We’ve got an AI debt problem, people.
And no, I’m not talking about financial debt, though there really is an awful lot of data center borrowing going on out there. And I don’t mean tech debt, though AI-generated spaghetti code is accumulating at a somewhat worrying pace too.
What I’m talking about instead is something far less discussed than it should be, especially by MSPs: governance debt, the kind that builds up when outdated, erratically enforced policies that may not have been terribly thorough to begin with encounter latest-generation AI productivity tools.
Like Microsoft Copilot, which is worth calling out not because Microsoft, Microsoft partners, or Microsoft users are unusually sloppy about governance, but because there are a lot of people using Copilot and more joining them rapidly. Indeed, paid Microsoft 365 Copilot seats jumped more than 100% quarter-over-quarter in Microsoft’s fiscal Q4 to north of 30 million.
Which is a problem for a lot of the companies using those seats based on what John Donovan (pictured), a principal research director in the infrastructure and operations practice at Info-Tech Research Group, has seen during some five years of advising Microsoft 365 subscribers. Many such companies rely on templated access and data sharing configurations from Microsoft that they haven’t updated in years, never applied consistently across apps, and may or may not have tuned to reflect compliance requirements.
“AI just really amplifies that existing governance weakness,” Donovan says.
The toll when the debt created by that weakness comes due can include PII exposure, IP leakage, and the morale hit that inevitably follows everyone in an organization discovering how much everyone else is making. Inspecting and refining governance settings before installing Copilot makes all of that largely avoidable, but many businesses—and MSPs—skip right past that step.
“They kick it off with a pilot, they don’t really know what they’re doing, and they start exposing data really quickly,” Donovan says. Mostly to avoid the better alternative, which takes time and effort, as Info-Tech itself discovered when it adopted Copilot over two years ago.
“We spent at least two or three months working on the security aspect of our tenant before we even released Copilot to our systems,” Donovan says.
The thing about a multi-month governance and security project, though, is that it’s billable. It also gets an MSP talking with clients about topics like data classification and cleansing that few SMBs have paid much mind to in the past, Donovan notes.
“It exposes that data debt, so to speak,” he says.
Super. Now we’re paying down two hidden debts, and given that IDC expects data from M365 and other productivity tools to amass at a 46.7% CAGR through 2030, the tab is compounding rapidly.
AI, it’s worth noting, is both responsible for that data growth and the biggest reason coping with data debt matters so much. People who feed Copilot redundant, messy records from 20 years ago, Donovan notes, will probably be unhappy with its output.
“They’ll be getting results based on inaccurate information, which could make or break a company,” he says.
Which is why a client’s senior-most leaders should always be part of governance engagements, and why addressing governance can help MSPs cultivate the kind of consultative relationships they badly need at a time when managed BPO is rapidly displacing managed IT as the service end users want and need.
“It’s really a business problem,” Donovan says. “It’s not an IT problem.” As a result, he adds, it should steer toward a “business-aligned governance” outcome that reflects the customer’s culture, goals, and regulatory obligations.
“For example, if I’m trying to put in a data retention policy, I need to understand what the business requirements are for that data,” Donovan says. Industry nuances can be crucial as well.
“If I’m a legal firm, I have to treat my data a whole lot different than if I’m just a commercial candy firm,” Donovan observes. Needs evolve over time too, he adds, so MSPs should revisit them with clients regularly.
“You need to be looking at your governance annually,” Donovan says.
Any MSP can master the technical parts of that conversation pretty quickly, he adds. “Microsoft has plenty of stuff out there for you.”
The harder part for MSPs still trapped in a speeds-and-feeds mindset, I suspect, will be learning how to augment the technical parts of the conversation with the business alignment parts.
Speaking of managed BPO…
Remember last week’s discussion of that topic here? I got the term straight from Top Down Ventures managing partner Mark Scott during the interview that he and fellow managing partner Joel Abramson did with us for the latest episode of MSP Chat, the podcast I co-host. Tune in to the entire interview to hear their informed, interesting thoughts on investing in AI and managed services. Then go here to explore our many other equally enlightening interviews with industry thought leaders.
Copulo’s plan for putting voice traffic to work for AI
Technicians are notoriously bad at documentation.
Some have to be bullied into doing it at all. Others do little more than jot down a few quick notes after closing a ticket. Not because they’re lazy in most cases, mind you, but simply because capturing documentation is time-consuming, tedious work that feels like a distraction from more urgent tasks. The upshot, though, is that they end up squandering data their employers badly need to deliver faster, more efficient service and train the AI-based automation tools more and more of them are adopting.
They’re not the only ones guilty of that sin either.
“Say you’re a salesperson and we’re talking, and we’re on the phone for twenty minutes. We hang up, and now you have to put notes into your CRM,” says Sam Skarloff. Except that takes time and patience you don’t have. “The notes end up being, ‘Talked to Sam. Great conversation. Have to follow up next Wednesday at 11.’”
Which is better than nothing, but only slightly, notes Sklaroff, who is chief strategy officer of Copulo, a recently launched AI startup whose first product, called SyncConnect, is designed to do the post-call work salespeople should do themselves but don’t, along with some things they probably couldn’t do terribly well even if they tried.
“We’re going to go into that contact, and we’re going to put in a link to the voice recording with sentiment analysis, we’re going to put in a link to the transcript, and we’re actually going to take the summary and put that right in so that you don’t have to take any notes,” Sklaroff says.
In doing so, he adds, SyncConnect augments the data in CRM, ERP, and other systems of record that AI tools depend on with additional fodder from a less widely utilized source. “There’s a lot of data in phone calls that people aren’t using.”
Sklaroff has a long history with phone calls, too. “I’ve been in the telephone business for forty years,” he says. He’s still there too: in addition to his role at Copulo (Latin for “to connect”) Sklaroff is CEO of telco provider Axxess Networks. He was also, some two decades ago, an early-generation managed service provider, which is why he and Copulo have MSPs squarely in their TAM sights. SyncConnect is great for sales reps, Sklaroff says. “It’s also great for help desk.”
With both audiences in mind, SyncConnect currently integrates with Salesforce, HubSpot, ConnectWise PSA, Zendesk, and roughly 20 other PSA, CRM, and ERP solutions. Within about six months, Sklaroff says, the list of supported platforms will number in the “hundreds”.
Other enhancements on the roadmap include calendaring functionality that will offer to schedule future meetings referenced during calls automatically and an AI assistant that will help users query their notes for facts and insights. A separate product currently in development called CallerIQ will enrich inbound calls seconds before the phone rings with information captured earlier by SyncConnect.
“If you’re a sales guy, we might pop up a summary of the last call you had with that prospect,” Sklaroff says. “If you’re on the help desk, we might pop up an open ticket or the last ticket they called in about.”
One can imagine a couple of things that could trip up a product strategy like this, beginning with the fact that there’s nothing to stop well-known UCaaS names like 8x8 and RingCentral from offering similar functionality. And indeed, Sklaroff assumes they all will.
“But they’re going to do it just on their platform,” he says. “We’re going to be carrier-agnostic, so you don’t have to throw out the unified communications platform that you are using now.”
Assuming you’re using a unified communications platform, that is, which raises issue two: plenty of MSPs don’t offer voice services at present and don’t wish to start. Copulo, working with Axxess Networks, has an answer here too.
“What we’ll do with our channel partners is take their customer list, whether they have 50 or 200 or 500, and we’ll run a campaign to both email and call their customers under their brand,” Sklaroff says. MSPs can pursue any leads that result themselves or outsource that as well.
“We’ll do the quote, we’ll do the demo, and we’ll close the sale for them,” Sklaroff says, adding that Axxess will pay the MSP commissions on that sale for as long as the customer’s account remains active.
Copulo, which is still pre-release on SyncConnect, has yet to finalize pricing for the service, but is currently planning to charge $10 per user per month plus 10 cents every time the solution writes call data out to a CRM or PSA solution. Alternatively, subscribers can purchase a plan with unlimited writes for $29.99 per user per month. If you’re in sales and doing 15 calls or more a day, Sklaroff says, the “all-you-can-eat” option usually makes good sense.
“And if you’re not as busy as 15 calls a day, you probably shouldn’t have a job,” he says with a grin.
Clairify builds a chief of staff for the rest of us
As we’ve noted here before, you can make good money selling software to MSPs but great money selling software through MSPs to their customers.
It’s a lesson the founders of Clairify have taken to heart. When I first met with them, on the sidelines of Pax8’s 2024 Beyond conference in Denver, they were hard at work on a product that would help busy, mobile MSPs organize and triage the information flying into their phone between meetings from ticketing systems, emails, texts, and other sources. Then they realized MSPs aren’t the only busy, mobile leaders inundated with tasks and information.
“The problems are always the same,” says Mehul Patel (pictured), a Clairify co-founder and its head of product. “When you have a small business, you’re typically surviving on the heroic effort of the owner.”
It’s apt that I first met with Patel and co-founder Randy Spangler, an ex-MSP, at Beyond given that multiple Pax8 executives at this year’s edition of that show endorsed building a digital chief of staff as a great first step toward mastering agentic AI. Clairify’s solution is designed to spare us all that work.
“We’ve taken that basic ChatGPT AI system model, and we’ve elevated it to a chief of staff role,” Patel says.
Specifically for leaders, the people who most need a chief of staff. “We’re really talking about protecting the user’s attention and making sure that knowledge is available and distributed throughout the organization, the typical things that a chief of staff would do if it was human,” Patel says.
Theoretically, you can accomplish that with Microsoft 365 Copilot, but Patel, Spangler, and a lot of SMB end users prefer the results they get from ChatGPT, Claude, and other frontier chatbots. Yet training those takes time and (as we’ve just seen) governance does too. Plus, recent headlines have underscored the security risks frontier agents can pose.
“It’s a ticking bomb if you don’t do it right,” Patel says. “An MSP’s not going to sell you cybersecurity and then deploy an unsecured agent in your workspace.”
Furthermore, he adds, ChatGPT and its peers are all-purpose tools with (as Lexful’s CEO would probably agree) too much context to perform a targeted set of functions effectively.
“They’re not fundamentally designed from a user experience perspective or from an information architecture perspective to serve executives,” Patel says. “[Clairify] is tuned to the executive level, and it’s a very specific corpus of data that we’re developing for a very specific purpose. Because of that, we think that we can produce higher fidelity output.”
While drawing on frontier AI’s inference power. Clairify automatically connects via natural language prompts to any LLM you use (including local ones), not to mention any SaaS solution in an end user’s stack with an API. It then consolidates data from those systems on the fly to process queries spanning multiple apps.
“You can get answers to just about anything and let the AI do the joining,” Spangler says. “It’s a lot less hassle for the backend and the user.”
It also, Patel adds, gives MSPs without AI expertise something compelling to offer customers immediately. “You get an AI that knows your business, connects to your tools, and is secured from day one,” he says. “And if you ever need help, just call support like you would if the printer jammed.”
Clairify is available for iOS at a launch-priced $12 per user per month, with additional subscription options coming soon. An Android version of the system is on the way as well.
Over on The Business of Tech
Host Dave Sobel has a theory that the reason end user executives are gravitating toward outcome-based pricing on AI services is because the AI services they purchased in the past have yet to deliver ROI:
“Three years ago they stood up in front of a board, or a staff meeting, or an earnings call, and committed to artificial intelligence. Budget moved. Headcount moved. In a lot of cases people were let go and the reason given was this technology. That commitment is on the record, and it cannot be walked back. And they cannot demonstrate the return … So the buyer changes what they are purchasing. Not the work — the certainty. If I cannot prove this paid off, I will stop paying until somebody else proves it. That is the whole appeal of paying on outcomes, and wanting it is completely rational.”
Also worth noting
Integris, which figured prominently in last week’s post, has completed its acquisition of First Focus, the mega mega MSP’s first overseas expansion.
Lots of news from CrowdStrike’s Fal.Con show last week, including the launches of SafeMind, a family of purpose-built cybersecurity AI models and harnesses; Falcon Guardian, an AI detection and response solution for agentic AI; and new integrations with Google Cloud’s enterprise AI ecosystem.
Also at Fal.Con, Rubrik said it’s integrating its identity-security and recovery technologies through CrowdStrike’s Charlotte Agentic SOAR.
Sophos is employing OpenAI GPT cyber models to help Sophos Managed Risk users better prioritize and manage exploitable vulnerabilities.
Proofpoint’s new SOC Analyst Agent, similarly, uses OpenAI Daybreak models to turn natural-language security queries into traceable findings and recommended next steps.
Palo Alto Networks has acquired AI-native platform provider Console to deepen agentic capabilities in its Cortex platform.
Cato Networks has launched SMB FlexPool, a pooled-capacity licensing program that lets MSPs provision and reallocate SASE licenses across SMB customers more easily.
CyberFOX is the newest addition to Ingram Micro’s line card.
MSP360 has added client-side encryption and expanded Object Lock functionality to the free edition of its cloud backup solution.
Cycode’s new Agentic Code Scanning solution uses deterministic and agent-based code analysis to uncover hidden vulnerabilities in AI-generated programs.
That hefty post-CompTIA endowment at work: GTIA has donated more than $3.2 million to 73 nonprofits worldwide, including $700,000 distributed to 19 organizations selected directly by GTIA members.
Kurt Mills is the new channel chief at KnowBe4.
Tom Bonos is the new CRO at Sumo Logic.
Jessica Krowel is the new SVP of North American private sector sales at Keeper Security and Bill Grabner is her new public sector counterpart.