Agency ops stack in 2026: fewer seams, not fewer tools

A seam is a point where the same fact has to be entered twice because two systems do not share it. The most useful thing published about agency software in 2026 is that the average company's stack got bigger, not smaller, which makes 'buy fewer tools' the wrong advice to give an agency. The target worth chasing is fewer seams.

Key Takeaways

The average business managed 144 SaaS applications in Q2 2026, up 4.3% in one quarter and 10% since June 2025, according to Vertice.

Consolidation does not hold. Stack counts dip during audit cycles, then rebound to record highs within a couple of months.

The three forces growing stacks are specialist AI tools, shadow IT, and integration gaps that spawn sidecar apps.

For a small agency, app count is the wrong metric. Count the seams instead: how many times a fact gets retyped or reconciled.

The stack grew, and nobody cleaned it up

Vertice is a procurement platform that publishes stack size data derived from over 75 billion dollars of global software spend it processed in 2026. Its July 2026 update puts the average business at 144 SaaS applications, up from 138 in Q1 2026 and 131 in June 2025: 4.3% growth in a single quarter, 10% in a year, and a record high for Q2 2026 (Vertice, 2026). The underlying figures are published at vertice.one/insights/saas-apps-per-business.

So the 2026 picture is not the tidy consolidation narrative that most vendor decks are still built on. It is consolidation as a widely stated priority that keeps losing to procurement. Worth noting the boundary: this is enterprise spend data, not agency data, so it tells you about the direction of the market rather than about any particular firm.

Cleanup without a home for the data does not hold

The more useful finding is what happened when organisations did try. Vertice tracks the dips as well as the growth, and they are consistent. The average stack fell from 136 in December 2025 to 130 in January 2026, a clear start of year audit cleanup, and then rebounded to record highs within a couple of months. Vertice's own reading is that these cleanups are historically temporary (Vertice, 2026).

Why they rebound is the part that matters for an agency. A cleanup cancels subscriptions, but it does not change where the data lives. If client scope still sits in a contract tool while delivery sits in a project tracker, then deleting the spreadsheet nobody trusted just moves the reconciliation somewhere less visible. The stack shrinks on paper and the work comes back. The only cleanup that holds is one that moves a fact to a single permanent home and deletes the other place it was kept.

Why app count is useless as a metric for a small agency

Here is the mismatch in most of this advice. A 144 app average describes a company large enough to have a procurement department, and no eight person agency will ever approach it. Chasing app count is like trying to lose weight by owning fewer clothes. The number that predicts an agency's actual friction is the count of seams, and that number tends to sit between four and twelve no matter how many tools you own.

The test is simple. Pick any client. Ask how many times this month somebody retyped or reconciled a fact about them across systems: the scope they approved, the hours worked, the amount invoiced, the date delivered, the contract they signed. Each of those is a seam if a person had to make two systems agree. Count them, then count the hours spent making them agree.

The three forces growing every stack right now

Vertice names three drivers of the growth, and all three show up in agencies at a much smaller scale.

• Specialist AI tools. Niche point solutions get bought to solve one narrow problem while the core platform stays put. Each one becomes a new place a piece of client context can live.

• Shadow IT. Low cost, click to buy subscriptions get purchased on a card at department level, outside whoever is responsible for the stack. In a small agency this is usually one founder's card.

• The integration gap. This one matters most here. When a core platform does not quite cover a specialised need, teams add what Vertice calls a sidecar app to fill it. A sidecar exists only because the main tool missed something, and it is defined by the fact that it re-states data the main system already has.

That third force is the mechanism behind most agency sprawl. The project tracker is fine, so nobody argues for replacing it. But it does not do billing cleanly, so an invoicing tool is added. The invoicing tool does not hold the signed scope, so a contract tool is added. Neither one holds the client's view, so a portal is added. Five tools, four seams, and at every single step the individual purchase looked reasonable.

A worked example: counting the seams in a six person agency

Take a six person agency running one retainer client. Their stack is a project tracker, a spreadsheet, an invoicing app, a contract tool, and email. Nothing here is a bad purchase. Here are the seams.

• A scope change is approved verbally on a call. Nobody updates the contract, so the tracker budget and the signed agreement now quietly disagree.

• Hours are logged in the tracker, the invoice is built in the invoicing app, and the two are reconciled by hand at month end.

• The client status update is a person reading the tracker and typing a summary into email, so it is current as of whenever they last looked.

• A new starter is added to the tracker on day one. The invoicing app finds out three weeks later, when the first invoice comes up short by their hours.

• Month end exports the tracker into the spreadsheet, and the spreadsheet is compared against the invoices by hand.

Five seams. Add up the human reconciliation and it comes to roughly eight hours a month: about three on the month end comparison, two chasing the scope change, one writing the status update that duplicates data the tracker already holds, one and a half copying hours into the invoice, and half an hour fixing the missed starter.

Eight hours a month is about 1,200 dollars at a 150 dollar blended rate, so 14,400 dollars a year, and it is the same eight hours every month. It is also invisible in any single tool, because the cost is not in the subscriptions. It is in the coordination between them. The honest caveat is that eight hours out of roughly 960 team hours is under 1%, which is exactly why it never gets assigned to anyone. It is too small to be a priority and too constant to be a one off.

What to consolidate first, in order

Do not start by choosing a platform. Start by choosing the fact that hurts most when it is wrong, and give that one fact a permanent home.

1. Pick the seam that already costs you money, not the one that is most annoying. Unbilled time is usually the right first candidate, because time tracking and billing carries a direct revenue number rather than a convenience one. The mechanics of that gap are covered in time tracking and billing belong together.

2. Move that fact, and only that fact, into one system. Deleting the other copy is what makes it stick, and it is the step almost every consolidation effort skips.

3. Re-run the count a quarter later. If the seam is back, you removed a tool without removing a job, and the work simply moved somewhere less visible.

4. Only then look at the client facing seams. A client portal can fix several at once because it reads the same project data rather than restating it, which is why portals beat email on mechanics and not just on feel.

The order matters because step two is the only step that creates a durable result, and moving one fact is far cheaper than migrating an entire stack at once.

When staying split is the right call

Consolidation is not a virtue by itself, and the 2026 data is a fair warning against treating it as one. If a tool is genuinely better at its job than whatever would replace it, and it shares data rather than restating it, then keeping it is the consolidation.

The honest test is whether the tool creates a seam. A tool that holds its own data and nobody has to reconcile is a reasonable addition. A tool that exists to re-state what another tool already knows is a seam with a subscription attached. Most sprawl is the second kind, and it is the only kind worth fixing.

This is also why the single workspace pitch is not automatically right. Depth still varies between platforms, and the wrong all in one tool simply relocates the seam from between tools to inside the tool, where it is harder to see. The argument for one workspace holds only if the chosen workspace actually holds the data rather than approximating it from the outside. For the wider category, read what is agency management software.