Name the operating constraint.
Management brings a workflow where time, cost, speed, quality, or capacity is materially holding the business back.
One possible engagement shape, agreed after consultation. Scope and commercial terms depend on the work.
The BGA AI Outcome Contract
If the agreed operating result does not work, you do not pay the outcome fee.
Choose one expensive, measurable workflow. We agree what “working” means, build the solution, and have up to 12 months to get it accepted.
ScopeOne measurable operating result
PaymentAfter agreed acceptance
Delivery windowUp to 12 months
Most AI projects begin with what the technology can do. We begin with one job the business should no longer do the same way.
A pricing team might spend most of its week collecting competitor data, checking rules, and preparing routine changes. Management knows the work should be lighter—but a prototype is not the same as an operating result.
A better management question
“Can this workload move from eight full-time equivalents to three—without losing control?”
That is concrete enough to baseline, design around, test, and put into a contract. It also keeps the human outcome clear: five people’s worth of capacity can move to supplier deals, inventory quality, analysis, or other higher-value work.
Risk reversal, with a real acceptance test
The commercial model aligns us around the same finish line.
No vague transformation program. Before the build starts, both sides can see the baseline, scope, acceptance criteria, and fixed outcome fee.
Baseline the work.
Together we document the current workload, volumes, exceptions, quality controls, dependencies, and what the team does with the capacity today.
Write “working” into the contract.
We define the measurable end state and the acceptance test: what must be true, under which conditions, and how both parties will verify it.
Fix the outcome fee.
The fee is linked to roughly one year of the agreed annual value released by the result. It is a known investment—not an open-ended transformation budget.
Pay when the result is accepted.
We have up to 12 months to deliver. If the contracted result misses its acceptance test within that window, the outcome fee is not due.
The objective is not “install AI.” It is to make the same pricing operation dependable with five fewer FTEs of repetitive workload.
Illustrative capacity modelAcceptance criteria agreed before build
8FTEs of current workload
3FTEs of steady-state workload
What changes
The routine collection, checking, and price preparation moves into a controlled system.
What stays human
Exceptions, commercial judgment, supplier conversations, and rule ownership stay with the team.
What value means
The five-FTE capacity difference becomes the basis for agreeing annual value and the fixed fee.
If the system is accepted after three months, we do not disappear. We support, tune, document, and stabilize it through the remaining agreed support period. Earlier delivery gives the business more time to realize value—and gives us more time to make the change stick.
What makes this credible
The hard part is not the model. It is the operating change around it.
The baseline is shared.
No savings claim without agreed starting volumes, workload, controls, and assumptions.
Acceptance is observable.
The result must hold under defined real operating conditions—not only in a demonstration.
The team remains in control.
Exceptions, overrides, permissions, auditability, and ownership are part of the solution.
Support is part of delivery.
An early result is refined and stabilized through the remaining contracted support period.
Production changes the conversation. A clever demo becomes useful only when it survives bad inputs, edge cases, economics, and Monday morning.
BGA’s team has built and operated pricing, advertising, creative, analytics, and AI systems inside a multi-market commerce business. That experience shapes the acceptance criteria as much as it shapes the software.
A relevant proof point
AI economics redesigned around the workload.
Before
A cloud API workflow cost roughly €800 each time it ran across the full catalog.
After
Observed hardware payback: about one month.
The right workload moved to local hardware with roughly electricity-only run cost.
This case illustrates BGA’s operating approach, not a forecast for another company. Every Outcome Contract uses its own baseline, scope, dependencies, value model, and acceptance criteria.
Is this the right shape of engagement?
The offer is strongest when the result can be measured and management can sponsor the change.
We will say no when the target is too vague, the data cannot support it, or the organization cannot give the workflow owner enough room to change.
A repeated workflow with a visible cost.
- Meaningful manual workload or delay
- A stable enough baseline to compare
- Management sponsor and workflow owner
- Access to the required data and users
- An outcome valuable enough to justify change
A technology idea without an operating owner.
- “Use AI somewhere” as the brief
- No baseline or measurable acceptance test
- A one-off task with little repeat value
- No access to the underlying workflow
- Unresolved compliance or data permissions
The questions management should ask
Plain terms before any build.
A fixed outcome fee tied to one precisely scoped and accepted operating result. The value model, payment trigger, support period, responsibilities, and any excluded third-party costs are written into the final contract.
If the contracted result does not pass the agreed acceptance test inside the delivery window, the outcome fee is not due. The final agreement will state treatment of any separately approved pass-through infrastructure or third-party costs.
Not necessarily. The contract is about releasing capacity. Management decides whether that capacity moves to growth work, quality, service, inventory, supplier relationships, natural attrition, or another priority.
Operational acceptance takes more than a technical build. The window allows time for discovery, implementation, real-world validation, training, controls, and stabilization. If acceptance happens earlier, the remaining agreed period is used for support and refinement.
No. We need a credible baseline, meaningful recurring value, management sponsorship, access to the work and data, and an outcome we can responsibly influence. The first working session is designed to test those conditions.
This page describes the intended commercial concept, not a binding offer. Scope, baseline, fee, acceptance criteria, dependencies, support, data protection, third-party costs, and termination terms must be agreed in a signed contract.
Bring the workflow, not a brief
Let us test whether the outcome can be contracted.
In 45 minutes, we will map the job, pressure-test the baseline, and decide whether there is a measurable result worth pursuing. No generic AI roadmap required.