Was this forwarded to you by a friend? Sign up, and get your own copy of the news that matters sent to your inbox every week. Sign up for the On EdTech newsletter. Interested in additional analysis? Upgrade to the On EdTech+ newsletter.

❝

The more important ‘S’ in software as a service now is service, service for the software.

In a recent board presentation I gave for Pathify, CEO Chase Williams offered this pithy summary of the both-and argument I was making about AI in EdTech. It is both getting easier for a small team to build software to address a specific campus problem, and it is getting more difficult to keep that software secure and dependable over the years of that application’s use.

For software as a service (SaaS) vendors in that argument: upkeep, security, privacy, and implementation support account for more of what a college or university values in its annual subscription. But there is also an implication for the higher ed customers: if AI allows a vendor to build and fix software faster, the customer has reason to expect a better product for the money it already spends.

Both points apply to home-grown software in EdTech. Build versus buy answers one question; who will provide the continuing service, and what should people expect from that team, is a separate decision.

Two Interrelated Decisions

Brian Robinson, senior vice president of product at Watermark, came at the issue from the campus side. Responding on LinkedIn to my October post on the PHP2Shell cybersecurity vulnerability, Brian acknowledged his stake as someone who sells software. He also made the case that home-grown software can be the right call for prototypes and small local needs. As he put it, “Creating the tool is usually in the budget.” The work of monitoring it, patching it, and responding when something goes wrong arrives later and lasts as long as the tool runs.

Consider a lab coordinator who uses AI to combine the course schedule and two spreadsheets into a list of open placements. Once advisors start placing students from that list, they expect someone to catch a change in how the registrar codes a full section. From the coordinator’s perspective, the project worked. From the advisors’ perspective, the institution now provides a service.

The room-booking example at McGill from the October post illustrates how an institution can handle both sides of this work. A neuroscience center was using an old application on a centrally maintained host. When I checked, the version of PHP underneath the application had already been patched for the PHP2Shell vulnerability, showing the different organizational levels involved.

  • At the department level, the center could keep a familiar application that met its scheduling needs. The people using it were in a good position to judge whether it still did the job.

  • At the university level, a team responsible for the host could update the underlying software. The center’s ability to keep booking rooms did not depend on someone in the department following PHP security advisories.

AI makes that division of labor more explicit and more important. An institution may find it worthwhile for more departments to build or adapt tools while simultaneously deciding that fewer departments should operate their own public-facing systems. Shared hosting takes some work off the department’s hands, but it still takes someone who understands lab placements to recognize a bad result, correct it, and tell the advisors who relied on it.

Service Provider Expectations

There is a temptation to read this as a budget argument. Add a maintenance line to the campus project, or pay the vendor enough to cover the extra security work. The problem is that this reading leaves out the change in expectations that both Brian and Chase described.

Vendors can use AI themselves to find and address weaknesses, and that changes what a customer can reasonably ask between security reviews. There may still be annual or semi-annual audits of system software, but in the new cybersecurity age we’re in, there needs to be a more proactive and ongoing examination of code, with much quicker identification of problems, software remediation, and communication with customers on what they might need to do.

The same question applies to a campus IT group running applications for a dozen departments. In either case, the people depending on those applications have a stake in how the operator uses its new capabilities.

Chase put the commercial pressure more directly.

❝

“If I’m paying you $100,000 for this every year, in addition to the services you’re offering, I expect it to get a lot better, quicker.”

End users are increasing their expectation of the services provided, even at the same price. They expect software to get a lot better, quicker. For an advising office, that might mean a recurring integration problem finally gets fixed in time for fall registration. The service element of SaaS is getting more important.

What the Institution Still Owns

One point that came up in a discussion with university leaders this week was how much judgment a campus still has to exercise when it buys software. The Canvas breach this spring provides a public example—when Instructure said Canvas was ready, many institutions waited days before restoring access. Simply bringing the platform back online was necessary but insufficient, and the real challenge was evaluating when faculty and students could safely use the LMS again, and institutions needed reliable information from the company to make that call.

This is where a vendor’s culture and infrastructure become more important in the value of the subscription. Do vendor executives and managers understand the decision a provost has to make, and can that information be shared while there is time to act? The equivalent question for a campus-built application is whether the department relying on it can get that answer from its own institution.

I expect AI to expand the range of software colleges and universities build locally and to increase demand for professionally managed services at the same time. Some of that work will go to vendors and some to campus teams. The change to watch is what counts as good service in either setting. Keeping an application available is part of the equation, but so is using the tools now available to find weaknesses earlier, resolve problems faster, and give the people who depend on the software enough information to make their own decisions.

The main On EdTech newsletter is free to share in part or in whole. All we ask is attribution.

Thanks for being a subscriber.