How Terrapinn Scaled Outreach with Two-Way Human–AI Agent
From manual calling to predictable, high-performance AI engagement
- 35%Average conversion rate
- 7×Uplift vs. traditional outbound benchmarks
- 1,731Real production calls executed
The Challenge
Terrapinn is a global events company producing hundreds of large-scale conferences and exhibitions each year across industries including rail, energy, mobility, and infrastructure. With audiences spread across regions and languages, repeat attendance is critical, yet increasingly difficult to drive at scale.
The team at Terrapinn wanted to automate this with AI and ruled out fully generative, black-box solutions. Outbound conversations directly impact revenue, brand reputation, and customer trust, and the business needed full visibility, predictability, and control over every interaction.
For their Rail Live event, Terrapinn set out to use an AI Agent (now called Terry) to proactively reach past attendees in order to:
- Increase repeat registrations
- Identify high-intent prospects
- Answer attendee questions in real time
- Enrich CRM data with structured insights
- Measure outreach performance reliably at scale
Their benchmark expectation was clear and conservative: ~5% conversion, consistent with historical human outbound performance.
GetVocal's AI Agent provided us with a new, scalable digital workforce capable of connecting with last year's attendees, answering their specific questions, and driving registrations. This is crucial for enhancing the end-to-end delegate journey while simultaneously delivering a measurable uplift in our event attendance metrics.
The Transformation
Terrapinn deployed a multilingual outbound AI voice agent built on GetVocal’s Conversational AI Platform.
Rather than relying on unconstrained generative AI, the agent was designed using GetVocal’s protocol-driven conversational graph. This ensured every conversation followed a deterministic, auditable flow, all while using LLMs strictly where they add value: natural language understanding and response flexibility.
The AI agent autonomously:
- Called past event attendees
- Personalised conversations using known attendee data
- Qualified intent to attend upcoming events
- Confirmed and corrected email details
- Answered common questions via a knowledge base
- Captured callback requests and speaker questions
- Handled English and Spanish seamlessly
When conversations moved outside predefined protocols, outcomes were explicitly captured, never inferred or guessed.
The rollout prioritised real production conditions over experimentation:
- Weeks to roll out: 4
- Total calls: 1,731
- Contacts reached: 1,282
This was not a pilot environment. The agent operated under live scheduling windows, retry logic, and real data constraints, exactly how Terrapinn would run it at scale.
The Results
The AI agent significantly outperformed expectations, and improved with scale.
- 35% average conversion rate across all batches
- 7× higher than the 5% benchmark
- Stable performance between 28% - 43%
- 74% engagement rate in the final batch
Performance improved with retries and volume, rather than degrading.
The Impact
Beyond conversion, the impact was operational and strategic.
- 122 interested attendees identified
- 114+ attendee questions captured
- 11 email corrections and 13 callback requests
- All insights delivered as structured, CRM-ready data
- Zero manual note-taking
In practical terms, GetVocal’s AI handled the equivalent of two full-time human agents working for four days, saving 34 human hours for a single event, with consistent quality and no fatigue.
How Get Vocal Works