Skip to content
Documentation

When to call SotsAI

SotsAI should not be called on every LLM request.

It is designed for situations where human interaction quality matters,
not for general knowledge or task execution.

This page gives you clear rules to decide when to call SotsAI.

The simple rule

Section titled “The simple rule”

Use SotsAI if all conditions are true:

  1. The response involves another person
  2. How the message is received matters
  3. A user psychometric profile is available

If any condition is false, your LLM should not call SotsAI.

Typical situations where SotsAI should be called

Section titled “Typical situations where SotsAI should be called”

Call SotsAI when the situation involves:

  • giving or receiving feedback
  • explaining a decision that may be questioned
  • handling resistance, tension, or disagreement
  • persuading or influencing someone
  • aligning stakeholders
  • managing up or managing down
  • addressing emotional reactions or misunderstandings

In these cases, content alone is not enough —
how something is said matters as much as what is said.

Situations where SotsAI is usually not needed

Section titled “Situations where SotsAI is usually not needed”

Do not call SotsAI for:

  • factual questions (“What is X?”)
  • technical explanations (“How does this API work?”)
  • summarization or rewriting without interpersonal intent
  • pure data extraction or transformation
  • internal reasoning with no human recipient

In these cases, your LLM can operate independently.

A decision checklist (practical)

Section titled “A decision checklist (practical)”

Before calling SotsAI, ask:

  • Is there an interlocutor (explicit or implicit)?
  • Could this message trigger resistance, emotion, or misunderstanding?
  • Would adapting tone, structure, or intent improve the outcome?
  • Do I have a user psychometric profile available?

If the answer is “yes” to all → call SotsAI.

One-person vs two-person situations

Section titled “One-person vs two-person situations”

One-person (self-adjustment)

Section titled “One-person (self-adjustment)”

Call SotsAI with:

  • context_summary (a PII-free summary of the user query)
  • user_profile (the user psychometric profile)

Example:

  • “How should I prepare feedback for my team?”
  • “How should I approach this conversation?”

SotsAI helps the user adjust their own communication style.

Two-person (adaptation + friction reasoning)

Section titled “Two-person (adaptation + friction reasoning)”

Call SotsAI with:

  • context_summary
  • user_profile
  • interlocutor_profile (the interlocutor psychometric profile)

Example:

  • “How should I give this feedback to this specific person?”
  • “Why did my message land badly with that person?”

This is where SotsAI provides the highest value.

What happens if profiles are missing?

Section titled “What happens if profiles are missing?”

If a user profile is not available:

Do not call SotsAI.

Instead, your LLM should handle the request autonomously using its own generic reasoning.

Valid strategies when profiles are missing

Section titled “Valid strategies when profiles are missing”
StrategyDescription
1. Gate the experienceAsk the user to complete a psychometric profile before proceeding.
2. Collect profiles via SotsAI (DISC)Invite users, then retry once profile data is available.
3. Fallback outside SotsAILet your LLM generate a generic response without calling SotsAI.

See:

Common anti-patterns

Section titled “Common anti-patterns”

Avoid these mistakes:

  • Calling SotsAI for every message “just in case”
  • Calling SotsAI without a user psychometric profile
  • Treating SotsAI as a generic coaching generator
  • Hiding missing profiles instead of handling them explicitly

SotsAI works best when called intentionally.