The Hidden Cost of the Modern Agency Stack

I've been in the agency business over forty years, and I want to start by saying the tools have got better. Genuinely better. What a five-person shop can do on a Tuesday would have taken a department and a month when I started.

But something has gone wrong underneath that, and it's been getting worse for about five years.

They breed like cockroaches.

It used to be a handful of subscriptions. Now it feels like every other week I'm pulling out a credit card for something — a service one client needs, for one project, possibly once. And honestly, I've gotten sick of reaching for my wallet.

Here's the part nobody warns you about when you're small. A new client lands on your desk with a job that needs a tool you don't own. You can't justify it as a standing monthly cost, because you have a handful of clients and exactly one of them needs this, and not even consistently. So you subscribe for the month, do the work, and cancel it before it renews. Then it happens again six weeks later with something else.

That gets old.

So here's how I'm seeing the modern agency, and I'll start where everybody starts — with the bill. I priced it, because I got tired of guessing.

I think an agency can run on just six categories of subscriptions — prospecting, an SEO suite, a design tool, a video editor, an AI ad generator, and something to report with. Cheapest workable paid plan in each, priced in September 2026: $117 a month. That's fine. That's a phone bill.

Now staff it. A five-person agency needs those six plus AI visibility, a reporting suite, social listening and a project tracker — ten categories, one realistic plan each, seats only for the people who actually touch the tool. $1,859 a month. Take the same ten to fifteen people and the seats scale it to roughly $2,866 to $2,946.

Around thirty-five thousand a year for a fifteen-person shop. And that number is the cheap part of this article.

Priced September 2026. These are real published plans, not worst-case enterprise tiers, and your own mix will differ — but the shape holds. As your agency grows and scales, your costs stack up fast.

That total buys ten disconnected products. None of them knows the client's brand. None of them passes work to the next one. And every one that offers a white-label report produces a different-looking one — different layout, different charts, a different idea of what matters — so nothing you hand a client looks like it came from the same agency.

And don't get me started on SEO reports that are just a data dump. Two hundred charts nobody reads, half of them the client couldn't interpret if they tried, delivered monthly as evidence that something happened.

Not categories of software — accounts. Things with a login, a renewal date and a card on file. Here they are in the four groups an agency actually works in, with the names you'd typically be paying.

Winning the business

Prospecting, outbound, pipeline and press.

Measuring and auditing

Search, AI answers, analytics and reputation.

Making the work

Design, video and social.

Running the agency

Reporting, coordination, approvals and glue.

Every tool in that table is good. I want to be straight about that before I say anything else, because the version of this argument where all the incumbents are stupid is a bad argument, and you've read it a hundred times.

Apollo and Clay are genuinely good at contact data and enrichment. Semrush and Ahrefs hold datasets neither you nor I could build. Figma is the best design tool ever made, and Frame.io solved review rounds properly. Asana and Monday keep work visible. Sprout schedules and listens. Make and n8n will connect anything to anything, if you're willing to become the person who maintains it.

None of them is the problem. Look down that last column instead. Not one of those gaps is a missing feature — every one is a different login, a different interface, a different way of doing the same small thing, and one more hop.

Which is why the browser has fifteen tabs open all day. You download from one, upload to another, copy out of a third, paste into a fourth. At the end of the month you compile a client report out of a dozen sources and try to make it look like one document. Or you give up on that and send four separate reports with four different layouts on them, and let the client work it out.

Follow one prospect through your own week.

They're discovered in the prospecting tool. Researched in a browser with eleven tabs open. Enriched somewhere else. The pitch gets written in a doc. The email goes out of a fourth system. The follow-up lives in a fifth, or in your head. They sign — and now everything you learned about them, the reason you called, the thing you noticed on their site, the competitor they kept mentioning, has to be typed again into a project tool that never met them.

Then the work happens in three more places, and at the end of the month you assemble a report out of screenshots.

Somewhere in that chain the reason the work mattered gets lost. Not dramatically. Quietly, one paste at a time.

And when it comes time to report, somebody has to pull all of it together. That's the owner, usually at night. Every piece of work that justifies the bill is sitting in a different system, so it ends up in a spreadsheet — because a spreadsheet is the only place all of it can be in one view.

That spreadsheet is what you actually bought when you bought ten tools.

This is where most agencies reach next, and it half works.

A project tool is a ledger of intentions. It holds the task, the assignee, the due date, and a status somebody remembered to change. What it doesn't hold is why the task exists — the audit finding, the competitor's ad, the client's offhand remark on the call. That context lived in another tool and didn't come across, so somebody re-explains it in a comment, badly, and the designer builds from the comment.

But the deeper thing is that a project tool has never done a single piece of the work. It tracks that the work should happen. You still open eight other tools and do it.

Which is roughly why I built what I built. I'd spent a couple of years watching how good this stuff had got at doing the actual work rather than describing it.

I designed StyleForge so I could work hand in hand with Claude instead of hand in hand with fifteen tabs.

So I connect the StyleForge MCP to Claude, and then I just talk to it. Onboard this new client. Build them a brand strategy. Pull a colour palette off their site. Here are three competitors — go and look at what they're running. Now audit the site and tell me what's broken.

And Claude goes out and does it. I don't create a ticket. I don't track the steps, or chase where they got to, or go back and mark them done. I ask, and I get the outcome.

I'll be honest about the motivation, because it isn't noble. I am one lazy man when it comes to ticketing. I don't like making them. I don't like tracking them. I really don't like reporting that I've done them. So when Claude takes that off my shoulders, I love it — and most of what used to be my whole Monday is the part I wanted back.

You stay in the loop the whole way, and that's a design decision rather than a limitation. You're the one accountable to the client, so you're the one who reads it, changes it and approves what goes out. Nothing publishes itself. What changes isn't who decides — it's that you spend the afternoon directing work instead of performing every keystroke of it.

That's a different category from project management, and I'd rather name it plainly than pretend we're a better Asana: this is get-the-work-done software. The brief still exists, and it carries the references that produced it — one click back to the workflow that made the winning ad, the campaign that ran it, the numbers that proved it. Status derives from real work rather than a dropdown, because a dropdown is somebody's opinion about progress.

None of that makes it a task tracker for a hundred-person company. At a five-person shop it removes the re-explaining and a good deal of the doing, which is where the month actually goes.

The unlock isn't features. It's that everything resolves the same client first — the strategy, the personas, the colours, the fonts, the story, the connected channels. The AI is never guessing the brand, because the brand is the substrate rather than a line in a prompt.

Which produces some unglamorous consequences. A prospect you win becomes that client's brand kit, carrying the research forward instead of being retyped. A pinned comment on a video at fourteen seconds can be acted on — the layer edited, the moment regenerated — instead of becoming a Slack message with nowhere to go. Reports come out of the record of what was actually done, so the report is a by-product rather than a monthly assembly job.

And seats stop being the unit. Team members draw from one shared pool, so a creative-heavy month spends where the creative is and a research-heavy month spends where the research is, without anyone doing licence arithmetic first.

StyleForge includes over fifty tools built for agencies. In the arithmetic below I'm only counting the core ones — the handful that map onto the subscriptions a modern stack genuinely can't run without. The rest aren't in the comparison at all.

So I priced StyleForge affordably for the small agency: Pro is $79 a month for a solo founder. Teams is $299 for five seats, Teams Plus $499 for fifteen — one shared token pool, a two-month rollover, and more tokens available when a month runs hot. Set that against $1,859 and $2,866 and the arithmetic makes its own argument. Current numbers always live on the pricing page; I'm not going to maintain them in twelve blog posts.

A design team will keep Figma, and should. A sales organisation running real pipeline will keep HubSpot. A mature outbound team with deliverability people will keep their sequencer. There are datasets inside Semrush and Ahrefs that nobody else has.

The goal was never to delete your stack. It's to stop paying senior people to be the glue between the parts of it that should have been connected in the first place.

Not the subscription total. The count of times a week your team rebuilds context that already existed somewhere else.

Small agencies don't need more dashboards. They need the opportunity found, the context kept, the work coordinated, and the client shown why it mattered — without a person in the middle carrying all of it between tabs.

The next agency stack won't be judged on how many subscriptions it has. It'll be judged on how little work gets lost between them.

• The Small-Agency Problem — why this platform exists, and the doorway most agencies walk in through.
• How to make money with AI visibility — what to sell, who pays, and what to charge.
• Why referrals poison an agency's growth — the channel you can't schedule, and the one you can.