Back to selected work
CASE STUDY 01 / VOICE & CONVERSATIONAL AI

AI Voice Calling & Contact Center Platform

Conversations that don’t just respond — they take action.

CONTEXTProfessional / client work
ORGANIZATIONWishtree Technologies
FOCUSVoice & conversational AI
OVERVIEW

AI voice. Real action.

An inbound and outbound AI voice platform connecting lead conversations, follow-ups, appointment scheduling, and tool execution into one continuous workflow.

PythonPipecatTwilioGeminiDeepgramElevenLabsFastAPI
CONVERSATION ENGINECONCEPT / 01
Demo schedulingInbound conversation
Live call
CALLER

I’d like to schedule a demo.

AI AGENT

Sure. Let me check availability.

TOOL EXECUTIONcalendar.check()
Available slot found Ready to schedule
01 / THE PROBLEM

Where the work begins.

Business calls rarely end with a simple answer. A useful conversation needs to lead to an action: check availability, schedule a demo, follow up, or record an outcome. Connecting speech with these workflows requires more than a conversational model.

02 / WHAT WAS BUILT

A system around the need.

A production-oriented voice platform supporting inbound and outbound conversations, campaign workflows, voicemail handling, scheduling, and automated conversation outcomes.

Core capabilities

  • Inbound and outbound calling
  • Appointment and demo scheduling
  • Contextual tool execution
  • Voicemail handling
  • Campaign and follow-up workflows
  • Automated conversation outcomes
03 / MY CONTRIBUTION

Connecting the pieces.

At Wishtree Technologies, Kevin worked on the voice AI system and its integration with tools and business workflows, bringing together Python services, conversational orchestration, and telephony.

Technology & approach

PythonPipecatTwilioGeminiDeepgramElevenLabsFastAPI
04 / ENGINEERING CONSIDERATIONS

Beyond the happy path.

The design challenges that matter for this kind of system.

Conversation meets execution

A voice system must coordinate spoken intent with tool execution and communicate the result clearly. Tool completion and a fluent response are separate responsibilities.

Multiple services, one experience

Telephony, speech recognition, the language model, and speech generation each contribute state. A coherent conversation depends on coordinating the full sequence.

Outcomes beyond the call

Scheduling and follow-up workflows need useful conversation outcomes, rather than a transcript alone.

05 / OUTCOME & TAKEAWAYS

What the system enables.

A unified workflow for AI-led calls and the actions that follow them, spanning scheduling, follow-ups, tool use, and campaign operations.

Treat the conversation as an interface to a workflow. The engineering question is not only what the agent says, but what happens next.
This case study presents a qualitative, conceptual overview. Client identities, internal implementation details, and performance metrics are not disclosed.
NEXT CASE STUDY / 02

Knowledge, connected.