Rethinking AI pricing for the enterprise

Mar 31, 2025
 • 
0 MIN READ
Gold share tray icon
Stephanie Keep
Content Marketing
Get updates into your inbox
Sample Metronome billing dashboard
Share

The landscape of AI pricing is evolving rapidly, and enterprises are facing a crucial challenge: finding a pricing model that aligns with their needs for predictability, scalability, and value. In a recent episode of Unpack Pricing, Metronome CEO Scott Woody sat down with Matan-Paul Shetrit, Director of Product Management at Writer, to explore the shifting dynamics of AI pricing and why the current per-token model is falling short for large enterprises. Here’s a dive into the highlights from their discussion, touching on current challenges in the landscape of AI pricing, how to support enterprise customers, and what the future might hold for AI pricing.

Per-token pricing works for experimentation, not for scale

One of the core issues with per-token pricing is that it creates uncertainty for enterprise buyers. While developers may be comfortable experimenting with AI model APIs on a credit card, large-scale adoption requires cost predictability. Matan explained how enterprises approach AI procurement similarly to cloud services—leaders need to know their budget upfront, not face unpredictable bills that fluctuate based on usage.

There's a problem with AI pricing. Which is predictability. If a CIO says, Hey, Matan, this is the use case we want to tackle. We have these types of docs, this type of information, and this is what we want to do. It's like, okay, great. And he asks, how much is it going to cost me? I'm like, I don't know. The problem with token-based pricing is that on average you have 3 words per token, but you can’t run a business unit with averages. You can say, okay it’s going to cost a million dollars. And then you start using it. And all of a sudden prices start going up to like $5, $10, $20 million.
– Matan-Paul Shetrit (13:29)

This unpredictability leads to two issues: some enterprises significantly underutilize their AI commits due to fear of excessive costs, while others deploy AI and later face unexpected ballooning expenses. Neither scenario is ideal for long-term adoption.

Enterprises aren’t just buying AI models—they’re buying platforms

For enterprises, AI adoption isn’t just about getting access to models. It’s about integrating AI securely and efficiently at scale. Writer doesn’t just sell LLMs; it offers a full-stack AI platform with security, observability, and control features. When the Writer team determines pricing, it’s done in a way that holistically takes the product value into consideration. As Matan notes, the core product offers one level of value, but it’s the combination of many secondary elements—like security, transparency, uptime, etc.—that makes the offering unique and allows you to differentiate even your pricing.

As we layer these experiences on top of the model layer, that's where I'm able to extract value. Obviously the models are great and without good models or great models, you won't be able to even do what's on top, but I can charge because I give you all these other features and experiences that are built in this foundational model layer.
– Matan-Paul Shetrit (29:58)

Similarly, AI providers that focus solely on model access will struggle to win over enterprise buyers. These companies need security, compliance, governance, and observability baked into the offering—pricing should reflect that full value, not just usage.

AI companies must guide enterprises through AI transformation

Unlike consumer AI applications, enterprises require hands-on guidance to adopt AI successfully. Selling AI to enterprises isn’t just about delivering technology—it’s about helping CIOs and IT teams navigate the transition. Just as they needed guidance during the era of cloud transformation, companies need education on how to go through this AI transformation, and the AI companies that realize this will make faster headway than those that don’t.

We work with companies to help them manage the transformation. The idea that an enterprise would build against an API self-serve will not happen. They need to be handheld because they have their way of doing things, and it’s been working for them. They're great companies and they want the transformation but they need your help. They need to understand what use cases are better and worse. They need to agree with you on what the success looks like.
– Matan-Paul Shetrit (37:41)

AI vendors need to invest in change management, proof-of-concept (POC) support, and training to ensure adoption. The companies that succeed will be those that offer more than just an API—they’ll provide a roadmap for how enterprises can effectively deploy and scale AI.

Large enterprises need cost certainty to scale AI

Ultimately, pricing models designed for startups and consumers don’t work for the Fortune 500. To drive AI adoption at scale, enterprises need predictable pricing structures. Writer has shifted to a model where enterprises pay a fixed platform fee that includes a generous token allocation. This approach allows companies to experiment without fear of runaway costs and ensures that AI adoption aligns with business value.

We want you to start using the platform, but we don't want you to be scared. Essentially, we're moving to a place where we're going to have a fixed platform fee. But we're going to bake in, as part of the platform fee, a monthly token usage for every month.
– Matan-Paul Shetrit (15:02)

This structure makes pricing discussions easier and helps internal AI champions justify additional budget allocations by tying AI usage directly to ROI.

AI pricing must evolve for the enterprise market

As AI adoption matures, pricing models will need to evolve to support enterprise scalability. The key lessons from this discussion highlight a few critical trends:

  • Per-token pricing is too volatile for enterprises and needs to be replaced with more predictable structures.
  • AI pricing will follow a trajectory similar to cloud pricing, balancing cost, commitment, and value.
  • Enterprises are investing in AI platforms (not just models) so pricing must reflect the full-stack offering.
  • Successful AI vendors won’t just sell technology. They’ll help enterprises navigate AI transformation.
  • Fixed platform pricing with built-in usage allocation offers a more viable path for enterprise AI adoption.

The future of AI pricing will be shaped by these lessons, as vendors refine their models to meet enterprise needs. Companies that can balance predictability with scalability will be best positioned to succeed in this fast-moving space.

Catch Matan and Scott’s full discussion on Episode 5 of Unpack Pricing.

Company Industry Outcome-Based Pricing Model Key Metrics for Pricing Notable Features
Salesforce (Agentforce) CRM / AI Customer Service

$2 per conversation handled by Agentforce (AI agent)

A conversation is defined as when a customer sends at least one message or selects at least one menu option or choice other than the End Chat button within a 24-hour period.

Number of support conversations handled by the AI agent

First major CRM to adopt a "semi"outcome-based pricing for AI; aligns cost with actual support volumes (clear ROI)

Addresses inefficiencies of idle licenses by charging only when value (a handled conversation) is delivered

Intercom (Fin AI) Customer Support Software

$0.99 per successful resolution by "Fin" AI chatbot - clients pay only when the bot successfully resolves a customer query

Fees accrue based on AI-solved issues

Count of support conversations resolved by the AI agent

Early adopter of AI outcome-based pricing in 2023

Lowers adoption risk by charging for resolved queries instead of a flat rate; combines usage- and value-based pricing to tie cost directly to support effectiveness.

Zendesk (AI Answer Bot) Customer Support

Per successful AI chatbot-handled resolution

No charge if the bot fails and a human must step in

Number of customer issues or tickets auto-resolved by the bot

Aimed at cost-conscious customers wary of paying for unproven AI

Aligns price with realized automation benefit; part of a broader industry shift from per-agent pricing to value-delivered pricing in support

Chargeflow Fintech (Chargeback Management)

Charges a fraction of recovered funds on chargebacks

Example: ~25% fee per successful chargeback recovery

No fees for chargebacks lost

Alert service charges $39 per prevented chargeback

Value/count of chargebacks recovered (disputes won) and chargebacks prevented (for prevention alerts)

4Ă— ROI guarantee on recoveries

No contracts or monthly fees

Revenue comes only from successful outcomes; pricing directly aligns with merchant's regained revenue, meaning Chargeflow only profits when the client does (win-win model)

Riskified*

(source: https://www.chargeflow.io/blog/riskified-vs-forter)

E-commerce Fraud Prevention

remain fraud-free

PAYGO, 0.4% per transaction

Only charges for transactions it approves that

Number or value of approved transactions without fraud (i.e. successfully processed legitimate sales).

Provider shares financial risk of fraud with clients; pricing tied to outcome of increased safe sales

Incentivizes vendor to maintain high accuracy (they only profit when fraud is stopped)

Foster continuous improvement in their fraud-detection algorithms

Subscribe

Keep up with the latest in
pricing and packaging