Why you can trust what comes out
Anyone can ask an AI about the future. The difference is whether you can defend the answer. Vorschau is built so every claim traces back to a source, diversity is enforced rather than hoped for, and the system checks its own work, with you in the loop on the decisions that matter. What a run actually does lives on How it works.
The trust moat
Six commitments behind every run
These are what separate auditable AI foresight from a clever prompt and a confident answer. Each one is built into the system, not bolted on afterward.
Source tracking on every claim
Every output carries a traceable line back from strategy → scenario → trend → signal → source. No claim exists without that chain, and fabricated citations fail the check. You can click any conclusion and walk it back to where it came from.
Two sources before it counts
A weak signal isn't promoted until it shows up in at least two independent sources. Single-source claims stay flagged as hints. The two-source rule is cheap, and it keeps rumour-driven scenarios out of your brief.
Diversity enforced, not hoped for
Scenario diversity is enforced by a quality gate, not left to chance. Breadth of signals, depth of interpretation and spread of scenario archetypes are all measured, and a run that comes up shallow is regenerated. AI tends to converge on the obvious. The system pushes it apart.
Built-in challenge, not flattery
A devil's advocate critiques the work as it goes, and a separate red team applies sustained adversarial pressure with its own evidence. Dissent is surfaced and shown to you, not smoothed over into a single confident answer.
You hold the high-leverage decisions
The system pauses at the calls that matter (your question, the scenario axes, the core assumptions, what to commit to) and waits for you. You approve the turning points; the AI handles the legwork in between.
Checked against reality
Every probabilistic statement is stored with the date it will be checked. As the early warning signs resolve, the system scores its own forecast accuracy and tells you which judgments to trust more, or less. Trust is earned against outcomes, not asserted up front.
Data & handling
Commitments, not aspirations
These hold for every session, not policies written after the fact. Every line below is either built into the platform or gates access to it.
- Your data stays on EU-hosted endpoints. Both live providers run in Sweden Central.
- No training on your data: the endpoints are configured for enterprise data handling, so your inputs are never used to train models.
- You can export or delete any session at any time, and you hold every high-leverage decision.
- When its confidence is low, the system says so, including which sources it had to skip, and why.
Beyond the guarantees
Two things most tools don't do
The commitments above are about how the work is checked. These are about what the work actually contains.
Stakeholders, per scenario
Most tools give you one static stakeholder map. We give you one per future.
Power, interests and alliances shift when the world shifts. The platform reruns stakeholder analysis inside every scenario, so you see which actors move with you, which move against you, and which change sides depending on the future.
Depth before prospecting
LLMs default to surface-level futures. The pipeline refuses to let them.
Before any scenario is drafted, every driver is pushed through a depth pass: the headlines, the system beneath, the core assumptions holding it in place, and the deep story keeping those beliefs alive. A run cannot advance without all four layers populated.