sovrGPT Decision Engine · Private Preview

AI decisions instead of long answers.

Classify, prioritise and route: the sovrGPT Decision Engine evaluates texts and system states against your criteria. You get structured results for your software – with clear rules for automation and human review.

Request pilot access Technical documentation

An example

Incoming message

“Since this morning I cannot log in. My September invoice was also charged twice – please check this urgently.”

Evaluation

  • Team → support 65 % · unclear 31 % · billing 4 %
  • Urgency → level 3 of 4 (urgent)
  • Human review? → yes, 99 %
  • Recommendation: review – no automatic reply

A real measurement from 2026-09-20. Four questions, one state, 397 milliseconds – and not a single generated line of text.

Three formats

Choice

One of up to 26 options – team, category, next action. With a probability for every option.

Yes/no

Does a statement hold? Review needed, complaint, escalation – as a probability, not a gut feeling.

Ordered score

One level of a scale: urgency, relevance, quality. With an expected value between the levels.

Why it is fast

A chat answer is produced token by token. A decision needs a single forward pass per question: the model reads the state and the options, we read off how likely each option is. Several questions about the same state run at the same time.

Measured on 2026-09-20, warm, ticket text with four options:

A chat answer to the same text takes 5–30 seconds. These are our measurements, not a guarantee.

One API. Controlled operation.

  1. 1Your application
  2. 2sovrGPT API – identity, tenant, scopes, limits
  3. 3Approved infrastructure in Germany or the EU
  4. 4Typed evaluation back – your workflow decides

The Decision Engine is being developed for operation on approved European or your own infrastructure. Data region, available models and operating conditions are stated for every approved deployment. A third-party model API call is not part of the core path.

We develop our own specialised decision models on top of open base models. Task-specific training, calibration and documented tests are meant to turn a general model capability into a reliably usable building block for concrete workflows.

Transparency

Even a cleanly structured AI output can be wrong. That is why the service states the model version and calibration status and can hand uncertain cases over for review. Binding approvals and access rights stay with your application.

State, questions and answers are not stored. Billing counts tokens. The processing location is part of every response.

Operating models

Decision API

Shared, approved inference pool. API access, quotas, documentation and playground. Now in private preview.

Dedicated Decision

Dedicated operation with a contractually defined region and performance. On request.

Custom Decision

Data preparation, customer-specific training, evaluation and model maintenance for your tasks. On request.

For developers

OpenAI-compatible authentication, one endpoint, JSON in, JSON out.

curl https://sovrgpt.com/api/v1/decisions \
  -H "Authorization: Bearer $SOVR_KEY" -H "content-type: application/json" \
  -d '{ "preset": "ticket-routing-v1", "state": "…" }'
Go to the API documentation

Private preview. Prices on request; limits and model versions may change during the preview.

sovrGPT Decision Engine – AI decisions instead of long answers