Buying and Selling AI: A Pricing Nightmare?

Software pricing has always been challenging, for both buyers and suppliers. Because of its granularity, costs were hard to manage, invoices were riddled with errors and buyer-supplier relationships were at best tense, at worst confrontational.

Now, AI is reshaping the economics of Software-as-a-Service and creating new challenges for those relationships. Many suppliers are moving from predictable per‑seat fees to usage‑based (API calls, credits) or opaque AI tiers (pre‑packed “AI bundles” with hidden ceilings). Examples are Salesforce (Einstein), Adobe (Firefly), Microsoft (Copilot) and also dozens of point‑solutions. It is easy to forget that SaaS was itself quite contentious just a few years ago. Indeed, as recently as 2018, WorldCC worked with industry to create a ‘fair contracting guide’. The dust has hardly settled, and here we are with another major shift in the commercial model, indicating once more the speed of change.

For buyers, any time an internal team adds an AI tool, it can trigger exponential compute and feature fees, causing annual spend to increase even if the headcount and usage footprint stay flat.

Added to this, since the old “one price covers all modules” model can’t account for variable inference costs or premium data‑processing tiers, suppliers constantly invent new chargeable line item

The supplier approach: reasonable … and opportunistic

AI R&D and cloud‑compute costs are real. Usage‑based pricing aligns revenue to an increasingly unpredictable cost base and reflects genuine consumption. In that context, the changes are reasonable. But on top of this, many suppliers create premium “AI add‑ons” without transparency on unit economics or clear metrics, while locking buyers into auto‑renewals and creating unexpected price increases.

Later in this article, we will reflect on whether it is in a supplier’s interests to take an opportunistic approach, or to invest in establishing a differentiated relationships with its customers.

Procurement’s role – are they part of the problem?

Procurement teams (not surprisingly) tend to put this shift down to unscrupulous and untrustworthy suppliers. But to an extent, they may have some accountability for the situation they now confront. They often operate with rigid negotiation practices, based on traditional procurement benchmarking (per‑seat, term‑discount). This approach breaks down when the key metric is “per thousand API calls.” Buyers may also accept default terms rather than demanding clear usage caps, audit rights or outcome‑linked SLAs.

There is also often a limited relationship focus, driven by traditional commodity thinking and an entrenched lack of trust.  A transactional “lowest total cost” mindset overlooks the value‑sharing and governance clauses that can provide greater control over AI costs. Teams that co‑innovate on pricing models – for example, gain‑share, volume‑discount thresholds, or pre‑paid credit blocks – may win better outcomes.

Since AI‑driven SaaS pricing is here to stay, Procurement needs to evolve from arguing about seat‑counts to more meaningful value‑and‑usage negotiations. They must insist on visibility into AI cost drivers, and embed dynamic guardrails in contracts, such as  usage alerts, true‑up caps, exit‑assistance clauses. These can help keep cost creep under control.

Becoming a smart buyer

If you don’t come prepared, suppliers will price and package to their advantage, so here is what buyers need to do.

1. Know the true cost drivers
   You should ask for breakdowns of AI‑compute, data‑processing and feature fees and to the extent possible obtain benchmarks from peers or industry surveys.
2. Negotiate on value, not just price
   Consider whether you can link fees to business outcomes such as processing time saved. You may want to build in gain‑share or volume tiers that reward your growing usage. However, initiatives like this depend on the quality of your data – you may need AI tools in order to accurately baseline and monitor.\Seek visibility and guardrails
3. Seek visibility and guardrails
   Incorporate real‑time usage reports and alert thresholds in your SLA (if your supplier is making use of AI, these should be embedded capabilities). Negotiate a cap on ‘true‑ups’ or require reset negotiations when costs exceed projections.
4. Treat AI tools like enterprise software
   You may want to use an MSA + order form model, rather than bespoke SOWs, so you have standard audit, data‑portability and exit rights. You should engage cross‑functional stakeholders (IT, security, finance) early to validate needs and controls.

The best way to control spend is by arming yourself with the right data, contract levers and governance.

A growing urgency: the shift to services as software

As the market moves towards Services‑as‑Software, the issues we are discussing become more critical. Among the changes we expect to see are:

1. Dynamic, Usage‑Driven Fees
   Metered billing (per API call, document, transaction) means costs can escalate overnight if left unchecked.
2. Opaque Packaging
   Bundled “AI tiers” hide compute surcharges and feature premiums, so unmonitored spend can quickly go out of control.
3. Evergreen Contracts
   Auto‑renewals and unilateral feature updates may lock you into the supplier’s next pricing model unless you negotiate strong exit and change‑control clauses up front.
4. Scale / Margin Leverage
   As you onboard more “micro‑services,” each one will add a subscription layer and this multiplies risk unless you standardize your negotiation playbook.

The key point with SaS is that every new service is a potential driver for cost increases which may be hard to control. Procurement processes need to focus on areas where you need transparent metrics, agreed caps on true‑up reconciliations, and perhaps introduce value or outcome‑linked pricing.

Opportunities for smart suppliers

Suppliers who seek to differentiate and lean into transparency and partnership can become “supplier of choice” without significant loss of revenue or margin.

1. Predictable, Value‑Linked Pricing
   This can be achieved by offering a clear “base” subscription fee plus tiered usage bands with steep volume discounts. Along with this, tie any premium AI features to demonstrable business outcomes (e.g. $100 per 1% reduction in processing time), so customers understand and can measure exactly what they are buying (though once again this depends on having an agreed baseline and accurate performance metrics).
2. Shared Governance & Analytics
   Suppliers should provide real‑time dashboards on consumption, costs and ROI and encourage joint reviews, perhaps quarterly, to fine‑tune forecasts, proactively identify and address overruns and build trust. This will help in shortening renewal cycles.
3. Risk‑Sharing Structures
   Again depending on the quality of data, gain‑share or cost‑smoothing models may now be more viable. Suppliers might also propose to refund a portion of over‑use fees if the service under‑delivers, or cap true‑ups at a small percentage of the annual contract value. Approaches like this give buyers confidence and reduce their fear of runaway costs.
4. Modular, Upgradeable Services
   Scrupulous suppliers might package their offering into “plug‑and‑play” modules so customers only pay for what they need and can easily expand when they see value. That flexibility lowers the barrier to entry for new seats or features.
5. Fast Time‑to‑Value
   Focus on mechanisms that allow rapid onboarding, providing templates and best practices so customers see impact in days, not months and the supplier speeds time ot revenue.

These approaches are all about positive differentiation. A supplier who shares data and aligns on outcomes stands out. Transparent, predictable economics cuts negotiation time. By structuring discounts around volume and value delivered, and by reducing churn and frequent re-bidding, a supplier can capture more share over time.

In short, a smart supplier doesn’t have to leave its margins on the table; it simply shifts from extractive pricing to a partnership model that drives growth for both sides.

‍