Skip to main content

Overview

Spinnable allows you to improve your AI agent’s performance through conversational feedback. Rather than technical configuration, you can simply tell your agent what it’s doing wrong and how to improve - just like training a team member.

How It Works

The Feedback Loop

  1. Observe - Notice when the agent provides incorrect or suboptimal responses
  2. Correct - Give feedback in natural language during the conversation
  3. Refine - The agent learns from your corrections and applies them to future interactions

Types of Feedback

Factual Corrections

Correct inaccurate information or outdated details

Tone Adjustments

Refine communication style and personality

Process Improvements

Optimize workflows and response patterns

Knowledge Gaps

Fill in missing information or context

Best Practices

Be Specific

"When customers ask about pricing, always mention our 14-day free trial first, then explain the three pricing tiers: Starter ($29/mo), Professional ($99/mo), and Enterprise (custom pricing)."

Provide Context

Give your agent the “why” behind corrections so it can generalize the learning:
"Don't use technical jargon like 'API endpoints' when talking to non-technical users. Instead, say 'integration options' or 'ways to connect'. Our customer base includes many small business owners without technical backgrounds."

Correct in Real-Time

The most effective feedback happens during actual conversations:
1

Notice the Issue

Observe when the agent makes a mistake or could perform better
2

Provide Immediate Feedback

Correct the agent in the same conversation thread
3

Verify Understanding

Ask the agent to confirm it understood the correction
4

Test the Learning

Try similar scenarios to ensure the improvement stuck

Real-World Example: MDS Portugal

MDS Portugal, a healthcare technology provider, uses Spinnable’s feedback system to continuously refine their customer support agent.

Initial Challenge

Their agent was providing technically accurate but overly complex explanations to healthcare providers who needed quick, actionable answers.

Feedback Applied

"When doctors ask about patient data security, start with the simple answer: 'Your patient data is encrypted and HIPAA-compliant.' Only provide technical details if they ask follow-up questions. Healthcare professionals are busy and need quick reassurance first."

Results

40% Faster

Average conversation time reduced

92% Satisfaction

Customer satisfaction score

60% Reduction

Escalations to human support

Feedback Patterns

Handling Edge Cases

When you notice the agent struggling with unusual scenarios:
Example
"When customers from the EU ask about data storage, mention that we have data centers in Frankfurt and Dublin. They're often asking because of GDPR requirements, even if they don't explicitly mention it."

Improving Empathy

Enhance the agent’s emotional intelligence:
Example
"When someone says they're frustrated or having trouble, acknowledge their feeling first: 'I understand this is frustrating' or 'I can see why that's confusing.' Then offer help. Don't jump straight to the solution."

Setting Boundaries

Teach the agent what it should and shouldn’t handle:
Example
"If someone asks for medical advice or diagnosis, you must say: 'I can't provide medical advice. Please consult with your healthcare provider about symptoms or treatment.' Then offer to help with questions about using our software instead."

Measuring Improvement

Track how your feedback is impacting performance:
1

Monitor Conversations

Review agent interactions regularly in the dashboard
2

Identify Patterns

Look for recurring issues or successful resolutions
3

Apply Systematic Feedback

Address patterns with targeted corrections
4

Verify Changes

Test the agent in similar scenarios to confirm improvements

Advanced Techniques

Scenario-Based Training

Create test scenarios to validate improvements:
  1. Prepare test cases - Write example conversations that previously caused issues
  2. Run the scenario - Have the agent respond to the test case
  3. Evaluate performance - Check if the agent applies previous feedback correctly
  4. Provide refinement - Add additional feedback if needed

Collaborative Training

Involve your team in the feedback process:

Team Training Best Practices

  • Share effective feedback examples with team members
  • Document common issues and recommended corrections
  • Assign team members to focus on specific agent capabilities
  • Review weekly to ensure consistent training approach

Common Pitfalls to Avoid

Contradictory Feedback: Ensure your corrections are consistent. If you tell the agent to “be brief” in one conversation and “provide detailed explanations” in another, clarify when each approach is appropriate.
Overfitting to Single Cases: Don’t over-correct based on one unusual interaction. Look for patterns before providing feedback.
Unclear Expectations: Vague feedback like “be better” or “improve responses” doesn’t give the agent actionable guidance.

Next Steps

Support

Need help refining your agent’s performance? Our team can help you develop an effective feedback strategy: