- 100% coverage
- 4 channels read
- Score + passage
- No sampling
Quality Intelligence
- Coverage
- 100%
- Channels read
- 4
- Modules
- 8
- Sampling
- 0
Every interaction read and turned into a score somebody can defend. The reasoning sits with the passage it came from.
REVEAL
Every interaction read, scored against your own rules, and signed so the result can be checked later. The judgement stays with the person who carries it.
[01 / THE PROBLEM]
One percent of your interactions decides what you believe about the other ninety-nine.
Quality assurance in customer operations still works by sample. A reviewer opens a handful of calls per agent per month, fills in a form, and that form becomes the number the business runs on.
SAMPLED · HOW IT WORKS TODAY
- A 300-seat operation handles roughly 150,000 interactions a month. Reviewing five per agent covers 1,500 of them.
- The rare failure, the one that costs you the account, almost never lands in the sample.
- Findings arrive days or weeks after the customer hung up, when nobody can act on them.
- Two reviewers score the same call differently, and neither can show their reasoning.
- The score moves and nobody can say why.
READ · HOW REVEAL WORKS
- All 150,000. Voice, chat, email and social, in the same pipeline.
- Rare failures surface because nothing was skipped.
- Scored as the interaction lands, not at the end of the month.
- One scorecard, applied the same way every time, in every language.
- Every score carries the rule it applied and the evidence it read.
The sample isn't wrong because the reviewers are wrong. It's wrong because it's a sample. Once a machine can read all of it, the sample stops being a method and starts being a limit.
[02 / ARCHITECTURE]
Three layers. One pipeline. Nothing hidden between them.
REVEAL reads an interaction once and passes it down. Each layer writes structured output against a schema, never free text, so the layer below can be built on it.
L1 · ANALYSIS
Eight modules read the interaction.
Root cause, contact driver, quality score, knowledge match, KPI contribution, customer profile, experience signal and social mention. Each returns a typed result with the passage it read. Adopt them in any order.
L2 · INTELLIGENCE
Patterns across interactions, not inside one.
What's moving, what's new, and what's about to show up in a KPI that hasn't moved yet. Trend detection, statistical analysis and the executive view are all built from L1 output, so every number drills back to the interactions behind it.
L3 · ACTION
The part that changes something.
Coaching recommendations, training material and process maps, generated from the intelligence and routed to whoever owns the fix. Flows adds the triggers and channels that carry them: an insight becomes a task, a call, a message, an escalation.
Two systems cut across all three. Channel Adaptations tunes the reading per channel, because a chat transcript and a phone call fail in different ways. The Agent Performance Index rolls the whole chain up to the person who handled the conversation.
[03 / MODULES]
Eight modules. One audited pipeline.
They share a single log, so adding one doesn't disturb the rest. Start with the two that hurt most and grow into the others.
Hand 01
Capabilities
Quality Intelligence · Voice & Flows · Local AI
- Task · call · message
- Escalation trails
- Human approval
- Idempotent retries
Voice & Flows
- Channels out
- 8
- Layers
- 3
- Approval
- Human
- Retries
- Idempotent
An insight becomes a task, a call, a message or an escalation. The trail of who did what stays attached.
Flows deep dive →- On your hardware
- Nothing leaves
- 4 model classes
- From 100 seats
Local AI
- Model classes
- 4
- Data out
- 0
- Runs on
- Your hardware
- From
- 100 seats
Models that run inside your infrastructure, sized to your hardware. Nothing sent out to be understood.
Local deep dive →Hand 02
Analysis
RCA · Top Call Drivers · Quality Monitoring · Knowledge Base
- DSAT / CSAT drivers
- First contact resolution
- Configurable subtype
- Every tier
Root Cause Analysis
- Code
- RCA
- Sub-types
- 4
- Tier
- Every
The driver behind a complaint, a cancellation, an escalation. Four sub-types: DSAT, CSAT, first contact resolution, and one you configure.
In every tier →- Live ranking
- Contact classification
- Forecast signal
- SMI deep dive
Top Call Drivers
- Code
- TCD
- Ranked
- Live
- Deep dive
- SMI
What customers are contacting you about, classified and ranked, before the volume shows up in next month's forecast.
SMI deep dive →- COPC-aligned
- Your scorecard
- 100% coverage
- Live calibration
Quality Monitoring
- Code
- QA
- Aligned
- COPC
- Coverage
- 100%
COPC-aligned scoring on every interaction, against your scorecard rather than a generic one. Calibration stops being a meeting.
QA deep dive →- SOPs & frameworks
- Source citation
- Answerable corpus
- Every tier
Knowledge Base
- Code
- KB
- Cited
- Source
- Tier
- Every
Your processes, SOPs, frameworks and reference material, ingested and answerable with the source passage cited.
In every tier →Hand 03
Signal & action
KPI Engine · Customer Profiling · CXE · Social Media Intelligence
- From interactions
- Targets & alerts
- Movement reasons
- Every tier
KPI Engine
- Code
- KPI
- From
- Interactions
- Tier
- Every
Contractual and operational indicators computed from the interactions themselves. Targets, alerts, and the reason the number moved.
In every tier →- Segment & effort
- Emotional state
- Value tier
- Lifecycle stage
Customer Profiling
- Code
- CP
- Per
- Interaction
- Tier
- Every
Who you were actually talking to: segment, effort, emotional state, value tier and lifecycle stage, per interaction.
In every tier →- Turn-by-turn read
- Sentiment & effort
- Brand voice
- Next best action
Customer Experience Engagement
- Code
- CXE
- Grain
- Turn by turn
- Tier
- Every
Sentiment, effort and emotion read turn by turn, with brand voice and the next best action for the turn after.
In every tier →- Reviews & posts
- Brand mentions
- Same voice pipeline
- One signal axis
Social Media Intelligence
- Code
- SMI
- Pipeline
- Same as voice
- Deep dive
- SMI
Reviews, posts and brand mentions read in the same pipeline as the calls, so public and private signal sit on one axis.
SMI deep dive →[04 / THE CHAIN]
Every output signs itself.
An AI score you can't check is an opinion. REVEAL writes five things with every result, and hashes them together so the record can't be edited after the fact.
-
01 · INPUT
The interaction as it arrived. Channel, language, timestamps, and the redactions applied before anything read it.
-
02 · RULE
The policy line or scorecard item the output is judged against, quoted in full rather than summarised.
-
03 · MODEL
Which model ran, at which version, with which prompt template. Swap the model and the record says so.
-
04 · SIGNATURE
A hash over all of the above, linked to the entry before it. Change one field and the chain breaks visibly.
-
05 · EXPORT
The bundle leaves as a file. Verify it on your own machine, offline, without asking us for anything.
The chain is what Constitution A1 to A8 is signed against. A3 forbids fabrication, A4 requires the reasoning to be readable, A7 marks every AI-produced line as AI-produced. They're rules the platform enforces at the log layer, which is the only place a rule survives contact with a deadline.
[05 / DEPLOYMENT]
Two ways to run it. Same modules, same chain.
Where REVEAL runs is a procurement question, not a product one. Nothing about the platform changes between them.
CLOUD
Hosted in the EU, running the day we ship.
EU region, EU processing, GDPR terms written for a European buyer rather than translated for one. You get each release as it lands, and the interaction caps scale with the tier.
PER AGENT · PER MONTH · PRICED ON YOUR NUMBERS
LOCAL
Inside your infrastructure, on your hardware.
No model weights leave your environment. No interaction text crosses a public network. Same eight modules, same audit chain, same console, running where your security team can point at it. From 100 seats.
FROM 100 SEATS · PRICED ON YOUR NUMBERS
[06 / THE MODEL]
Four models. Yours follows from your shape.
Per agent, per month, in euros · the rate falls as the operation grows, and every model includes the L1 analysis modules. Which one you land on is a fact of your seats and volume, not a negotiation. The price is worked out on a call, with your own numbers on the table.
STARTER
- SEATS
- 10 to 25
- INTERACTIONS
- 10,000 per month
- LAYERS
- L1 analysis
PROFESSIONAL
- SEATS
- 25 to 100
- INTERACTIONS
- 50,000 per month
- LAYERS
- L1 · L2 intelligence
BUSINESS
- SEATS
- 100 to 500
- INTERACTIONS
- 150,000 per month
- LAYERS
- L1 · L2 · L3 action
ENTERPRISE
- SEATS
- 500 and up
- INTERACTIONS
- Uncapped
- LAYERS
- All, plus the DMAIC suite
REVEAL Local runs in your stack from 100 seats · its own model, same rule. Volume and multi-year terms move the effective rate down. No renewal escalation, no add-on stacking.
Take the exercise → The coverage calculator takes your seat count, your volume and your QA sample, shows what the sample misses today, and resolves which model your operation lands on. Then arrange the call · thirty minutes, your own numbers on the table, and you leave with the price.
[07 / CONSOLE]
Open the console.Audit the chain yourself.
Operator, auditor and executive. Three views onto the same signed chain, running the instance our clients run. No login, no sales call.
quality visible.