BBVA is a global financial services group operating in more than thirty countries with roughly one hundred and twenty thousand employees. Its OpenAI estate covers ChatGPT Enterprise seats and OpenAI API consumption.
The proposal on the table was a three year commitment. In a category where the models, the prices and the capabilities all changed materially within twelve months, three years is a long time to promise anything.
BBVA avoided the lock in and cut twenty eight percent from the cost.
The customer profile
One hundred and twenty thousand employees across more than thirty countries, in a sector where data handling terms carry as much weight as price.
Banking adds a constraint most buyers do not face. Where data is processed, how long it is retained, and whether it can be used for training are regulatory questions before they are commercial ones.
That changes the negotiation. Terms you might trade away elsewhere are not tradeable here, which means the commercial flexibility has to come from somewhere else.
The opening position
A three year commitment covering seats and API consumption together, priced against an adoption curve nobody had yet measured.
Long commitments are sold as price certainty. In a fast moving category they are better understood as a transfer of technology risk from the vendor to the buyer.
The blended seat and API structure compounded it, removing the ability to adjust either side independently.
The approach
We separated the two purchases and sized each from evidence. Seats against a segmented addressable population, API against modelled workload volumes.
Then we tested the three year term itself. The pricing advantage of committing for three years turned out to be modest against the flexibility it removed, particularly in a category where a cheaper capable model can appear inside a quarter.
The data terms were negotiated separately and in writing, rather than accepted as platform defaults, which for a regulated institution is not optional.
The eleven moves
- Segment the workforce by role. Document, inbox and meeting intensity decide who is genuinely addressable.
- Separate seats from API. Two purchases with different economics and different failure modes.
- Normalize every quote to cost per task. Token rates across vendors are not comparable.
- Tier workloads by model. Reserve the most capable model for the work that needs it.
- Model caching and batching. Restructuring a workload often beats anything won at the table.
- Commit on the proven baseline. Never on the roadmap, which moves faster than the contract.
- Keep a second provider viable. A working integration, not a stated intention.
- Get price protection in writing. Including what happens when list prices fall.
- Cover model deprecation. Establish your entitlement when a committed model retires.
- Negotiate rate limits alongside price. A cheap rate you cannot consume at peak is not cheap.
- Put data terms in the contract. Retention, training use and residency, not a console setting.
The commercial outcome
- Saving. Twenty eight percent against the opening position.
- Term. The three year lock in avoided, preserving the ability to move as the category changes.
- Structure. Seats and API separated and sized independently.
- Data terms. Retention, residency and training use written into the contract.
- Alternatives. A second provider integration kept working.
In a category this young, the length of the commitment is a bigger decision than the price attached to it. Certainty is worth paying for only when the thing you are certain about is not about to change.
How we engage
- Enterprise AI scoping. A six week engagement that segments the addressable population, separates seats from API, and normalizes vendor quotes to cost per task. GenAI vendor services practice.
- Negotiation. We run the seat and API conversations separately, with price protection, model deprecation cover and data terms on the table. OpenAI enterprise procurement playbook.
- Generative AI procurement. The same discipline across OpenAI, Anthropic, Google and AWS Bedrock. AI platform contract negotiation.
- Vendor Shield. Always on cover across the AI platforms and the wider software estate. Vendor Shield.
- Run the numbers. The software spend assessment sizes AI spend against the wider portfolio.
