The single most important shift in business software isn't a new app — it's the death of the app as we knew it. For a decade and a half, the model was fixed: identify a task, buy a tool for it, learn its interface, pay per seat, repeat. Do that enough times and you end up where the average enterprise sits today — juggling 291 separate SaaS subscriptions, up from 110 in 2020. Small businesses aren't far behind. Every one of those tools makes you do the work inside it.

An AI agent flips that. Instead of handing you a form to fill out, it takes the goal, breaks it into steps, calls the tools itself, and comes back with the result. One is a spreadsheet. The other is a co-worker who never sleeps. That's why Gartner expects task-specific agents inside 40% of enterprise apps by 2026 and why analysts project roughly 35% of point-product SaaS tools will be replaced or absorbed into agent ecosystems by 2030. The two-track economy Michael Dermer keeps naming is showing up in software too — the tools that give you leverage, and the ones that just rent you an interface.

Seat-based pricing charges you for headcount. Agent economics charge you for outcomes. Once you feel that difference, you can't unfeel it.
— Michael Dermer

Read one way, that's a threat to a stack you've spent years assembling. Read another way, it's the biggest cost-and-leverage opportunity to hit small business in a decade. The question isn't whether to adopt agents — it's which tools to keep, which to pilot, and which to let go.

Why "just add another tool" is the wrong lever — and what the data proves

Here's the trap the heat-grid exposes. The default response to a new business need has always been linear: new problem, buy new software. That's how a company ends up with 291 apps, a tangle of subscriptions that don't talk to each other, and a monthly bill that scales with every hire. Each tool solves one slice and leaves the stitching to a human — usually you. The cost isn't just the subscriptions; it's the hours spent moving data between tools that were never designed to cooperate.

Agents attack exactly that seam. Instead of a tenth tool, one agent can read from your existing systems, do the multi-step work a human used to do by hand, and write the result back — no new interface to learn, no per-seat tax that grows with your team. This is the leverage principle rendered in software: a SaaS seat is worth exactly one seat, every time, and its cost climbs with headcount. An agent that automates a whole workflow keeps paying off as you scale, because its cost is roughly flat. Same problem, radically different economics.

Adding another app is not a strategy — it's a subscription you'll forget you're paying. The move is to make one workflow run itself, not to buy a tenth tool to babysit.
— Michael Dermer

So how does the money actually shake out when you put a per-seat tool next to a single agent doing the same job? Slide the numbers yourself.

Why the "boring, ownable" workflow beats the shiny new tool

Run the calculator up to a mid-size team and a quiet inversion appears: the per-seat tool that felt cheap at three people becomes the most expensive line item at fifty, while the agent barely moves. The tools most exposed to replacement are the crowded, single-purpose ones — CRM data entry, tier-1 support, weekly reporting — precisely because they automate a well-defined task that everyone runs the same way. The resilient layers are the boring, foundational ones: your database, your cloud, your identity system. Agents don't replace those; they run on top of them and make them more valuable.

This is Dermer's "playground where nobody else is playing" rendered in software economics. The barrier to capturing agent leverage isn't technical genius — it's the willingness to spend a few unglamorous hours mapping one high-volume workflow and piloting an agent on it, instead of reflexively buying the next tool. Most teams skip that step because buying is instant and building takes a week. That urgency is understandable, and it's also the exact mechanism that keeps companies trapped under 291 subscriptions. The businesses that pull ahead are the ones that protect a little time to build the thing that runs itself.

Everyone buys the tool that ships today. The advantage goes to whoever spends one week building the workflow that runs itself next year.
— Michael Dermer

The stakes compound. A company that keeps stacking per-seat tools has a bill that grows with every hire and nothing that scales without more humans. A company that moves its highest-volume workflows to agents has infrastructure that gets cheaper per unit of work as it grows. Same problems, same effort — a canyon of difference in outcome. So before you renew, it's worth knowing exactly which tools are safe and which are on the clock.

How to move from a bloated stack to an agent foundation — starting this week

The good news is you don't have to rip anything out on Monday. The teams that fail at this attempt a big-bang replacement and drown; the ones that win go one workflow at a time. Start with an audit: list your tools and, next to each, write the actual workflow it runs. Most of the 291-app sprawl turns out to be a handful of genuinely load-bearing systems plus a long tail of single-purpose tools automating narrow, repeatable tasks — and that long tail is exactly where agents pay off first.

The second move is to pick your first hill: the one workflow that's high-volume, well-defined, and expensive to run with humans — think CRM data entry, tier-1 support triage, or weekly reporting. Run an agent on it in parallel with your existing tool for 30 days, measure the result honestly, and only then phase out the subscription. The third move is to protect the foundation: don't touch your database, identity, or cloud layers — those are what your agents will run on, and they get more valuable, not less. Audit, pilot, scale. That's the whole playbook, and it's boring on purpose, because boring is what survives contact with a real business.

It's not OK to just tell someone to "adopt AI" — that's handing them a buzzword. You give them one workflow to automate this month, and the stack starts working for them instead of billing them.
— Michael Dermer

The failure mode here isn't caution — being deliberate is right. It's paralysis: staring at 291 subscriptions, deciding it's too big to touch, and renewing everything for another year. Effort spent maintaining a bloated stack burns money and hours. Effort spent moving one workflow at a time to agents compounds into a business that scales without scaling its software bill.

What the agent shift is really measuring

Zoom out from the adoption curves and the agent shift is measuring a structural change, not a passing hype cycle. For most of the SaaS era, software was sold as access: you rented a place to do the work, and the work stayed yours to do. The fact that agents can now take the goal and do the work end-to-end signals that the "rent an interface" model has quietly stopped being the ceiling of what software can be. That's the disruptive reading, and it's real. But underneath runs a more hopeful current: for the first time, a solo founder or a five-person shop can command the kind of operational leverage that used to require a whole operations team.

Every agent you deploy is, whether you'd use the word or not, a hire that doesn't churn, doesn't need a seat license, and doesn't cost more as you grow. That reframe is why the agent shift, disruptive as it is, is genuinely fertile ground for small business specifically. The incumbents with 291 apps and thousands of seats have the most to unwind; the small, nimble builder has almost nothing to unlearn and everything to gain. The pressure on the old model isn't going away. The question is whether you use it to build a foundation that scales — or keep paying per seat for the privilege of doing the work yourself.