September 30, 2026
In a piece published by Fortune on September 26, 2026, Bryan Goode -- Microsoft's Corporate VP of Business Applications and Agents, who leads product marketing for Dynamics 365, Power Platform, and Copilot Studio -- made the case that SaaS isn't being killed by AI agents. It's being upgraded by them. According to Fortune, Goode's argument goes like this: every SaaS app a business buys today -- a CRM, an ERP, a project management tool, a productivity suite -- was designed for a human to operate, which means a person has always had to be the connective tissue between them. Someone copies a deal from the CRM into a spreadsheet, re-keys a customer request into the support tool, chases an approval across email and a project board. Goode calls the toll that takes on people's time and attention the 'toggle tax,' per Fortune's reporting. His forecast is that AI agents take over that connecting work -- not the applications themselves. The apps stay. The tabbing between them disappears. People, in his framing, stop being 'human glue' and spend their time on decisions, relationships, and running the business instead. Fortune also cites a Gartner forecast in support of the trend: agentic AI will be embedded in 33% of enterprise software applications by 2028, up from under 1% in 2024.
If you're currently deciding whether to add another SaaS seat or consolidate, this argument matters because it's the most credible version of the case for staying the course. It doesn't deny that AI is about to change how work gets done -- it just insists the change happens on top of your existing stack, not underneath it. That's a comfortable answer for anyone who has already signed multi-year contracts with a CRM vendor, an ERP vendor, and a project management vendor, because it means the money keeps flowing to the same five or six line items, just with an agent layer bolted on. It's also, not coincidentally, the answer that happens to be best for Microsoft, whose Copilot Studio and Power Platform are built to be that agent layer sitting across other people's software.
Goode's framing is genuinely useful -- the toggle tax is a real, describable cost, and naming it is smart. But look closely at what the argument assumes rather than proves: that the five or six licensed applications a team pays for today are the fixed, permanent home for that workflow, and the only open question is who, or what, does the connecting between them. That's a narrower claim than it sounds like. It quietly reduces the entire disruption story to 'agents replace the person who was tabbing between apps' -- and skips past a much less comfortable question: why did this workflow need five separate applications in the first place?
For a lot of mid-size teams, the honest answer is that it didn't need five apps -- it needed five apps because a person had to be the integration layer, and off-the-shelf software gets sold, priced, and marketed one department at a time. A sales team buys a CRM. Ops buys a project tool. Support buys a helpdesk. Finance buys something else. Each purchase made sense in isolation. The stitching-together problem Goode describes is downstream of that fragmentation, not separate from it. Once an AI agent -- or better, one purpose-built application generated for the exact workflow a team actually runs -- can perform the underlying task directly, instead of just relaying data between five systems built for someone else's org chart, the case for licensing and administering all five platforms gets weaker. Not just the case for manually toggling between them.
This is the same tension we've written about in the context of buy vs. build vs. ViibeStack: the real cost of a stacked SaaS bill isn't only the per-seat price, it's the ongoing coordination overhead of keeping several broad platforms in sync for one narrow, specific workflow. Agents that sit on top of that stack can reduce the coordination pain. They don't reduce the number of vendors you're paying, the number of admins you're managing, or the surface area you're responsible for securing -- something worth weighing against the practical steps in our security and compliance checklist before you add another layer to an already-tall stack.
None of this makes Goode wrong on the facts he's describing. But it's worth saying plainly: he's a Microsoft VP whose own product portfolio is the direct beneficiary of a future where SaaS stays intact and Microsoft's agent layer sits on top of it. A world where agents dissolve the need for five separate licensed platforms is a much worse outcome for Dynamics 365, Power Platform, and Copilot Studio than a world where those platforms remain the substrate and Microsoft just gets to sell the connective agent that plugs into all of them. That doesn't invalidate the toggle-tax framing -- it's a real cost, and it's smart to name it -- but it does explain why the story stops exactly at the point where it stops threatening the licenses Goode's team sells.
The toggle tax is a good concept. We just think Goode understates who ends up paying less of it. Plenty of teams are going to keep their CRM, ERP, and project tool and add an agent on top -- that's a reasonable, low-disruption path, and it's exactly what products like Copilot Studio are built for. But a growing number of teams are going to look at that same toggle tax and decide the cheaper fix isn't a smarter agent stitching five platforms together -- it's not needing five platforms for one workflow at all. That's the logic behind building a narrow internal tool purpose-fit to the actual process, instead of paying for and administering several broad platforms that were never built to talk to each other in the first place. We've seen teams make exactly that trade when they replace a general project tool or consolidate a CRM and support workflow into one app instead of coordinating across two. The toggle tax doesn't just get automated away by an agent -- it gets avoided entirely by teams who stop paying for the toggle in the first place.
Sources