BENJAMIN NWANDU / CONVERSATIONAL AI SPECIALIST

I build Conversational AI systems that turn conversations into business actions.

From the first hello to the next business action.
Voice agents, connected workflows, and human oversight.

View my work Work with me Walk through my thinking

The human side of Voice AI

Voice AIAI automationQuality & evaluationAgentic systems

01 / What I build

Beyond the conversation.

The voice is the interface. The system behind it makes the difference.

Conversational systems

Voice experiences that understand intent, handle routine requests, and know when to bring in a human.

Connected automation

Workflows that carry conversational context into calendars, CRMs, APIs, and business processes.

Reliability by design

Evaluation, validation, and approval checkpoints that make system behavior easier to inspect and improve.

02 / The core areas

Three jobs, done properly.

Every call starts the same way — a person with a phone and a reason to ring.
These are the three reasons I build agents for.

Core areaAppointment SchedulingBooked, moved, or cancelled — without a callback queue.Full scope

Core areaLead QualificationAsk the qualifying questions every time, then route the answer.Full scope

Core areaCustomer SupportAnswer what is known. Escalate what is not.Full scope

03 / Selected work & explorations

Systems, with purpose.

All projects

The flagship / 01In Development

Voice AI Workflow Builder.

Describe the conversation. Design the system behind it.

A capability-first compiler for turning natural-language requirements into validated, platform-neutral workflows.

Explore the project Try the compiler Walk through my thinking

01 / THE REQUIREMENT

“Build a voice agent that qualifies leads and books a call.”

UseCaseSpecWorkflowSpec

Platform adaptern8n

Proposed architecture · currently in development

Customer supportENTIN Bank AI Voice SupportA voice agent for a fictional Nigerian bank that handles failed transfers, card disputes and fraud — with identity checks and risk rules enforced in the workflow, not the prompt.Full scope

AgentToolsn8nWebhooksContract checkURL · authphone vs phone_numberWorkflow inactiveCIBlockedReliability engineeringAgent PreflightA pre-production check for a voice agent’s webhook tools and the n8n workflows behind them. It prints what will break, and fails the build when it matters.Full scope

Call logTierOutcomeIntent · hang-upCounted byDeterministic metricsThenHuman reviewPublishedClient reportReporting & QA operationsVoice Agent Reporting HubA reporting system that turns raw voice-agent call logs into client-ready weekly reports — built for a voice AI agency and used on real client accounts.Full scope

Lead qualificationOriki Homes: Real Estate Speed-to-Lead AgentAn AI assistant, Joy, who answers a Nigerian property enquiry in under a minute on WhatsApp or by live voice, qualifies the buyer against a tested scoring rubric, and alerts a human the moment a hot lead appears.Full scope

Voice AIFairy Share Dental AI ReceptionistAn inbound dental receptionist that books into the practice management system — and only says “booked” after it has read the appointment back and checked every field.Full scope

Call libraryRecorded callsExpected behaviourReplayCandidate versionBehaviour diffBooking flowCaller-name accuracyEscalation pathRelease gateBlockedReliability engineeringVoice Agent Regression HarnessA proposed replay harness that runs recorded calls against a new agent version and blocks the release when behaviour regresses.PlannedFull scope

Proposed actionAccount change · pendingRisk evaluationAbove thresholdApproveRejectEither wayAudit trailControl & reliabilityHuman-in-the-Loop AutomationLet AI prepare the action. Keep consequential decisions with a human.PlannedFull scope

Where the turn goesEndpointingModel first tokenSpeech first byteNetworkMeasured againstBudget per stageBarge-inStop · keep state · resumeVoice performanceLatency & Barge-in LabA proposed measurement rig for where the seconds go in a voice turn, and how cleanly the agent handles being interrupted.PlannedFull scope

As a QA Analyst at VoiceROI, I reviewed inbound and outbound Voice AI interactions across client systems — examining how agents actually behaved in real conversations, not just how they were supposed to.
Benjamin NwanduQA Analyst · VoiceROI

04 / What I reviewed

HallucinationsCaller-name accuracyBooking behaviorScript adherenceCall drop-offsConversation quality

That experience informs how I think about building: the happy path is only the beginning.

05 / Technical direction

The right tool for the system.

Platforms, models and infrastructure I build with. Project pages distinguish planned tools from verified implementation.

Retell
Vapi
n8n
Cal.com
Calendly
Python
Cursor
Claude
ChatGPT
Google Calendar
ElevenLabs
Deepgram
Make
Zapier
Airtable
HubSpot
Notion
Retell
Vapi
n8n
Cal.com
Calendly
Python
Cursor
Claude
ChatGPT
Google Calendar
ElevenLabs
Deepgram
Make
Zapier
Airtable
HubSpot
Notion

06 / How I build

Thoughtful by architecture.

Useful automation starts with clear boundaries, and improves through evidence.

Understand first

Define the task, the context, and what a successful outcome actually looks like.

Design the boundaries

Separate reasoning from rules. Make permissions and escalation paths explicit.

Build for the handoff

Connect the conversation to business systems and the people responsible for the next step.

Evaluate, then improve

Test failure paths, review behavior, and turn findings into regression checks.

07 / A little about me

I’m Benjamin. I think in connected systems.

My focus sits where conversation, automation, and quality meet. I’m interested in what happens after the AI responds: the booking, the handoff, the business action.

More about my approach

Let’s build something useful

Good conversations.
Better systems.

Have a conversational AI challenge, an automation idea, or a role in mind?

Email LinkedIn