Group Intelligence

Every group has a power structure.
Export it. Read it.

Team chats. WhatsApp groups. Deal rooms. Communities. Meeting rooms. The dynamics between the people that matter are already in the conversation — you just couldn't read them. Until now.

Privacy as architecture · Without ever reading your messages · Learn how →
A new observable

Groups are not just people.
They are systems with structure.

Every group conversation has an underlying dynamic — who pushes, who resists, who aligns, who couples with whom, where friction lives, what the group is becoming. This structure exists whether you see it or not. Group Intelligence makes it visible.

Before

You read hundreds of messages, form an intuition, and act on a feeling.

The gap

The structure was always there. You just had no instrument to read it.

After Hermion

Export. Assign. Read the dynamics — without reading a single message.

How it works

Three steps. Any group. Any platform.

01

Export any group conversation

WhatsApp group, Slack channel, Discord server, meeting transcript with 3+ speakers, or any format you can export. Upload to the web app, or pipe through the API if you're automating.

02

Assign the actors that matter

Name your two groups. Drag speakers into sides. Anyone not assigned is silenced — stripped from the analysis entirely. A 50-person group becomes a 2-faction analysis the moment you assign the people that matter.

03

Read the dynamics

Interaction state, resistance vectors, who is investing more, cross-actor patterns, what is working, optimal next move — and more. Framed around your stated goal: deal, professional, sales, whatever the interaction is for.

For developers — automate entirely

Self-host the open source classifier. Convert group exports to Hermion signals on your own servers — messages never leave your infrastructure. Group actors programmatically before sending. Call the API. Get intelligence on your schedule, automatically.

Run quarterly team dynamics reports. Trigger analysis after every deal room session. Monitor community health automatically.

Read the API docs →
The core capability

Focus on who matters.
Silence the rest.

Group Intelligence is not about reading the whole group. It is about isolating the dynamics between the actors that actually determine the outcome — and silencing everyone else.

Group actor mapping interface

Any group size → two sides

3+ speakers. 20 people. 200 messages. You assign two sides. The engine reads the dynamics between them. Scale is irrelevant — the structure is always binary at the decision level.

Noise is not neutral

Observers, low-signal contributors, and tangential actors dilute the dynamics. Silencing them is not just a UX choice — it makes the intelligence sharper. Focus is precision.

Same actor, different assignments

You can assign people from the same real-world team to opposing sides to test internal dynamics. A GP on your side, a GP on theirs. Who is actually pushing for you? Now you know.

Works across all formats

Group chat export, meeting transcript, Slack history, Discord thread. If it has multiple speakers and you can export it, Group Intelligence can read it.

Use cases

Novel uses. Real outcomes.

01

Enter a new group cold

Join a new community, partner ecosystem, or deal room. Export the group. Assign the key actors. Before you've said a word, you understand who couples with whom, where the alliances are, who to lean on, and what your strategy should be.

Useful for: new market entry, community evaluation, partner due diligence.

02

Team dynamics without the awkward evaluation

Export your team group chat. Set two team members as the two actors. Goal: professional. See how their professional relationship is structured — who drives, who resists, alignment levels. Do this for multiple pairs. Better task assignments, better team composition. No performance review required.

Can be automated via API. Run quarterly. Get team health reports without anyone knowing they were evaluated.

03

Isolate the decision maker

Your investor group has two GPs: Chen and James. Don't lump them together. Assign Chen to one side, James to the other. If Chen reads at 45% progress and James at 10%, the overall deal reads 10% — James is the bottleneck. Now you know exactly where friction lives and where to focus energy.

Works for any group where individual actors have different weights on the outcome.

04

Internal ally detection

Even within the other side — a partner team, an investor group, a client organisation — assign one of their members to your side as a test. See who is actually pushing for you versus holding the group back. Subtle power made visible without a single explicit conversation.

05

Meeting transcripts with 3+ speakers

A meeting with three or more speakers where you split them across two sides is Group Intelligence running on a meeting transcript. The same engine, the same signals. Map your team versus their team. Get cross-actor patterns, resistance vectors, and your recommended move — before you send the follow-up.

Supported formats: Zoom, Google Meet, Otter.ai, Fireflies, Notion, plain paste.

06

Community and ecosystem intelligence

Enter a new community, market, or partner network. Export a representative thread. Assign the voices that shape the conversation versus the rest. Understand the power structure of the space before you invest resources, build relationships, or make public moves.

Works anywhere you can export

Any multi-actor conversation.

WhatsApp Groups
Slack
Discord
Zoom Transcripts
Google Meet
Notion Notes
OTOtter.ai
FFFireflies
GRPAny group export

Supports .txt · .vtt · .md · .zip · Auto-detected on upload · 2+ speakers required

Intelligence output

Some of what surfaces.

The engine produces more than any list can capture.

Interaction state

Where the group dynamic stands right now — progressing, stalled, volatile, recovering.

Resistance vectors

What is specifically blocking progress and which actor is driving the friction.

Cross-actor patterns

How individual speakers relate to each other across the full conversation arc.

Who is investing more

Signal asymmetry across sides — who is carrying the dynamic and who is coasting.

What is working

The interaction patterns that are building progress, however small.

Optimal next move

A concrete, context-aware recommendation based on your stated goal.

Projections

Where this dynamic is heading if the current pattern continues unchallenged.

Risk signals

Ghosting probability, stall risk, rejection signals — surfaced before they become outcomes.

Privacy as Architecture

Intelligence without surrender.

Group conversations are sensitive. Your messages never leave your device. The open source classifier runs locally — only mathematical signal patterns are extracted and sent. Hermion never sees who said what.

Messages

on your device

Signals

extracted locally

Encrypted

before transmission

Hermion

models dynamics

Diagnosis

returned to you

Full privacy architecture →

The group has been talking.
Now you can read it.

Upload any group export. Assign your actors. Get the dynamics.

No account required · Messages never leave your device · Free to try