Propagandometer BUSINESS

Focus on the work.
Not the power games.

Spot pressure, personal attacks and suppressed objections in workplace discussions before they turn into prolonged conflict.

Discuss a pilot See an example
Your messages and AI model stay within your infrastructure
Quotes without direct identifiers Conversation context Human decisions Private deployment

Pressure doesn't always raise its voice.

A conversation can sound professional while making it harder for people to raise concerns and defend their views.

01

“You're against the team again”

A question about risk turns into a test of loyalty.

02

“Get some experience first”

Discussion of an idea is replaced by attacks on its author.

03

“You misunderstood me”

Decisions are retold. Responsibility for mistakes becomes someone else's.

See the pattern. Understand the context.

Example analysis

Launch discussion

[ EMP-017 ]10:42

Let's review the estimate. We haven't included support costs.

[ EMP-042 ]10:44

Everyone has already agreed. If you care about the outcome, stop holding the team back.

Analysis result

Potential pattern

Pressure through loyalty

Supporting the team is tied to dropping an objection. The question about costs remains unanswered in this excerpt.

What to check

Was the question answered elsewhere? Is this a recurring response to concerns about risk?

Illustrative example. A single remark does not establish a person's intentions or character.

From a signal to a constructive conversation.

01

Set the scope

Choose sources, a time period and risk categories.

02

Find relevant episodes

Identify signals and the messages that support them.

03

Review the context

Have a specialist confirm or dismiss the finding.

04

Track changes

Review recurring patterns after feedback.

Personal data minimisation

Follow the patterns. Keep names private.

The model doesn't need names or email addresses to analyse rhetoric. Direct identifiers are replaced with stable codes before analysis. Profiles are updated as new episodes arrive and are reviewed.

  • One code connects episodes over the selected period.
  • Your company stores the code-to-name mapping separately.
  • Identifiers are also removed from quotes and metadata.

This is pseudonymisation. Conversation content may still identify someone. Your company controls access and retention.

[ EMP-042 ]

Coded profile

Communication risk profile · last 30 days

Pressure through loyalty
2 confirmed episodes
Personal attacks
1 episode under review
Last reviewed
Today
Trend
Insufficient observations
How is the profile kept up to date?

New episodes are added to the history. Statuses change after review; dismissed signals do not count as confirmed. Older events leave the selected observation window.

The profile describes observed interactions, not personality traits. All example data is fictional.

Private corporate deployment

Your communications. Your infrastructure.

Corporate deployment is designed to keep storage, identifier replacement and the AI model inside your infrastructure, without sending text to external AI services. We agree architecture and access with your IT team before the pilot.

Company infrastructure
  1. 01Replace names with codes
  2. 02Local AI model
  3. 03Coded profiles for HR

Explore the value for your team.

Start with a focused pilot and agreed quality criteria. Evaluate signal accuracy and usefulness to HR.

Discuss a pilot