Bring one decision.
Leave with evidence.

A good first pilot has a named decision owner, an existing working baseline and a way to check whether a candidate earns adoption. Resource placement, scheduling or a bounded software configuration can fit when the objective and constraints are measurable.

• The decision: what is chosen, how often, and why it matters.
• The baseline: the current rule or system and at least one credible simple alternative.
• The objective: cost, time, quality or another measurable outcome.
• The constraints: limits every answer must respect, including failure cases.
• The evidence: a sanitized frozen workload and an evaluator or checker your team accepts.

Keep the first message non-confidential. We discuss data access, retention and transfer arrangements before any business dataset is shared.

Before work starts, agree on the problem, engine and methods, data access, matched baselines, evaluation count, compute spend, delivery dates, acceptance threshold and handover. The fee and payment terms are quoted for that scope; there is no automatic subscription or unlimited solving entitlement.

The success threshold must be fixed before results are seen. The same input data, constraints and measured budget apply to the comparisons. Demand changes and failure conditions belong in the test plan. A tie or no-go is a legitimate outcome.

The free company showcase supplies fixed synthetic scenarios and aggregate summaries. Your company's task and deployable deliverables are separately agreed paid work.

• An agreed executable or mechanically checkable candidate when one meets the acceptance checks.
• Baseline and stronger-reference comparisons under the agreed conditions.
• Constraint checks, failed tests, limits and results across the accepted evaluation set.
• Reproduction instructions for the delivered artifact and your evaluation harness.
• A written improvement, tie or no-go readout, with the evidence needed for your adoption decision.

The underlying Istimthal engine remains private. Access and rights to the agreed candidate are settled in writing before work starts. A no-go still receives the evaluation evidence; a deployable improvement is not guaranteed.

Your team owns the adoption decision and agrees who handles staging, support, monitoring and rollback. A pilot result alone does not authorize production deployment.

Any future case study will identify its accepted baseline, constraints, test scope, measured outcome and limitations. The customer approves the exact public wording and any attribution. Their name, data and quotations are not published by default.

Our current benchmark is synthetic evidence from the bounded web model. It is distinct from a result on your fleet or business process.