Skip to main content
Cavefish
Communication Risk Intelligence

Every major decision
has a number.
Except this one.

Cavefish measures communication risk — whether what you're saying is actually believed by the board, the market, your customers, and your own people — before the disconnect becomes a business problem.

The only UK-built communication risk intelligence platform with ICO registration, cultural calibration and enterprise governance.

ICO registered · UK GDPR compliant · FCA Regulatory Sandbox

13 industry awards since 2023

Communication Risk Report — Live Output
Earnings Call Rehearsal · Q2 2026
73/ 100
Credibility Signal↓ 4:22
0:004:22 — drop8:00
Communication Risk Score73 / 100
Delivery Drop↓ at 4:22
Engagement RiskElevated
Audience AttentionHigh — 87%
Coaching NotesAI-generated
AI Coaching Note

Slow delivery pace by 15% at line 47. Remove upward inflection on "expect." Repeat in rehearsal until Trust Score at 4:22 reaches 80+.

ECSO Cyber Solution Days — Winner
Fintech Awards — Best Use of AI
Cardiff Life Awards — Winner
IoD Director of the Year — Shortlisted
UK Technology Awards — Shortlisted
FCA Regulatory Sandbox
13 Awards Since 2023
ICO Registered ZB915633
44 Action Units · FACS Standard
14 Cultural Cohorts · 6 Countries
Anthropic
OpenAI
Mistral
AI Wales CIC
Swansea University
ECSO Cyber Solution Days — Winner
Fintech Awards — Best Use of AI
Cardiff Life Awards — Winner
IoD Director of the Year — Shortlisted
UK Technology Awards — Shortlisted
FCA Regulatory Sandbox
13 Awards Since 2023
ICO Registered ZB915633
44 Action Units · FACS Standard
14 Cultural Cohorts · 6 Countries
Anthropic
OpenAI
Mistral
AI Wales CIC
Swansea University
0
Facial Action Units
0
Cultural Cohorts
0
Industry Awards
0
Turnaround
Market Position

The only UK-built communication risk intelligence platform with ICO registration and a full governance layer — built for enterprise teams.

01
UK-built
Cardiff, Wales · Welsh Government backed
02
ICO registered
ZB915633 · Verifiable at ico.org.uk
03
Full governance layer
LIA · DPIA · DPA standard · FCA sandbox
Four claims. Four proof points. →Every claim on this page is independently verifiable.

The Problem

Gut feel has never been measured.

It decides hires, deals and boardroom votes — yet it appears in no dashboard, no board pack and no risk register. Every other input to a major decision carries a number: cost, risk, headcount, runway. There is no standard metric for Communication Risk — the probability that a well-designed decision fails because it isn't believed.

📊
Measured
Cost, risk, headcount, runway, market size
Not measured
How the announcement will land. How the pitch will be received. Where trust breaks.

This is not sentiment analysis. Sentiment tells you what people said after the fact. Cavefish tells you whether it will be believed — before you deliver it.

VAD
V
VALENCE
A
AROUSAL
D
DOMINANCE

The Science Behind the Score

44 Action Units.
One number you
can act on.

Cavefish analyses communication delivery using the FACS framework — 44 facial Action Units calibrated across 14 cultural cohorts in 6 countries. The result is a Communication Risk Score that is objective, reproducible and auditable.

Congruence scoring: Cavefish reads face, voice and text simultaneously — measuring the gap between what is said and what is felt. Single-channel systems cannot detect this; the gap only exists across channels.

  • FACS-standard 44-Action-Unit decoder
  • Calibrated across 14 cultures in 6 countries
  • Timestamped per-second Credibility Signal
  • Sub-200ms inference latency
  • On-device processing available
  • UK data residency · GDPR compliant · ICO ZB915633
Full technical explanation →
🔒
Per-Client Small Language Model

Each enterprise deployment receives a dedicated SLM, trained and continuously tuned on your organisation's own communication data — within your security boundary. The model becomes progressively more accurate for your people, culture and communication patterns.

Process

Three steps. No specialist equipment.

01
📤

Send what you already have

A video recording, audio file, text transcript or image. Zoom call, earnings rehearsal, sales demo, campaign creative — any format, shared via a link. No installation. No hardware.

02
🧠

Cavefish analyses the signals

Linguistic credibility mapping and hedging detection for text. Vocal pitch, pace and hesitation markers for audio. FACS-aligned delivery pattern analysis for video. All modalities, all in one report.

03
📋

Receive a scored, timestamped report

Trust Score 0–100, Credibility Signal timeline, key moment identification with timestamps, specific coaching recommendations. PDF format. Within 5 working days. Board-ready.

Sample Output · Earnings Call

What you receive in the report

01
Communication Risk Score 0–100
Composite delivery effectiveness across all modalities
02
Delivery timeline
Per-second — where engagement dropped and why
03
Key moment identification
Timestamps for intervention and coaching
04
AI coaching notes
Specific, written, actionable recommendations
05
PDF report
Shareable, board-ready, auditable — see sample
73RISK SCORE
AI Coaching Note

What's working: Strong authority projection (0:00–0:45).
Where it drops: Guidance delivery loses confidence at 4:22. Remove upward inflection on "expect."

Download a sample report →

FAQ

Questions before you reach out

A structured PDF with Communication Confidence Score (0–100), Delivery Effectiveness timeline (per-second), key moment timestamps, and specific coaching recommendations. Where relevant: vocal hesitation analysis, linguistic pattern mapping and an Engagement Signal Map.
Text and scripts (PDF, DOCX, TXT), audio (MP3, WAV), video (MP4, MOV, Zoom recording) and images (JPG, PNG). Share via Google Drive, Dropbox or WeTransfer. Most organisations start with a script or audio recording.
Both. Cavefish can be deployed as an ongoing software integration (API or dashboard) or delivered as a consultancy-led analysis. Most organisations start with a scoping consultation, then discuss which deployment model suits their use case.
Cavefish uses the FACS standard — 44 Action Units calibrated across 14 cultural and linguistic cohorts in 6 countries. It measures observable delivery signals, not guesses about individuals. Full accuracy and methodology documentation is available on the Proof page.
Yes. Cavefish is ICO registered (ZB915633). All analysis requires explicit informed consent. Data is processed under UK GDPR with a signed DPA. On-premise deployment is available for regulated environments.
We begin with a scoping consultation to understand your use case. Pilot programmes start from £1,500/month, typically covering 10–20 analyses with dedicated analyst support. Enterprise licensing, API access and on-premise deployment are priced on scope.

Data Sovereignty & Deployment

On-premise, UK sovereign, EU sovereign

For regulated environments and data-sensitive deployments, Cavefish can be deployed to meet your specific data residency and sovereignty requirements.

🇬🇧
UK Sovereign
Data processed exclusively within UK jurisdiction
🇪🇺
EU Sovereign
GDPR-compliant EU data residency
🏢
On-Premise
Deployed in your own infrastructure
☁️
Private Cloud
Dedicated instance, your security controls
Learn about regulated deployment options →

Start Here

Give your next decision
a number.

Cavefish works with leadership teams to measure communication risk across the moments that matter — board updates, campaigns, and customer conversations. Book a consultation to discuss your use case.

ICO registered ZB915633 · UK GDPR compliant · hello@cavefish.ai