A constrained API targets classification and routing
OpenAI has moved the Decisions API into general availability after a public beta. The service is designed for applications that need to check conditions, choose from fixed options or score text, images and JSON against a rubric. Because it does not generate a full response, it can support classification, moderation, routing and evaluation tasks where a bounded answer is more useful than open-ended prose.
The distinction can simplify production systems. Developers can define questions and permitted outputs rather than parsing an unconstrained model response. That reduces some formatting risk, but it does not make the underlying judgement deterministic. Ambiguous evidence, biased rubrics or distribution shifts can still produce a valid-looking option that is wrong for the real situation.
The general release expands input and scale
The general release adds native GPT-Live integration so voice applications can react while people speak. Prompt caching lowers cost for repeated context, native JSON accepts structured application data, and individual questions can include images for comparison against a shared reference. A request can now contain up to 400 questions, twice the previous limit reported in the announcement.
OpenAI lists GPT-6 Luna pricing at US$0.10 per million non-cached input tokens and US$0.01 per million cached input tokens, with no output-token charge. Low unit cost may encourage teams to apply the API at high volume. Budgets should therefore follow business events and completed decisions, not only token prices, and cached context needs the same privacy and lifecycle controls as uncached input.
Fixed choices still require governance
A decision endpoint can sit directly in consequential workflows, such as prioritising cases, approving content or choosing the next agent action. Teams should define abstention and escalation options rather than forcing every input into one of several choices. Thresholds need calibration against false positives and false negatives, with separate measures for groups or conditions where errors have different consequences.
General availability signals operational support, not universal fitness. Before replacing rules or human review, organisations should replay historical cases, inspect disagreement and test adversarial inputs. Logs should preserve the rubric, model version, evidence and selected option. The API is most valuable when it makes a narrow decision faster while keeping the reasons, limits and accountable owner visible.
What teams should validate before production
For implementation, teams should record the exact model, software, dataset, connector and policy versions behind every result. Representative evaluation must include normal work, boundary cases, deliberate misuse and component failure, with outcomes separated by user role, data sensitivity and consequence. A strong average should never hide a serious failure in a critical path. Changes should pass the same regression set before promotion, and operators need a documented fallback that preserves evidence rather than silently broadening access or lowering quality.
Procurement and planning should account for the human work required to review exceptions, maintain controls and investigate incidents. Published performance or efficiency claims are useful starting points, but production confidence comes from evidence collected in the organisation’s own environment. A named owner, measurable acceptance threshold, cost model and tested shutdown path turn a promising release into a service that can be governed over time. Where decisions affect safety, rights, money or regulated records, accountable professional review remains essential.
Security and operational evidence
Security review should map every identity, credential, network destination, stored artefact and administrative access path involved in the workflow. Logs need enough context to reconstruct who initiated an action, which evidence was consulted, what the system proposed and what ultimately executed. Sensitive prompts and traces should not be retained merely because they are useful for debugging. Retention, redaction and access must follow the underlying data classification, with a practical process for correcting or removing material when policy or law requires it.
Operational monitoring should measure quality and service behaviour together. Latency, throughput and availability matter, but so do unsupported claims, missed constraints, unsafe tool calls and the rate at which people overturn recommendations. Teams should rehearse dependency outages and partial failures before unattended operation. If a model, connector or external service becomes unavailable, the workflow should fail predictably, preserve its state and tell the user what remains unfinished. Clear degradation is safer than an agent improvising around a missing control.
Transparency should scale with adoption
Organisations should explain the scope and limitations of the deployed capability to the people affected by it. Users need to know when automation is involved, how to challenge an outcome and where a human can intervene. That transparency improves feedback and prevents a pilot result from being interpreted as blanket approval for broader autonomy. Expansion should follow demonstrated evidence, not novelty or pressure to match a competitor announcement. The durable advantage comes from dependable integration, careful measurement and learning that survives the next model or product update.
Finally, teams should schedule a review after real usage rather than treating launch as the end of governance. The review should compare expected benefits with observed quality, cost, incidents and user behaviour, then retire workflows that do not justify their complexity. A service that works well can expand through the same evidence-based process. A service that fails should be narrowed or stopped without losing the records needed to understand why. This feedback loop is what turns experimentation into accountable operational capability.