Advice

Generate structured behavioral guidance for complex human interactions
(feedback, conflict, persuasion, alignment, difficult conversations).

This endpoint is designed for LLM tool calling, not for direct human consumption.

SotsAI returns a neutral, explainable reasoning structure that your own LLM can transform into natural language, in your tone, your language, and your UI.

When should your system call this?

Call this tool when the user’s request involves:

interpersonal tension or misunderstanding
adapting communication to another person
emotional reactions, resistance, or disengagement
influence, feedback, negotiation, or alignment

Do not call it for:

factual or informational questions
pure rewriting or translation
generic advice with no interpersonal context

If the question is “how should I talk to this person in this situation?”
→ this tool is probably the right one.

High-level flow

Your orchestration layer determines that behavioral reasoning may be needed
Your LLM calls Advice with:
- a short situation summary
- one or two psychometric profiles
SotsAI returns a structured advice object
Your LLM renders the final response (email, checklist, script, coaching text…)

Advice request (input)

Required fields

Situation summary

A short, sanitized English description of the situation.

Focus on behaviors, stakes, and intent
Avoid names, emails, or sensitive identifiers
2–6 sentences is usually enough

Example:

“The user needs to give corrective feedback to a direct report whose work quality is inconsistent. When the user is very direct, their direct report tends to shut down and become quiet. The user wants improvement without damaging trust.”

User profile

You must provide a psychometric profile describing the person asking for advice.

This profile may come from:

your own systems (bring-your-own), or
data previously collected via SotsAI DISC

In all cases, the profile data must be included explicitly in the request.

Optional fields (strongly recommended)

Interlocutor profile

The psychometric profile of the other person involved.

If omitted, SotsAI will focus on self-adaptation strategies only. If provided, the interlocutor profile must use the same psychometric framework as the user profile.

Relationship type

Helps frame power dynamics and expectations.

Typical values:

manager
direct_report
peer
self
other

Situation hint

A short semantic hint such as:

giving_feedback
conflict_management
persuasion
change_management

This is optional; SotsAI will still classify internally.

Example request

{
  "situation_type_hint": "giving_feedback",
  "relationship_type": "direct_report",
  "user_profile": {
    "tool": "disc",
    "raw_scores": {
      "natural": { "D": 78, "I": 64, "S": 22, "C": 36 },
      "adapted": { "D": 70, "I": 58, "S": 30, "C": 42 }
    }
  },
    "interlocutor_profile": {
      "tool": "disc",
      "raw_scores": {
        "natural": { "D": 18, "I": 32, "S": 70, "C": 76 },
        "adapted": { "D": 22, "I": 28, "S": 74, "C": 80 }
      }
    }
  },
  "context_summary": "The user needs to give corrective feedback to a direct report whose work quality is inconsistent. When the user is very direct, their direct report tends to shut down. The user wants improvement without damaging trust.",
  "language": "fr"
}

Advice response (output)

The response is a structured advice object, not a text answer.

It is designed to be:

stable
explainable
safe to re-use across languages and channels

What it contains

Primary framing: Whether the situation is mostly about friction, alignment, or both
Tone guidance: How the final message should sound (e.g. reassure, soften, be specific)
Profile lenses: Short, neutral summaries of the user (and interlocutor if present)
Interaction dynamics: Key friction or synergy dimensions in the interaction
Behavioral levers: Concrete communication strategies and why they work
Risk patterns: What typically goes wrong if no adaptation happens
Reflection handles: Prompts your LLM may use to encourage self-reflection

Example response (excerpt)

{
  "primary_tension_frame": "friction",
  "tone_guidance": ["be_clear", "ground", "soften"],
  "impact_estimate": "high",

  "user_profile_lens": {
    "style_summary": "Direct, big-picture oriented, and assertive in communication, with a strong focus on results.",
    "dominant_drivers": ["achieving results", "innovation"],
    "sensitive_zones": ["perceived lack of progress", "feeling unheard"]
  },

  "interlocutor_profile_lens": {
    "present": false,
    "style_summary": "Not available",
    "dominant_drivers": [],
    "sensitive_zones": []
  },

  "dynamics_lens": {
    "friction_axes": [
      {
        "axis_id": "directness",
        "intensity": "high",
        "description": "High directness may be perceived as harsh in a feedback context.",
        "likely_effect": "Defensive reactions or withdrawal."
      },
      {
        "axis_id": "big_picture_orientation",
        "intensity": "medium",
        "description": "Emphasis on goals may overshadow concrete guidance.",
        "likely_effect": "Lack of clarity on what to improve."
      }
    ],
    "synergy_axes": [
      {
        "axis_id": "assertiveness",
        "intensity": "high",
        "description": "Clear assertiveness helps communicate expectations.",
        "likely_effect": "Better understanding of priorities."
      }
    ]
  },

  "behavioral_levers": [
    {
      "lever_id": "balance_directness_with_empathy",
      "related_axes": ["directness"],
      "description": "Maintain clarity while softening delivery.",
      "intended_effect": "Reduce perceived harshness while keeping expectations clear."
    },
    {
      "lever_id": "focus_on_specifics",
      "related_axes": ["big_picture_orientation"],
      "description": "Pair high-level goals with concrete examples and next steps.",
      "intended_effect": "Increase clarity and actionability."
    }
  ],

  "risk_patterns_if_ignored": [
    {
      "pattern_id": "harsh_feedback",
      "description": "Feedback remains blunt and unbuffered.",
      "consequences": ["defensive reactions", "relationship strain"]
    },
    {
      "pattern_id": "lack_of_specificity",
      "description": "Feedback stays abstract and non-actionable.",
      "consequences": ["confusion", "limited improvement"]
    }
  ],

  "reflection_handles": [
    {
      "handle_id": "reflect_on_feedback_style",
      "focus": "How does your delivery style affect the other person’s openness to feedback?"
    }
  ]
}

Field names and descriptions may evolve across versions. You should rely on the semantic meaning of each section, not on fixed phrasing.

How your LLM should use this

Think of the response as raw reasoning material, not a script.

A good rendering:

follows the tone guidance
uses 1–2 behavioral levers only
adapts language and format to the user context
avoids sounding like a psychological report

A bad rendering:

dumps the structure verbatim
over-analyzes personalities
introduces assumptions not present in the request

Writing good situation summaries

describe observable behavior
mention intent and constraints
keep it concise

Avoid

names, emails, company secrets
emotional diagnoses
long transcripts

If the user provides sensitive data, sanitize it before calling the tool.

Errors & limits

Requests are authenticated via your organization key
Usage counts toward your monthly quota
Standard error responses are returned for invalid input, quota exhaustion, or internal failures

Error responses include a stable error code that can be used for programmatic handling.

In short

This tool gives your LLM:

clarity about interpersonal dynamics
structure for safer reasoning
freedom to generate the final message in your own style

You stay in control of voice and UX. SotsAI stays focused on behavioral intelligence.