The Hidden AI Subscription Stack: Why Your $7/Seat PM Tool Now Costs $30/Seat

You signed up for ClickUp Unlimited at $7 per user. A year later, ClickUp Brain rolled out at $9 per user on top. Then Notion forced anyone wanting AI onto the Business plan at $18. Then Microsoft 365 Copilot landed at $21 per user. Then Asana AI Studio quietly started metering credits in the background of your existing plan. Your "we just want to track tasks" tool is now a stack of three subscriptions and an unpredictable credit meter, and the line on the invoice that used to read $35 a month for a 5-person team is closer to $200. This is the new shape of AI add-on pricing in project management — the second wave of per-seat billing, dressed up as innovation.

The first wave of per-seat pricing took a decade to fully reveal itself. The hidden costs of renewals, seat minimums, and guest reclassifications only became visible to most teams once they tried to grow past the free tier. The AI add-on wave is moving faster. In 2026, four of the largest project management vendors have already split AI out as a separate per-seat surcharge — and the math hits small teams hardest because they pay for capacity nobody on a five-person team uses.

The Math at 5, 10, and 20 Users

Before discussing why this is happening, look at what it actually costs. Each of these numbers is the public 2026 pricing for the base plan plus the standard AI tier, billed annually.

ClickUp Unlimited + ClickUp Brain

Unlimited base: $7/user/month. Brain AI add-on: $9/user/month standard, or up to $28/user/month for "AI Autopilot" (Everything AI). At the standard tier:

5 users: $35 base + $45 AI = $80/month
10 users: $70 base + $90 AI = $160/month
20 users: $140 base + $180 AI = $320/month

The AI add-on costs more than the base plan. At Autopilot pricing, AI is four times the base. (ClickUp Brain pricing page)

Notion Plus → Notion Business (forced upgrade)

Old setup: Plus at $10/user/month + standalone Notion AI at $8/user/month = $18/user/month for AI features. New setup in 2026: AI is removed as a standalone add-on and bundled exclusively into the Business plan at $18/user/month. The kicker: the Business plan is paid for every member of the workspace, not just the ones who use AI.

5 users: $90/month ($72/month if you stay on Plus without AI)
10 users: $180/month ($120/month without AI)
20 users: $360/month ($240/month without AI)

A team where two people want AI and eight don't ends up paying the AI premium for all ten. (Notion 2026 pricing)

Microsoft 365 Copilot Business

Microsoft 365 Business Standard: $12.50/user/month. Copilot Business add-on: $18 promotional through June 2026, then $21/user/month standard on annual billing. At post-promo pricing:

5 users: $62.50 base + $105 Copilot = $167.50/month
10 users: $125 base + $210 Copilot = $335/month
20 users: $250 base + $420 Copilot = $670/month

Copilot is 168% the cost of the base productivity suite it sits on top of.

Trello Premium (Atlassian Intelligence)

The model here is slightly different — you can't add AI to Standard. To get Atlassian Intelligence at all, you must be on Premium at $10/user/month (annual). Asana plays the same game with AI Studio: 50,000 included credits at Starter, then a credit meter that depletes silently mid-month if your automations run more than expected. (Trello pricing · Asana AI Studio pricing)

For a 10-person team that adopts the AI add-on across two of these tools — say, ClickUp Brain and Notion Business — the monthly bill goes from a notional $70 to roughly $340. Annualized, that's about $3,300 in AI surcharges before anyone has done the math on whether the AI is actually shipping more work.

Why AI Add-On Pricing Has Become the Default

The structural reason is straightforward: AI inference costs vendors real money. GPU time, model API fees, and infra scaling are non-zero on every prompt your team sends. Vendors need to recover that cost somehow, and the easiest mechanism that fits their existing billing infrastructure is — surprise — another per-seat line item.

But there's a second reason that's less often stated: per-seat AI pricing defends the SaaS revenue multiple. If a vendor bundles AI into the base plan and absorbs the cost, their margin compresses and Wall Street punishes them. If they split AI into a separately priced surcharge, they get a brand new revenue line tied to seat count — and the model the market already knows how to value. ClickUp's stock story, Notion's series-C narrative, Atlassian's earnings call — all of them benefit from "AI ARR per seat" as a separate growth metric.

For the vendor, this is structurally rational. For the buyer, it's the same problem per-seat pricing always created: fast-growing teams pay disproportionately for capacity they don't use. Five-person teams pay for AI seats for the two part-time freelancers who never log in. Twenty-person teams pay for the marketing intern who only opens the tool to mark their own tasks done. The "average AI usage per seat" math that justifies the price doesn't survive contact with how small teams actually staff a project.

The Numbers Behind the Trend

This isn't a vibe — it's a measurable shift. According to Zylo's 2026 SaaS Management Index, 78% of IT leaders experienced unexpected charges on a SaaS bill due to consumption-based or AI pricing models in the past year. McKinsey's 2026 Software Pricing Report found that 62% of SaaS platforms introduced AI-premium tiers, and buyers reported budgeting 25–35% higher to add AI functionality to existing stacks. Average per-employee SaaS spend hit $7,900 per year in 2026 — up 27% in two years, with a meaningful chunk of that increase attributable to AI surcharges layered on tools the employee was already paying for. (Zylo)

A separate consumer analysis put the average AI power user's monthly subscription stack at $116 across ChatGPT Plus, Claude Pro, Notion AI, GitHub Copilot, Grammarly Premium, and a couple of others — and concluded that 60–75% of that stack is redundant, underused, or replaceable at zero cost. The same dynamic is playing out at the team level: tools your company is already paying $X per seat for now also charge $Y per seat for AI features that overlap with the AI subscription a different vendor sells your engineers.

The Two Ways Out

If the per-seat-AI surcharge model is now the default, what are the actual alternatives? There are two, and they're meaningfully different.

Way 1: Flat-rate tools where AI is included or excluded — never priced by seat

The cleanest exit is a tool that doesn't price AI by seat at all. Flat-rate project management tools like Basecamp, ProofHub, and Heimin charge a fixed amount for the whole team — no per-seat layer for AI to attach itself to. Some flat-rate tools include AI features in the base price; others omit AI entirely and let you bring your own. Either way, a 10th, 20th, or 30th teammate doesn't trigger an AI add-on bill.

This route is honest about a tradeoff: vendor-built, deeply-integrated AI features (the kind ClickUp's Brain or Asana's AI Teammates are trying to ship) are not what flat-rate tools usually compete on. If your team has decided that proprietary AI features inside the PM tool are worth a 3x bill, this isn't the route for you. If your team's AI workflow already lives in Claude or ChatGPT and you mostly want your task tool to stay out of the way, this route adds zero AI surcharge to your monthly invoice.

Way 2: MCP — your existing AI subscription reaches into your tools

The second route is more interesting and fairly new. Model Context Protocol (MCP) is an open standard that lets AI assistants — Claude Desktop, ChatGPT, Cursor, custom agents — read and write to your tools without the tool vendor having to ship its own AI product. If your task management tool exposes an MCP server, your team's existing AI subscription does the work: you ask Claude to summarize this week's open tasks, and Claude reads them directly out of your tool. No second AI seat, no credit meter, no Brain add-on.

The numbers: a Claude Pro subscription is $20/user/month and covers everything from coding to email triage to PM workflows. Adding ClickUp Brain at $9/user/month for the same AI capabilities scoped to one tool is a clear duplication of spend. The MCP route makes your existing AI subscription portable across every MCP-enabled tool you adopt — you pay once, the AI works everywhere.

Heimin made the MCP bet explicitly — the entire task system is exposed as an MCP server, and your AI agent can manage tasks, comments, projects, and the CRM through your existing AI subscription. There's no "Heimin Brain" add-on on the roadmap because we don't think there should be one. We're not the only tool taking this approach, but in 2026 we're still in a small minority.

Practical Takeaways

Audit your stack for AI duplication. List every tool that charges per-seat AI. Then list the AI subscriptions your team already has (ChatGPT, Claude, Copilot). Anywhere the per-seat AI overlaps an individual subscription, you're paying twice.
Calculate your team's "AI inflation rate." Take your current per-seat PM tool cost. Add the published AI add-on for that tool at your current headcount. The percentage increase is the AI tax your vendor is hoping you'll quietly absorb at renewal.
Distinguish forced bundles from real choices. Notion's move to bundle AI into Business at $18 is a forced bundle — you can't get AI any cheaper. ClickUp Brain at $9 is technically optional, but every onboarding and upsell flow is built to make it feel mandatory. Read the fine print.
Pilot the MCP route on one workflow. Connect Claude Desktop to one tool with MCP support and see if your team actually loses anything by canceling that tool's native AI add-on. Most teams discover the answer is "nothing meaningful."
Renegotiate at renewal. AI add-ons are the easiest line item to drop in a renewal conversation, because vendors know most customers paying for them aren't using them at the rate that justifies the price. Try removing the AI tier and see what happens.

The Heimin Take

We built Heimin's MCP integration before we built any "AI features" of our own — and that ordering was deliberate. Our 5-to-20-person customers were already paying for Claude or ChatGPT. Building a parallel "Heimin Brain" priced at $9 per seat would have been a clean revenue line for us, but it would have meant our customers paying twice for AI capabilities they already had. The flat-rate model and the MCP model point in the same direction: don't charge for capacity nobody on a small team uses.

For the 5-to-20-person team running the math at renewal, the question isn't whether AI in PM tools is valuable — it often is. The question is whether the value scales with seat count, and whether the AI features your vendor is selling are duplicating something your team already pays for elsewhere. In 2026, the honest answer for most small teams is "yes" and "yes."