Today's AI price isn't tomorrow's: how to budget AI in project management
AI pricing trends in 2026
AI prices are moving on three levels at once: the cost of the models themselves, subscriptions to AI services, and pricing models in everyday software. And they are not all moving in the same direction.
Level 1: model costs. Over the long run, AI is getting cheaper: according to the BenchLM index, prices of leading models as of September 2026 are 84% lower than in March 2023. But it is not a straight line down. Over the last 12 months, prices of the most powerful models rose by 72%, while mid-tier models fell by only 7.7%. The gap between the most expensive and the cheapest model on the market is thousands of times. Even the rate of decline varies widely: from 9x to 900x per year, depending on the level of capability.
Level 2: subscriptions. About $20 a month has become the market standard: that is what ChatGPT Plus, Claude Pro and Perplexity Pro cost as of mid-2026, with Google AI Pro at $19.99. But above that level, $100 and $200 plans have appeared for heavy users; in April 2026, OpenAI added a $100 plan in response to Claude Max. Prices rise less through direct increases and more quietly: the most valuable features move to more expensive plans, and credits and tokens replace clear dollar amounts. In early 2026, software prices rose by 12.2–14.5%, even though the technology itself is getting cheaper.
Level 3: pricing models in software. The SaaS market is moving away from a simple per-user price. According to Bain, about 65% of more than 30 major vendors have added an AI usage meter on top of per-seat pricing. In a Cruxy survey, 97% of 300 SaaS CEOs plan to move away from seat-only pricing within two years. Vendors are changing how they monetise AI, cutting free credits and selling agents as separate products. Gartner forecasts that by 2030 at least 40% of enterprise SaaS spend will shift to usage-, agent- or outcome-based pricing.
As a result, the price a team pays depends both on model costs and on the vendor's pricing experiments. All three levels are moving right now.
AI in task trackers: why the question has become pressing
AI prices and terms in task trackers change every few months. That makes an AI budget today a forecast rather than a fixed amount, and choosing a tool on price alone is becoming risky.
In 2026, almost every task tracker has built-in AI: ClickUp Brain, Asana AI, Atlassian Rovo, Notion AI, monday Sidekick. Most teams already use it. In a Tempo survey of 300 PMO leaders, 91% are piloting or using AI in projects. Yet 42% cannot connect their AI spend to a return.
Part of the reason lies not with the teams, but with the market itself. Let's look at what exactly is changing and how to plan for it.
What changed in 2026
In just a few months, major vendors rewrote the rules for paying for AI, both towards discounts and towards new charges.
Tellingly, even reviews describe the same pricing differently. Some say that Asana's AI is part of the Advanced plan at around $24.99, others that AI is free on all paid plans. When the market moves this fast, even comparison articles can't keep up.
Why prices don't stand still
The AI market is still young, and vendors are still looking for a pricing model that works for both them and their customers.
- Underlying costs change. Language models get cheaper and more expensive, and new ones appear. A task tracker vendor pays the model provider and passes these changes on to customers.
- There is no common unit. Some count credits, some count "actions", some include AI in the seat price, and some sell it as a separate add-on. Comparing "the price of AI" across tools directly is impossible.
- The cost of a single action is opaque. In Jira, a Rovo Chat request uses 10 credits and Deep Research uses 100. But no one knows in advance how many credits a specific team task will take. Reviewers say outright that this makes planning difficult.
- Vendors are testing the market. Generous allowances today are often a way to get teams used to a feature. Terms get revised once usage grows.
What this means for teams
An AI budget approved for the year will most likely need revisiting within a quarter.
The usual approach of "calculate the annual licence cost and close the question" doesn't work here. The amount depends not only on headcount, but on how actively the team uses AI and on terms the vendor can change. On top of that, a bill for credits says nothing about the return: according to Tempo, most teams that use AI can't even separate AI's work from people's work.
So it is more honest to treat an AI budget as a forecast with a range, not a fixed line item.
How to plan an AI budget when everything keeps changing
Instead of an exact figure, you need a system that can withstand price changes.
- Build in a buffer. Plan a range, not a single number: a base scenario and one in which the vendor starts charging for overage.
2. Read the terms for pricing changes. How many days' notice they give before new prices, what happens when you exceed the limit, whether you can set a hard spending cap.
3. Recalculate every quarter. Compare actual usage with the plan and with the vendor's current terms, not the ones that applied when you bought.
4. Measure results in hours, not credits. Credits show costs. The return shows in the time the team has freed up: on reports, planning, task assignment.
5. Don't tie your processes to a single AI feature. If a feature becomes more expensive or changes, work shouldn't stop.
How to choose a task tracker when prices are unstable
A $2–3 per-user difference is a weak argument: in a quarter it may disappear or reverse.
Most often, a task tracker is chosen by comparing price and credit tables. In 2026, such a table goes out of date faster than a pilot finishes. Meanwhile, the cost of getting functionality wrong remains: if a tool doesn't cover key tasks, the team loses far more than it saved on the plan. Missed deadlines, overloaded people, reports compiled by hand.
A more robust order of choice:
- Start with business goals. What should change: deadlines, visibility for managers, team workload, quality of results.
- Translate goals into tasks for the tool. For example: plan realistic deadlines, assign tasks without overloading people, flag risks early.
- Test these tasks in a pilot. Two or three weeks on a real project will tell you more than any price table.
- Only then compare cost, together with the terms for changing it, not just the current price.
Checklist: what to ask a vendor before buying
- Which of our key tasks does the tool cover right now, without customisation?
- What AI is included in the plan, and what is paid separately?
- How is AI usage counted: credits, actions, seats — and how much does a typical task use on average?
- What happens when we exceed the limit: blocking, extra charges, a warning?
- How many days' notice do you give before changing prices and terms?
- Can we set a spending cap for a team or department?
- Can we see which work was done by AI and which by a person?
- How long does onboarding take, and who helps at the start?
Conclusion
Today's AI price is just a snapshot of a fast-moving market. Plan your budget with a buffer, check the terms for pricing changes and measure the return in time saved. And choose your task tracker by how well it helps you reach your business goals: functionality will stay with you longer than any current price.