The Dirty Little Secret Behind Gen AI Functionality Pricing

Archive | Upgrade to Premium | Sponsor

Hi {{FIRST_NAME|Readers}},

Software vendors are quietly abandoning “per-seat” and “per-module” pricing in the “era of AI”…

And they're hoping you won't notice until it's too late…

Take this line I heard in a recent procurement software demo:

"We're transitioning to outcome-based pricing so you only pay when our AI functionality delivers measurable business results to your procurement organization."

Sounds great, right? Very customer-centric!

But here’s the dirty little secret behind that line…

Access to generative AI models is charged by usage at the enterprise level.

(contrary to your personal usage of Chat GPT, Claude, Gemini or Co-Pilot)

Giving you unlimited Gen AI usage under “fixed price” licenses would trigger an unsustainable "cost death spiral" for software vendors…

After all, they are caught between a rock and a hard place: paying their Gen AI provider “per use” in the background!

So they're passing variable costs to you using "outcome-based pricing."

The question is… What outcomes are you actually paying for? Is it a good deal?

Let’s find out in tonight’s note.

Onwards!

📰 In this week’s edition:

Note: Some of the content listed above is only available in the email version of this newsletter. Don’t miss out! Sign up for free to get the next edition.

👀 In Case You Missed It…
My Best Linkedin post this week:

If you can swing it, I believe this is THE ProcureTech software vendor strategy.

That additional license cost for Gen AI supported functionalities? It’s not about access to cutting-edge technology... It's about basic survival math ($).

Here's the dirty little secret behind every "Gen AI-powered" feature:

You ask for contract analysis…
→ Vendor calls OpenAI API
→ Costs them $0.30-$3.00 per document

You want spend insights…
→ Multiple Claude API calls
→ Costs them $0.50-$5.00 per analysis

You chat with their "AI assistant"…
→ Could be pennies... or dollars per conversation depending on the length

Gen AI currently burns through computing power like nothing we’ve ever seen before… And, there’s less than a dozen commercial, enterprise grade models on the market… And, everyone is building software that taps into them!

Unlike your personal usage of Chat GPT, Claude, Perplexity or any other model, businesses need to pay every time they call a Large Language Model (LLM)…

That’s a small nuance with a BIG impact!

This means that under a traditional "per-user" or “per-module” pricing model, heavy Gen AI functionality users would decimate your software vendor’s margin.

This is unsustainable…

So vendors are pivoting to "outcome-based pricing" for AI functionalities to preserve their margin and transfer risk…

If your vendor can't predict their AI costs, they definitely don't want to absorb them…

So they’re passing them onto you!

Let's clean house. Here's what's really driving the pricing panic:

Old Model (being killed quietly):
→ Fixed cost per user/month regardless of usage
→ Predictable budgets
→ Vendors eat the variable costs

New Model (being pushed hard):
→ "Pay for results achieved"
→ Costs scale with your "success"
→ You absorb the usage risk

Sounds reasonable, right?

The problem is most vendors proposing "outcome-based pricing" haven't figured out how their AI usage actually improves your bottom line...

…And if we’re being honest, neither have you!

If this is the case, you should be hearing “usage-based pricing” but with a nice marketing lipstick applied. 😅

Let's break them down by what vendors call them vs. the dirty little secret behind each one.

What They Say: "Pay a percentage of savings our AI identifies"

The Dirty Secret: Pay-per-analysis with value-based marketing

Use your calculator. When "outcomes" perfectly correlate with AI usage volume, it's not outcome-based pricing.

What They Say: "Only pay when AI delivers measurable value"

The Dirty Secret: Usage-based pricing with a value veneer

Hook you with "value-based" language while hiding per-transaction charges.

What They Say: "We share the risk and reward of AI implementation"

The Dirty Secret: You absorb all the AI cost volatility

Perfect for vendors who want guaranteed revenue while you handle unpredictable AI expenses.

Here's how to cut through the marketing spin:

If your vendor can't answer these, they're asking you to subsidize their AI experiments.

So demand transparency. Make “should cost” economics a prerequisite.

If a vendor is truly looking to partner, these conversations will be transparent.

Instead of accepting vague "outcome-based" proposals:

This is the same exercise as working out the costs of your raw materials… Just applied to the Gen AI functionality market.

🔥 Pro tip: Ask vendors to show you their AI cost breakdowns. The honest ones will explain their external dependencies. The sketchy ones will dodge the question.

Your vendors decide in the first pricing conversation whether they'll admit the “dirty little secret”…

True outcome-based pricing: "Pay 5% of procurement savings we generate"

Fake outcome-based pricing: "Pay per AI analysis that identifies savings opportunities"

The difference? One actually scales with your success. The other scales with vendor costs disguised as your success.

True outcome based pricing is GREAT! It ties costs to the bottom-line of your P/L.

BUT the bigger the scope of the software you’re looking at, the harder to tie usage directly to P/L benefits…

The bottom line: Outcome-based pricing CAN work. But only when both parties understand the true economics and do the hard work of figuring out how they’re going to share the risk… And the reward.

Unfortunately, most businesses aren't ready for real outcome-based pricing… It’s hard to get right… For starters, you need great baselines on the underlying KPIs… Then, you need the right culture around partnerships. That’s already a big ask in most businesses…

In most cases, outcome based is just a fancy word for "pay per use."

Don’t “pay per use” for Gen AI functionalities unless the ROI is crystal clear!

Otherwise, the invoice amount will be the only thing that’s crystal clear 😅.

What has been your experience on AI functionality pricing with your software vendors?

Let me know in the comments 👇