Talk to us
Menu
Talk to us
Services
AdvisoryBuying AIProcurement strategyProcurement planSpend analysisStrategic sourcingSupplier managementContract managementCompliance and processProcess improvementProcurement technologyRisk managementExecutionCategory managementRFP and tender developmentSupplier diversitySustainability initiativesTraining and developmentProcurement advisoryImplementationOngoing supportStart hereAll servicesBuying AIAI tools
AI tools
Live productsProcurement Maturity Assessment Procurement Skills Assessment On this siteHow the tools work
ClientsCase studiesOur approachBlog

Buying AI: procurement advisory and execution

Buying AI is not like buying software. It is priced per token, per seat, per GPU hour or as a commitment against a cloud agreement; the same model is sold through several channels at different prices; and the product you sign for may be replaced within months. We help organisations decide what to buy, run the sourcing, and sit at the negotiating table with them.

The problem this solves

Demand arrives before the strategy. Departments run pilots with different providers, assistant licences are bought one team at a time, and usage-based charges appear on cloud invoices nobody owns. By the time procurement is asked, the organisation is already committed to overlapping tools on terms nobody negotiated.

The commercial models are unfamiliar. Pricing per token, committed spend, reserved capacity and drawdown against an existing cloud agreement all behave differently as usage grows, while list prices keep falling. A deal that looks generous in its first year can be expensive in its second, and few buyers have a benchmark to test it against.

The contract terms that matter are new. Whether your data can be used to train a model, where it is processed, what happens when a model version is retired, who owns the outputs and who carries the risk if they infringe someone's rights: standard IT templates rarely cover any of it, and the vendor's paper favours the vendor.

What we actually do

The steps, in the order we run them.

How the engagement runs1Map the demand and thespend2Set the buyingstrategy3Test the marketproperly4Negotiate thecommercials5Close the contractterms6Govern the consumption

  1. Map the demand and the spend

    Every live AI use case, pilot, subscription and cloud line item, with its business owner, the data it touches and what it costs today. Most organisations find overlapping tools and spend they did not know they had, and that alone pays for the exercise.

  2. Set the buying strategy

    What to standardise and what to keep open; when to buy direct from a model provider, through a cloud marketplace or via a partner; and how much to commit, given how quickly prices and models change. Aligned with your AI policy, data classification and security requirements.

  3. Test the market properly

    Provider and partner long lists, requests for proposal written for AI rather than adapted from software templates, and evaluations run on your own use cases and data, including Arabic language performance where it matters, rather than on the vendor's benchmarks.

  4. Negotiate the commercials

    Volume tiers, commitments that can move between models, price protection as list prices fall, credits for pilots, rate limits and capacity guarantees. We build the position, model the scenarios on your volumes and sit at the table with your team.

  5. Close the contract terms

    Data use and retention, processing location, model changes and notice periods, ownership of outputs, intellectual property indemnities, service levels, security, audit rights and exit. Worked through with your legal team so the commercial intent survives the drafting.

  6. Govern the consumption

    Budgets, quotas and usage reporting by department, so AI is managed like any other category rather than discovered on the invoice. We set it up, your team runs it.

What you get

We build AI tools of our own: the procurement maturity and skills assessments on this site run on an AI engine we developed, so we buy model capacity and deal with AI pricing and terms from the buyer's side ourselves. That sits on fourteen years of sourcing and negotiating for government bodies and large enterprises in the Gulf.

Four ways to buy AI, by commitment and controlOpen models on demandOpen models on hourlycapacity. You run them.Reserved capacityCommitted or dedicatedcapacity. Lowest unit cost.Pay as you goPer-token access. Fasteststart, priciest at scale.Seats and enterprise plansPer-user plans. Predictable,little room to move.CommitmentControl over data and models
Most organisations need more than one route. The skill is knowing which spend belongs where, and moving it as volumes grow.

Questions we get asked

What counts as buying AI?

Access to large language models and other AI models, whether through an API, a cloud provider or an enterprise subscription; AI assistants and features sold per seat inside software you already use; the cloud and GPU capacity to run models yourself; data services; and the partners who implement and support all of it. Each has its own commercial model, and most organisations are buying several at once.

Should we buy from the model provider or through our cloud provider?

It depends on what you already have. Buying through an existing cloud agreement can count towards commitments you have already made and keeps billing and security in one place; buying direct can bring earlier access to new models and different terms. We model both on your volumes before you choose, and avoid locking the decision in for longer than the market justifies.

How do you negotiate when AI prices keep falling?

By not fixing what the market is about to move. Shorter terms, or commitments that can move between models; price protection that follows published price reductions; benchmarking rights; and ramp schedules that match real adoption rather than the vendor's forecast.

Do you replace our IT and data teams?

No. They own the architecture, security and the technical evaluation of models. We own the market view, the commercial structure, the sourcing process and the negotiation, and we make sure their requirements end up in the contract.

Are you tied to any AI provider?

No. We do not resell AI models or cloud capacity, so the only deal we have an interest in is the one you sign.

Can you work with our procurement team rather than instead of it?

That is how we prefer to work. Your team runs the process with us, and keeps the templates, the evaluation framework and the governance model when we leave.

Where to start

Run the free procurement maturity diagnostic for a scored view of where the gaps are, or talk to a consultant about this specific problem. Both take less time than a meeting about having a meeting.

Talk to a consultant

Other services