Proactive Alignment

By leveraging Fastino’s Personalization API, your AI systems can stay one step ahead — adapting in real time while remaining transparent, explainable, and user-controlled.

Overview

Proactive alignment is the ability for an agent to act before being told — but only when those actions are predictably beneficial, low-risk, and grounded in learned user intent.

Using Fastino’s world model, agents can:

• Detect emerging goals or shifts in behavior.

• Align task prioritization with user routines and energy windows.

• Make small, helpful moves (like scheduling, summarizing, or reminding) before explicit instruction.

• Learn when not to act, maintaining appropriate deference and transparency.

Key Components

Example: Anticipating a Scheduling Need

Fastino helps agents make time-aware proactive decisions.

POST /query
{
  "user_id": "usr_42af7c",
  "question": "What tasks or meetings should I proactively reschedule based on user preferences?"
}

Response

{
  "answer": "Ash prefers meetings after 1 PM and tends to move morning calls later in the week. Suggest rescheduling the Tuesday 10 AM sync to 2 PM."
}

Your scheduling assistant can use this context to propose — or automatically execute — an aligned update.

Example: Detecting Goal Drift

When users’ behavior deviates from past routines, agents can query Fastino for trend-based adjustments.

POST /query
{
  "user_id": "usr_42af7c",
  "question": "Have Ash’s recent work patterns shifted compared to previous weeks?"
}

Response

{
  "answer": "Ash has been working later in the evenings and skipping morning focus blocks — suggest adjusting daily planning routines."
}

This enables proactive adaptation before frustration or inefficiency occurs.

Example: Proactive Reminders

Use deterministic summaries to identify upcoming needs:

Response

{
  "summary": "Ash prefers early-week planning reminders and status check-ins on Thursday afternoons."
}

Your assistant can use this to trigger helpful, context-sensitive reminders automatically.

Continuous Alignment Loop

This loop allows assistants to evolve from reactive tools into adaptive collaborators.

Example: Capturing a Successful Suggestion

When the agent acts proactively and the user approves:

POST /ingest
{
  "user_id": "usr_42af7c",
  "source": "scheduler_agent",
  "events": [
    {
      "event_id": "evt_proactive_01",
      "type": "outcome",
      "timestamp": "2025-10-27T15:00:00Z",
      "metadata": {
        "context": "meeting_reschedule",
        "status": "accepted"
      },
      "content": "User accepted proactive meeting reschedule suggestion."
    }
  ]
}

Fastino stores this success as a positive outcome and reinforces proactive confidence.

Example: Logging a Declined Suggestion

Agents can also log declined actions to avoid overreach.

POST /ingest
{
  "user_id": "usr_42af7c",
  "source": "scheduler_agent",
  "events": [
    {
      "event_id": "evt_proactive_02",
      "type": "correction",
      "timestamp": "2025-10-27T15:10:00Z",
      "metadata": {
        "context": "task_priority",
        "status": "declined"
      },
      "content": "User rejected automatic prioritization; prefers manual ordering for now."
    }
  ]
}

This feedback ensures the agent’s proactive behavior remains trustworthy and aligned.

Integration with Other Use-cases

Fastino unifies these capabilities into a single feedback-driven personalization loop.

Example Implementation (Python)

import requests

BASE_URL = "https://api.fastino.ai"
HEADERS = {"Authorization": "x-api-key sk_live_456", "Content-Type": "application/json"}

def get_proactive_summary(user_id):
    r = requests.get(f"{BASE_URL}/summary?user_id={user_id}&purpose=proactive-alignment", headers=HEADERS)
    return r.json()["summary"]

def log_outcome(user_id, content, accepted=True):
    event_type = "outcome" if accepted else "correction"
    payload = {
        "user_id": user_id,
        "source": "proactive_agent",
        "events": [
            {
                "event_id": "evt_" + event_type,
                "type": event_type,
                "timestamp": "2025-10-27T15:30:00Z",
                "content": content
            }
        ]
    }
    requests.post(f"{BASE_URL}/ingest", json=payload, headers=HEADERS)

# Example
print(get_proactive_summary("usr_42af7c"))
log_outcome("usr_42af7c", "User accepted proactive reminder to finalize slides.", accepted=True)

Proactive Confidence Scoring

Agents can maintain a confidence score for proactive decisions using a combination of:

• Historical approval rates

• Feedback signals

• Boundary summaries

Example query:

POST /query
{
  "user_id": "usr_42af7c",
  "question": "How confident should I be in making proactive scheduling suggestions?"
}

Response

{
  "answer": "High confidence — Ash accepts 80% of proactive meeting reschedules and provides positive feedback consistently."
}

This gives agents quantifiable, data-backed thresholds for autonomous behavior.

Use Cases

Best Practices

• Always query the latest purpose=proactive-alignment summary before acting.

• Use feedback ingestion to calibrate proactivity level dynamically.

• Explain every proactive suggestion — never act silently.

• Pair with Action Boundaries for ethical safety.

• Keep proactive windows (time, scope, and domain) well-defined.

• Use summaries, not raw events, in latency-sensitive workflows.

Example: Deterministic Proactive Summary

Response

{
  "summary": "Ash appreciates proactive nudges when aligned with weekly goals and prefers concise, transparent suggestions before execution."
}

This summary can be embedded in agent system prompts or workflows to ensure predictable, user-aligned behavior.

Summary

The Proactive Alignment use case transforms reactive assistants into adaptive collaborators — ones that think ahead while staying transparent, explainable, and safe.

By grounding proactive behavior in Fastino’s personalization engine, your agents can predict needs, suggest actions, and evolve continuously — acting as intelligent partners that anticipate without overstepping.

Next, continue to Personalization Use Cases → Routine Prediction to explore how Fastino models user activity and daily rhythms.