Untitled Proposal
Executive Summary
BusUp is a leading corporate mobility company operating across six countries—Brazil, Portugal, Spain, Mexico, the United States, and beyond—providing bus transportation services for marquee clients including Amazon, Accenture, BBVA, and SpaceX. With a 16-person customer care team operating 24/7, 365 days a year, BusUp delivers exceptional service but faces mounting challenges in its voice channel, which accounts for approximately 20% of inbound contacts yet creates the greatest operational friction and customer dissatisfaction.
The core issue is clear: when a passenger calls, they wait. When they wait, they grow frustrated. When an agent picks up, they're blocked for five to ten minutes per call—unable to handle other requests simultaneously, unlike text-based channels. During peak demand moments—such as when a bus breaks down and 50 passengers all call at once—the system buckles under pressure.
BoSar proposes building an intelligent AI Voice Agent powered by Retell AI that answers every call immediately, identifies the caller, routes through the appropriate workflow, resolves what it can autonomously, and seamlessly escalates to human agents with full context when necessary. This solution will be multilingual (Portuguese, Spanish, and English), integrated with BusUp's proprietary dashboard and Zendesk, and designed to scale across all markets with full GDPR compliance for European operations.
Our goal is simple: zero wait time, fewer agent errors, dramatically reduced call handling costs, and a customer experience that feels personal, fast, and accurate—regardless of which company the passenger works for.
Project Understanding
The Voice Channel Problem
While WhatsApp represents the largest share of inbound contacts (~53%), voice is where the pain concentrates. Unlike WhatsApp, where an agent can juggle multiple conversations, a phone call locks an agent into a single interaction for five to ten minutes. During peak events—a bus breakdown affecting 50 passengers, a service disruption, or morning rush-hour delays—the voice queue becomes a bottleneck that drives dissatisfaction and overwhelms your team.
Knowledge Complexity Across Clients
BusUp doesn't offer a one-size-fits-all service. Each corporate client—Amazon, Accenture, BBVA, SpaceX, and many others—has unique policies. One client may allow Uber reimbursement when a bus is 15 minutes late; another may prohibit it entirely. Your agents need to hold extensive knowledge about dozens of clients' specific rules, and mistakes here directly translate into passenger dissatisfaction and operational issues. Training time is long, and errors are frequent.
Well-Defined Workflows Ready for Automation
BusUp has already invested significant effort in mapping approximately 30 distinct contact reason flows—from lost items, to bus location inquiries, to ticket cancellations, password resets, and complaint registration. Each flow includes decision trees, API call points (highlighted in orange in your internal documentation), and escalation triggers. This is an exceptional foundation for AI agent development.
Dynamic, Real-Time Situations
Not everything lives in a static knowledge base. Live events—bus breakdowns, route deviations, delays—require real-time information that flows between your traffic monitoring team and customer care. The AI agent must be able to handle these dynamic situations gracefully, either by querying live systems or by escalating with full context.
Multilingual & Multi-Region Requirements
The agent must operate fluently in Brazilian Portuguese, European Portuguese, Spanish, and English, with the flexibility to add additional languages over time. European operations require GDPR-compliant infrastructure with servers hosted in Europe.
Integration Requirements
- Zendesk — Ticket creation and management for every interaction
- BusUp Proprietary Dashboard — Caller identification, account lookup, route/service information, and operational data via API endpoints (some existing, some to be built)
- Retell AI — Voice agent platform with per-minute pricing
- JIRA — Complaint and operations ticket creation
Proposed Solution
Intelligent AI Voice Agent on Retell AI
We will design, build, and deploy a production-ready AI Voice Agent using Retell AI as the telephony and voice platform. The agent will be purpose-built for BusUp's customer care operations with the following capabilities:
1. Instant Call Answering & Caller Identification
- Every inbound call answered immediately—zero wait time, regardless of volume
- Automatic caller identification via phone number lookup against BusUp's dashboard API
- Identity confirmation flow (\"I see you're calling from a number registered to Thibaud. Is that correct?\")
- Graceful handling of mismatched identities with escalation to human agents
2. Multi-Flow Conversation Engine
- Implementation of all ~30 contact reason flows using Retell's flex mode, allowing the agent to navigate dynamically between conversation paths
- Non-rigid routing: if a conversation shifts mid-flow, the agent adapts and transitions to the appropriate workflow without breaking
- Access to general knowledge base for policy questions that don't fit a specific flow
- Client-specific policy enforcement—the agent knows whether to offer an Uber reimbursement or not, based on which company the passenger belongs to
3. Smart Escalation with Warm Transfers
- When the AI cannot resolve a call—identity mismatch, frustrated caller, complex situation, missing information—it escalates to a human agent
- Warm transfer protocol: before the human agent picks up, the AI provides a spoken summary of the call context, the caller's identity, the issue, and what has already been attempted
- This eliminates the \"please repeat everything\" experience that frustrates passengers
4. Guardrails & Safety
- Anti-hallucination safeguards: when the agent doesn't know an answer, it acknowledges this honestly and escalates rather than fabricating information
- Strict adherence to client-specific policies—no cross-contamination of rules between companies
- Conversation monitoring and logging for quality assurance
5. Multilingual Operation
- Full support for Brazilian Portuguese, European Portuguese, Spanish, and English
- Language detection and switching capabilities
- Architecture designed to add additional languages in the future
6. System Integrations
- Zendesk: Automatic ticket creation for every call, with full conversation summary, contact reason classification, and resolution status
- BusUp Dashboard: Real-time API queries for caller identification, account information, route status, and service details
- JIRA: Automated complaint ticket creation routed to the operations team
- Dynamic Knowledge Updates: A mechanism for your traffic team to flag live incidents that the AI agent can reference in real-time
7. European Infrastructure
- Separate deployment with European-hosted servers for all EU operations
- Full GDPR compliance in data processing and storage
Scope of Work
Phase 1: Discovery & Architecture (Weeks 1–2)
- Deep-dive workshops with BusUp's customer care and traffic teams
- Review and map all 30 contact reason flows for AI implementation
- Identify existing API endpoints and document new endpoints required from BusUp's engineering team
- Define client-specific policy matrix for all major corporate accounts
- Design conversation architecture, escalation rules, and guardrails
- Define dynamic knowledge update mechanism for live events
- Deliverable: Technical Architecture Document & Conversation Design Blueprint
Phase 2: Core Agent Build — Priority Flows (Weeks 3–5)
- Set up Retell AI infrastructure (including separate EU servers)
- Build caller identification and verification flow with BusUp dashboard integration
- Implement first batch of high-priority flows (estimated 5–8 flows based on volume and impact)
- Integrate Zendesk ticket creation
- Integrate JIRA for complaint workflows
- Configure multilingual support (Brazilian Portuguese, European Portuguese, Spanish, English)
- Implement warm transfer escalation protocol
- Deliverable: Functional AI Voice Agent with priority flows, ready for testing
Phase 3: Testing & Iteration (Weeks 6–7)
- Internal testing with BusUp team across all supported languages
- Stress testing with concurrent call scenarios (peak simulation)
- Fine-tuning conversation quality, response accuracy, and escalation triggers
- Adjust client-specific policy handling based on test results
- Deliverable: Test report with performance metrics and optimized agent
Phase 4: Expanded Flow Implementation (Weeks 8–10)
- Implement remaining contact reason flows (approximately 20–22 additional flows)
- Build any additional API endpoints identified during Phase 1 (in coordination with BusUp's engineering team)
- Implement dynamic knowledge base update system for live events
- Refine general knowledge base responses for edge-case questions
- Deliverable: Complete AI Voice Agent with all 30 flows operational
Phase 5: Production Launch & Optimization (Weeks 11–12)
- Gradual rollout: start with one region/language, expand to full deployment
- Real-time monitoring during initial live period
- Performance optimization based on live call data
- Knowledge base and flow adjustments based on real interactions
- Final handoff documentation and training for BusUp's team
- Deliverable: Fully deployed, production-ready AI Voice Agent across all regions
Timeline
| Phase | Duration | Milestone |
|---|---|---|
| Phase 1: Discovery & Architecture | Weeks 1–2 | Architecture Document & Conversation Blueprint |
| Phase 2: Core Agent Build | Weeks 3–5 | Functional Agent with Priority Flows |
| Phase 3: Testing & Iteration | Weeks 6–7 | Optimized Agent Ready for Expansion |
| Phase 4: Expanded Flow Implementation | Weeks 8–10 | All 30 Flows Operational |
| Phase 5: Production Launch | Weeks 11–12 | Full Production Deployment |
Total project duration: approximately 12 weeks from kickoff.
Investment
The following is the one-time setup fee for the complete design, development, and deployment of your AI Voice Agent. This does not include Retell AI's per-minute usage fees, which BusUp will pay directly to Retell based on call volume.
One-Time Setup Fee
| Component | Investment |
|---|---|
| Phase 1: Discovery, Architecture & Conversation Design | $5,500 |
| Phase 2: Core Agent Build (5–8 priority flows, Zendesk + JIRA + Dashboard integrations, multilingual config, EU infrastructure setup) | $14,500 |
| Phase 3: Testing, Stress Testing & Optimization | $4,000 |
| Phase 4: Expanded Flow Implementation (~22 additional flows, dynamic knowledge base, additional API integrations) | $12,000 |
| Phase 5: Production Launch, Monitoring & Handoff | $4,000 |
| Total One-Time Setup Fee | $40,000 |
What's Included
- Complete AI Voice Agent design and build across all 30 contact reason flows
- Multilingual support: Brazilian Portuguese, European Portuguese, Spanish, English
- Full Zendesk, JIRA, and BusUp Dashboard API integrations
- Warm transfer and smart escalation logic
- Client-specific policy engine for differentiated service per corporate account
- Separate European infrastructure deployment (GDPR-compliant)
- Concurrent call handling for peak demand scenarios
- Anti-hallucination guardrails and conversation safety measures
- Documentation and team training
Optional: Ongoing Retainer
After launch, we recommend an ongoing retainer to ensure your AI agent stays sharp, adapts to new clients and policies, and continuously improves based on real-world performance data.
| Retainer Plan | Monthly Fee | Includes |
|---|---|---|
| Standard Maintenance | $2,500/month | Monitoring, bug fixes, minor flow adjustments, knowledge base updates, up to 10 hours/month of development |
| Growth & Optimization | $4,500/month | Everything in Standard plus new flow development, new language additions, advanced analytics, up to 25 hours/month of development |
The retainer is optional and can be discussed after the initial deployment is complete.
Not Included
- Retell AI per-minute usage fees (paid directly to Retell by BusUp)
- Development of new API endpoints on BusUp's dashboard (to be handled by BusUp's engineering team, with our guidance and documentation)
- Zendesk AI tool licensing (separate from this engagement)
Why BoSar
Proven Retell AI Expertise
We are recognized on Retell AI's partner ecosystem, which is how BusUp found us. We have deep, hands-on experience building production AI voice agents on the Retell platform—including projects handling 50+ concurrent calls with reliable performance.
Custom Integration Specialists
Our team, led by CTO Nick Romanenko, has spent the past two years building custom API integrations for AI agent deployments. Whether it's Zendesk, JIRA, proprietary dashboards, or CRM systems—we have the technical depth to connect your AI agent to every system it needs.
Multilingual & Multi-Region Experience
We understand the nuances of deploying AI across multiple languages and regions, including the critical difference between Brazilian Portuguese and European Portuguese. Our infrastructure approach ensures GDPR compliance for European operations while maintaining performance globally.
Conversation Design That Works
We don't just plug in an AI model and hope for the best. We design conversation architectures with guardrails, fallback logic, warm transfers, and anti-hallucination measures that ensure your passengers have a reliable, trustworthy experience every time they call.
Phased, Pragmatic Approach
As Thibaud noted on the call: the goal isn't to build everything from day one. We share this philosophy. Our phased approach lets BusUp start seeing value quickly with priority flows while we continue building toward full coverage—reducing risk and enabling real-world learning at every stage.
Next Steps
We're excited about the opportunity to transform BusUp's voice channel from its biggest pain point into a competitive advantage. Here's how we move forward:
- Step 1: Review this proposal and share any questions or feedback with us
- Step 2: Schedule a follow-up call to align on priority flows, API documentation exchange, and any scope adjustments
- Step 3: Upon agreement, we issue the project contract and kick off Phase 1 immediately