Think Build Implement Repeat
London, UK +44 7367 067226
WhatsApp FOLLOW f in X
  1. Home
  2. Blog
  3. Some Customers Use Our AI Features So Much They Cost More Than They Pay. What Do We Build?
Problems We Solve

Some Customers Use Our AI Features So Much They Cost More Than They Pay. What Do We Build?

Heavy users on flat SaaS plans can cost more in AI calls than they pay. We meter usage per account, show who is unprofitable, and add fair limits and tiers.

Updated 3 min readBy SpiderHunts Technologies

Free estimateNo obligation

Get a free estimate

Tell us what you need. A senior engineer reads every enquiry.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →

Quick answer — TL;DR

When plans were priced before anyone knew what each account would cost in model calls, a few heavy users on a flat plan can quietly cost more than they pay. We attribute every AI call to the account that caused it, show you usage and cost per account against its plan, and build the limits, credits or usage tiers that let your pricing match what customers actually consume.

The plan price was set before anyone knew

You launched with a simple monthly price per seat or per account, like most SaaS products. The AI features were the reason people signed up, and they use them. Some use them lightly. A few use them all day, run bulk jobs, upload enormous documents, or have wired your product into their own automation so it calls the AI far more than a person ever would.

The model provider's invoice arrives as one total. You can see it rising, faster than revenue some months, but you cannot say which customers are behind it. Your gut says a handful of accounts are costing more than they pay. You have no way to prove it, so you cannot decide what to do about it.

Why flat plans and AI features clash

Traditional SaaS pricing assumes a heavy user costs you roughly the same as a light one, because serving another page is close to free. AI features break that assumption. Every request has a direct cost, and it grows with how much a customer sends and how often.

  • Plans were priced on competitor prices or gut feel, before usage data existed.
  • Model calls are not recorded against the account that made them.
  • There are no limits, so one account can consume without any ceiling.
  • Longer inputs and bigger outputs cost more, and heavy users tend to send both.
  • Customers who automate against your API or run bulk jobs use far more than any person clicking.

The founder's question is not really technical. It is whether the pricing model still makes sense. But that question cannot be answered without the numbers, and the numbers need building.

What not knowing is doing to the business

Blind spotWhat it leads to
Cost per account unknownUnprofitable customers look like your best ones
No usage ceilingOne account can move your monthly bill on its own
Plans priced by guessworkLight users overpay, heavy users underpay
No usage data at renewalNothing to base a price change or upsell on
Investor questionsGross margin figures you cannot stand behind

There is a trap in the obvious fix. Raising prices for everyone punishes the light users who were paying their way, and they are the ones most likely to leave. Without per-account data, you cannot target the change where it belongs.

How we measure AI usage and price for it

  1. Route every model call through one place. A thin layer in your code that all AI requests pass through, whichever provider they go to, such as OpenAI or Anthropic Claude.
  2. Record usage against the account. Each call is logged with the account, user, feature, model and the tokens or units it used, so cost can be worked out from the provider's own rates.
  3. Show usage and cost per account. A dashboard that lists accounts by AI usage next to what they pay, so the heavy users and the unprofitable plans are visible for the first time.
  4. Reduce cost where it is waste. Caching repeated work, trimming oversized inputs and using a cheaper model for simpler tasks, so heavy use costs less before any price changes.
  5. Add limits that fit your product. Monthly allowances per plan, soft limits that warn before hard ones, and rate limits on automated use, set as configuration rather than code.
  6. Build the pricing mechanics. Usage tiers, credit packs, an add-on for heavy AI use or metered billing through Stripe, depending on what suits your customers.
  7. Tell customers where they stand. An in-app usage meter and warnings as they approach a limit, so nobody hits a wall without notice.

We give you the data and the mechanisms. Which pricing model to choose is your decision, and we will lay out the trade-offs of each option with your own usage figures in front of you.

Pricing on evidence

You can see, account by account, what AI usage costs you against what each customer pays. The accounts that were quietly costing too much are known, and you can talk to them with facts rather than a blanket price rise. New plans are designed around real usage patterns. Limits protect you from a single account running up the bill, and customers can see their own usage, which makes an upgrade conversation straightforward.

Margin figures for the board or investors come from records of actual calls, not an estimate spread evenly across customers.

Could this be your pricing problem?

  • Your plans are flat and include unlimited or loosely limited AI features.
  • The model provider's bill is growing faster than revenue.
  • You cannot say which accounts account for most AI usage.
  • Some customers automate against your product or run large batch jobs.
  • You are considering a price rise but do not know who it should apply to.

FAQ

Frequently asked questions

The questions readers ask us after this guide.

Still have a question?

Ask us directly — a senior engineer will get back to you.

Ask about your project

Do we have to move to usage-based pricing?

Not necessarily. Many products keep simple plans with sensible allowances and an add-on for heavy use. The data tells you which approach fits your customers.

Can we measure usage for past months?

Usually only roughly, unless calls were already logged per account. Accurate per-account figures start from when the measurement layer goes in.

Will limits upset existing customers?

They can if introduced abruptly. We usually suggest starting with visibility and warnings, then limits with notice, and letting existing customers keep terms for a period if you choose.

Does this work if we use more than one AI provider?

Yes. The layer records usage per call whichever provider handles it, using each provider's own rates.

What do you need from us?

Access to the codebase, your plan structure, and recent invoices from your AI providers.

Keep reading

More on Problems We Solve

Start here

Talk through your SaaS product with us

Tell us what you are building or running, who it is for, and the decision or problem in front of you. We will give you a straight view of the options, what each involves and which we would pick, and if you do not need us for it, we will say so.

  1. You tell us what you needTwo minutes on the form, or a message on WhatsApp.
  2. A senior engineer reviews itAnd comes back with questions, a realistic range and an honest view on fit.
  3. Free 30-minute scoping callWe talk through scope, options and a realistic estimate — with no obligation.
Free estimateNo obligation

Talk to someone who builds this

Send a short brief and we will come back with an honest view and a realistic range.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →