Claros Arbiter
Stop deciding important things in Slack threads.
Arbiter gives decisions structure: defined criteria, weighted scoring, collaborative evaluation, and a searchable record of what was decided and why. AI surfaces relevant past decisions and flags contradictions in real time.
Your decisions have no paper trail.
The team chose Vendor A over Vendor B in a Slack thread six months ago. Three people voted with emoji reactions. One person had concerns but didn't push back. Nobody documented the criteria.
Now Vendor A isn't meeting expectations and someone asks: why did we pick them? Nobody remembers the reasoning. The same discussion happens again from scratch, with different people, reaching a different conclusion based on the same incomplete information.
This cycle repeats for technology choices, process changes, hiring decisions, and vendor evaluations. Each decision is made in isolation with no reference to what came before, no structure for evaluation, and no record that survives the Slack retention window.
Step by step
How Arbiter works
Define the decision
Start from a template or from scratch. Describe the context, identify stakeholders, and set the visibility level (public, team, or private). The AI sidebar immediately begins searching for semantically similar past decisions in your organisation.
List options and define criteria
Enumerate the options under evaluation. Define the criteria that matter — cost, implementation time, risk, team capacity, strategic alignment — and assign weights. The criteria and weights become the shared framework the team evaluates against.
Score and evaluate
Each stakeholder scores each option against each criterion. Weighted totals are computed automatically. The highest-scoring option is highlighted. The AI flags if the new decision contradicts any active decisions in the system.
Approve and close
Submit for approval if required. Record the chosen option with a reasoning statement. Set a review date for outcome tracking.
Track outcomes
Record expected vs. actual outcomes. Mark as met, partial, or missed. Over time, the organisation builds a searchable record of decisions and their results — a feedback loop for better future decisions.
Interactive demo
See it in action
A monitoring tool evaluation — three options, four weighted criteria, and an AI sidebar surfacing a past decision that validates the current scoring.
What Arbiter can do
Capabilities
Structured Decision Wizard
Five-step guided process: template selection, context definition, option enumeration, weighted criteria, and per-option scoring with automatic winner highlighting. AI sidebar shows similar past decisions throughout.
Weighted Evaluation Matrices
Criteria with configurable weights scored per option. Weighted totals computed automatically. Transparent evaluation that the whole team can see and challenge.
AI Similarity Search
Vector embeddings power real-time similar-decision retrieval as you draft. Past decisions above 55% similarity are surfaced to prevent duplication and inform current choices.
AI Contradiction Detection
Two-stage pipeline: vector screening finds semantically close active decisions, then LLM analysis identifies genuine logical contradictions. Explanations are validated against candidate IDs to prevent hallucinated references.
Approval Workflow
Submit decisions for approval before activation. Approvers receive notifications and can approve or reject with comments. Rejection blocks activation until resolved.
Outcome Tracking
Record expected vs. actual outcomes for closed decisions. Status tracking (met, partial, missed) builds an organisational learning record.
Risk Register
Formal risk assessment with 5x5 likelihood-impact scoring. Treatment plans with four options (accept, mitigate, transfer, avoid) and a status lifecycle through to verification. Reassessment scheduling with currency tracking.
AI Governance Module
Full AI system lifecycle management: EU AI Act risk classification, model cards with versioned documentation, fundamental rights and DPIA assessments, incident tracking, conformity assessments, and system timelines.
The platform
Part of the Claros platform
Decision context draws from Scribe's knowledge base. Risk assessment currency feeds into the cross-product compliance readiness score.
Strategic improvement recommendations from Cadence analysis can be structured as decisions in Arbiter for team evaluation.
AI governance conformity assessments reference Protocol's compliance evidence. Risk register entries can trigger protocol reviews.
Pending decisions can be included in handoff briefs so incoming team members inherit decision context and upcoming review dates.
Decision review reminders and pending approvals appear as tasks. Upcoming evaluation deadlines are tracked.
Who uses Arbiter
Built for
Engineering Managers
Evaluate technology choices with structure. Reference past decisions so the team doesn't relitigate settled questions every quarter.
CTOs / VPs Engineering
Maintain a searchable record of architectural and tooling decisions across the organisation. Track whether decisions delivered the expected outcomes.
Compliance Officers
Formal risk register with treatment tracking and reassessment scheduling. AI governance module for EU AI Act conformity.
Product Managers
Structure feature prioritisation, vendor evaluations, and resource allocation decisions with transparent criteria the team can inspect.
See Arbiter in action.
Book a 20-minute demo. We'll run a live decision with the scoring matrix and AI similarity search.
Already a customer? Go to Arbiter →