when an MSP should build instead of block.

Is Locking Down AI Enough for MSP Clients?

No, because clients don't feel the value of being locked down. Gary said most of what he hears from MSPs is restriction: govern AI, make sure nobody uses it who shouldn't, and keep unwanted data out of it. He agrees with all of that, but he struggles to see how stopping there benefits any client. When he ran his own MSP, he leaned toward lockdown too, because he didn't want the call about leaked data or a breach.

The catch is that security work turns invisible when it succeeds. Clients stop getting hit with ransomware and assume that is simply how a business runs, while they do feel the blocked websites and the heavy-handed approvals for new devices. In Gary's view, the MSP that asks how it can generate more leads, more revenue, or lower costs for a client, and then uses AI to deliver it, is the MSP that wins.

  • Treat AI governance as the floor of your offer, not the whole offer.
  • Remember that security work goes unnoticed when it succeeds.
  • Ask each client how AI could raise their leads, revenue, or margin.

Is It Safe to Share an AI-Built App With a Client?

It can be, if you share it by invitation and never by public link. The episode opens with a small demo app that held a client's real prospect list and had been shared through a public link, and Gary's advice was to stop. A public link puts the page on the internet for anyone who has the URL. Gary recalled that when one AI platform first let users share chats publicly, search engines indexed those links and information leaked that nobody realized was exposed.

Invite-only sharing is different. The recipient is invited by email, has to sign in with an account tied to that address, and gets an editor, commenter, or viewer role. Gary compared it to sharing a spreadsheet from your file storage to one address: you are trusting the platform's back end. He also flagged two checks. An invitation fails when the address isn't tied to the person's account, and a published page may keep showing an old version after you edit it.

  • Never put client data behind a public share link.
  • Invite people at the exact address tied to their account.
  • Check what recipients actually see after every edit.

Which Sharing Rules Should an MSP Apply to AI Tools?

The same ones you already enforce for file sharing, chat, and email. Gary pointed out that an MSP can stop employees from sharing documents with people outside the organization through a single policy setting in its collaboration platform, and it can watch outbound email for sensitive information. An AI tool's share button has none of that around it, which is why MSPs cringe: a user can load a data set into a model and then publish an app built on that data through a link the MSP never sees.

Josh's takeaway was that sharing is sharing, and the governance around it can be borrowed from non-AI applications. Gary added a step that comes first. Before a client's data goes into an AI tool at all, check with the client. He separates a firm's own internal data, where the productivity gain may be worth the risk, from client data, where the client decides.

  • Get the client's consent before their data enters an AI tool.
  • Extend your external sharing policies to AI platforms.
  • Use team or enterprise plans so users sit inside a managed account.

Should an MSP Rebuild a Client's Legacy App With AI?

Often yes, if the app keeps generating tickets and you are honest about what you are. Gary described a smaller, AI-forward MSP owner who kept getting tickets about a client's old custom scheduling app. He told the client's CEO he could probably rebuild it, and the CEO agreed the app was old, possibly insecure, and never did what the business wanted. Over a weekend, the owner gave an AI coding tool a recording of his conversation with the client, screenshots, and the old code.

In about five hours of supervised work, he had a new app with the missing features that signs users in with their existing company credentials. He is not a developer. Josh's caveat is candor: tell the client you are exploring the build together, not posing as an expert, and that fixes will be billed. Gary's counter is that custom developers were never uniformly good, and old code written by rotating teams is often insecure and inconsistent.

  • Start with the legacy app that keeps generating tickets.
  • Scope it as a defined project and say who maintains it.
  • Tell the client plainly that you are exploring the build with them.

Should Your MSP Take Over Client Websites?

Gary thinks AI tools now make it a real option, because the MSP already owns the fallout. Most MSPs say they don't do websites, so an outside contractor builds one. When that site gets hacked and starts sending spam, and the client's email system goes down with it, the MSP still takes the call, whatever the contract says. In the past, an MSP had little ability to do anything about that.

Now, Gary argues, it can own the site, secure it, and connect it to the client's scheduling and calendar so it adds real value to their workflow. His comparison is direct: a website that brings a client more leads every month is worth more to them than a web filter, and you don't have to choose between the two. Josh closed with advice for every owner: start by building something for yourself, because you can't see what these tools do until you use them.

  • Count the website risks you already absorb without any control.
  • Offer websites as a lead and security service, not a side job.
  • Build something for your own MSP before you offer it to clients.

Frequently Asked Questions

Is locking down AI enough for MSP clients?

Not on its own, according to Gary Boyle. Governing AI, limiting who uses it, and keeping unwanted data out are all necessary, but clients don't feel the value of a locked-down environment. Gary argues that the MSP that also asks how AI can bring a client more leads, more revenue, or lower costs is the one clients will value.

Is it safe to share an AI-built app with a public link?

Not if it holds client data. A public link lets anyone with the URL open the page. Gary Boyle recalled that when one AI platform first allowed public chat sharing, search engines indexed those links and exposed information nobody meant to publish. Treat a public link to client data the same as posting the client's financial files online.

How does invite-only sharing protect client data in AI tools?

The owner invites specific people by email, each person has to sign in with an account tied to that address, and the owner assigns an editor, commenter, or viewer role. Gary Boyle compares it to sharing a file to one address from your file storage. Confirm the address is right and that recipients see the current version.

Should an MSP get client permission before using client data in AI?

Yes. On the episode, Gary Boyle separates a firm's own internal data, where the productivity gain may justify some risk, from client data, where the client has to agree first. Once the client says yes, the same external sharing rules you already apply to files, chat, and email should govern what leaves the AI tool.

Can an MSP build software for clients without being a developer?

Gary Boyle described a small MSP owner, not a developer, who rebuilt a client's legacy scheduling app with an AI coding tool in about five hours of supervised work. Josh Peterson's advice is to be candid: tell the client you are exploring the build together, that you are not a software house, and how fixes will be handled.

Why should an MSP try AI tools again if it gave up on them?

Gary Boyle says the tools owners gave up on in 2025, the ones that hallucinated or got math wrong, are not what they would be using today. He and Josh Peterson agree that the only way to see what is possible is to build something for yourself first, then guide clients through the same process.


Listen on Your Favorite Platform

  • Apple Podcasts
  • YouTube
  • YouTube Music
  • Spotify
  • Amazon Music
  • Podbean
  • iHeartRadio
  • Player FM
  • Listen Notes
  • Podchaser
  • BoomPlay

Episode Highlights

  • Why a public share link holding client data is a leak.
  • Invite-only sharing and roles for AI-built apps.
  • Borrowing file sharing and email governance for AI tools.
  • Asking clients before their data goes into an AI tool.
  • Why clients don't feel the value of a locked-down environment.
  • The shift toward advisement and what AI adds to it.
  • Why today's AI tools aren't the ones owners gave up on.
  • Rebuilding a legacy scheduling app in about five hours.
  • Trusting AI-written code against the average developer.
  • Owning client websites instead of leaving them to contractors.

About the Co-Host: Gary Boyle

Gary Boyle is COO and Management Consultant at Bering McKinley. He ran his own MSP before joining BMK and works with MSP leadership teams on operations, planning, and execution.

About the Host: Josh Peterson

Josh Peterson is the CEO of Bering McKinley and host of The BMK Vision Podcast. Since 2004, Josh has worked with hundreds of MSP owners to build operationally sound, profitable businesses through consulting, peer teams, and direct coaching.

Connect with Josh Peterson on LinkedIn →

Related Resources from Bering McKinley

Want to Continue the Conversation?

Whether your MSP decides to build with AI or only govern it, that choice belongs in your strategic plan. Vision gives your MSP a strategic plan, a consultant who holds your team accountable every month, and your numbers tracked against BMK's standards.