AI that argues
with itself.
Most AI tools generate one answer. GPS Intelligence generates many — and then tries to prove each one wrong. Only what cannot be falsified is surfaced to decision-makers.
This is not a safety feature. It is the core reasoning architecture.
HOW INTELLIGENCE REASONS
Three steps. No shortcuts.
Generate N hypotheses
When a question is raised — what caused this incident, where is the risk, what should the CIO do — Intelligence generates multiple competing explanations simultaneously. No single hypothesis is favoured. All are treated as equally plausible until evidence distinguishes them.
Falsify each one
Every hypothesis is paired with a set of conditions under which it would be false. Intelligence then evaluates each condition against live platform data: discovery signals, health scores, service relationships, historical patterns. Hypotheses that cannot survive the falsifier are marked refuted — not archived or deprioritised, but formally eliminated.
Surface only survivors
Only hypotheses that pass every falsification condition reach the surface. Each survivor is annotated with the confidence band earned by its surviving conditions, the assumed facts that underpin it, and the data lineage that supports it. Decision-makers receive conclusions — not raw AI output.
CAPABILITIES
Six properties. All structural.
Intelligence is not a module you bolt on. These properties are embedded in how the platform reasons — every product that uses Intelligence inherits all of them.
Competing Hypotheses Engine
Multiple explanations generated and evaluated simultaneously. The engine does not rank by plausibility — it eliminates by falsification. What remains is what can be defended.
Falsifier Framework
Every AI assertion must carry a defined falsification condition before it can be asserted. This is a structural property of how Intelligence reasons — not a post-hoc filter applied to softened output.
Confidence Bands
Confidence is expressed as a band, not a point. A hypothesis at 87% means: given all current signals, this is consistent with the evidence at 87% confidence — and these are the assumed facts that bound that number.
Assumed Facts Audit
Every conclusion is preceded by an explicit list of what the system assumed to be true. Assumptions are surfaced — not hidden in a model. If an assumption is wrong, the affected conclusions are automatically invalidated.
Persona-Aware Narratives
The same underlying conclusion is expressed differently for a CIO, CFO, COO, or CRO. Intelligence understands what each persona needs to make a decision — and generates language calibrated to their authority and risk tolerance.
Cross-Product Synthesis
Intelligence draws simultaneously from Pathfinder's discovery signals, Bearing's health data, Contour's service map, and Vantage's incident context. The synthesis is coherent because the underlying data lives on one fabric.
WHY IT MATTERS
The difference is structural.
Other AI systems give you answers. Intelligence gives you answers you can defend in front of a board, an auditor, or a regulator.
- —Generates one answer
- —Confidence is a percentage with no derivation
- —Assumptions are hidden inside the model
- —Cannot be audited or appealed
- —Output cannot be falsified after the fact
- ✓Generates N competing hypotheses
- ✓Confidence is a band with explicit assumed facts
- ✓Every assumption is surfaced and auditable
- ✓Conclusions carry data lineage
- ✓Every claim has a defined falsification condition
POWERED BY INTELLIGENCE
Three products. One reasoning engine.
Intelligence evaluates competing explanations for score degradation — distinguishing genuine risk from noise before surfacing a recommendation.
Learn more →Competing hypotheses about what caused the incident are evaluated and falsified in real time, converging on a confirmed root cause rather than a ranked guess list.
Learn more →Persona-filtered narratives are generated from surviving hypotheses, ensuring each executive receives a conclusion calibrated to their decision authority and context.
Learn more →Most platforms ask you
what to measure.
GPS already knows — because it watched. Every indicator in the platform is grounded in live discovery data, not manual input. The questionnaire is gone. The scorecard arrived on day one.
WHY INDICATORS AREN'T HARD IN GPS
The questionnaire
vanished.
Traditional CMDB health programs begin with a data collection campaign — 47-question surveys sent to application owners who are too busy to answer them. The result is stale data, disputed ownership, and a model that starts broken.
GPS doesn't ask. Pathfinder discovery populates the configuration model directly from your environment. Every application, service, capability, and offering is observed — not reported. Indicators inherit that grounding automatically.
Q1 OF 47
Every indicator traces back through intelligence to discovery. No manual input required. The stack is the guarantee.
ARCHITECTURE BUILT ON INTELLIGENCE
Measured in reality.
Not in surveys.
Every indicator in GPS sits at the top of a grounded stack. Below it: an intelligence layer that derives patterns from live signals. Below that: Pathfinder discovery, continuously populating the ServiceNow fabric.
This isn't a reporting layer bolted onto a CMDB — it's a measurement system that is architecturally connected to the thing it measures. Change propagates upward. Indicators update automatically. The stack is the guarantee.
DAY-ONE INDICATORS
Four domains.
Lit on arrival.
GPS ships with indicators pre-built across the four architectural domains that matter most: Business Applications, Capabilities, Services, and Offerings. They activate the moment discovery runs — no configuration sprint required.
NEED MORE? SIMPLICITY AND BREADTH
When defaults
aren't enough.
Not every organisation needs 40 indicators. Some need 400. GPS offers five levels of indicator breadth — from the defaults that ship with the platform, to AI-assisted authoring that proposes what to measure before you've asked.
Each level builds on the grounded stack. You never leave the safety of discovery-grounded data regardless of how deep you go. Complexity scales with intent — not with consultant hours.
THE RESULT
One lens for
collective decisions.
Four signal streams — Reality, Intelligence, The Lens, Decision — converge into a single view that every stakeholder can read. CIOs see strategy. Architects see design. Engineers see debt. Boards see risk.
The compass doesn't change. The signal doesn't shift. Decisions made from a shared indicator system don't collide in the room — because they started from the same ground truth.
Read the full brief.
The complete “Indicators, Grounded in Reality” document — how GPS discovery grounds every indicator, the authoring model, and the Signal-to-Decision Ribbon explained in full.
See Intelligence in a live environment.
Book a demo and we will run a live falsification cycle against your own environment data. Watch what gets refuted — and what survives.