I will design an interactive voice agent and conversational AI voice persona

I will design an interactive voice agent and conversational AI voice persona that sounds human, stays on-brand, and handles real conversations end to end. From persona and dialogue flows to a working, testable prototype, I build voice experiences your customers actually want to talk to.

What you get

• A documented voice persona: name, personality, tone, speaking style, do/don't language, and a sample-line library so the voice stays consistent across every reply.
• A conversation design map: intents, entry points, happy paths, fallback paths, disambiguation, confirmations, and graceful hand-off to a human.
• A dialogue flow specification (flowchart + written prompts) covering greetings, slot-filling, error recovery, interruptions (barge-in), and end-of-call wrap-up.
• A system prompt / instruction set tuned for an LLM-driven agent (intent handling, guardrails, refusal language, and out-of-scope responses).
• A working interactive prototype you can speak or type to, built on a platform we agree on up front (for example a Retell / Vapi / LiveKit / Twilio-style stack, or your existing IVR/LLM backend).
• A voice and TTS recommendation: suggested voice model, speaking rate, pronunciation fixes (SSML where supported), and a short A/B of candidate voices.
• A test script and QA checklist with representative utterances, edge cases, and pass/fail notes so you can verify behavior yourself.
• A handover document explaining how each piece fits together, how to edit prompts, and how to extend the agent later.

Plans

	Basic	Standard	Premium
Voice persona definition	Core persona + sample lines	Full persona guide	Full persona + multi-voice variants
Intents / use cases covered	Up to 3	Up to 8	Up to 20
Dialogue flow design	1 primary flow	Multi-flow + fallbacks	Complex multi-flow + escalation logic
System prompt / instruction set	Yes	Yes, tuned	Yes, advanced guardrails
Interactive prototype	Text-based demo	Voice prototype	Voice prototype + integration scaffolding
TTS / voice recommendation	Basic	A/B of voices	A/B + SSML tuning
Test script & QA checklist	Short	Full	Full + edge-case suite
Revisions	1	2	3
Handover document	—	Yes	Yes, detailed

How it works

1. Discovery. You share your goal (booking, support, lead qualification, FAQ, etc.), your brand voice, sample calls or transcripts, and any platform you already use. I send a short questionnaire to capture the essentials.
2. Persona draft. I define the voice persona and a handful of sample lines for your sign-off, so we lock tone before building flows.
3. Conversation design. I map intents, happy paths, and fallback/recovery behavior, then write the actual prompt language for each turn.
4. Build. I assemble the system prompt, dialogue logic, and a working prototype on the agreed platform, including a recommended TTS voice.
5. Test and refine. I run the QA script, share the prototype for you to try, and apply your revisions within the plan's scope.
6. Handover. You receive the persona guide, flow spec, prompts, prototype, and a document explaining how to operate and extend everything.

Why choose this

I treat a voice agent as a product, not a chatbot with audio bolted on. Spoken conversation has its own rules: people interrupt, trail off, change their mind, and expect short, natural replies. I design for those realities first, then engineer the prompts and flows to match. You get a persona that feels deliberate, fallback behavior that prevents dead ends, and a prototype you can actually hear and judge before committing to a full rollout. Everything is documented so you are never locked into me to make a small wording change.

Who it's for / use cases

• Startups and SMBs adding an inbound or outbound phone/voice assistant for bookings, support, or lead qualification.
• Product teams prototyping a voice interface for an app, kiosk, or in-car/IoT experience.
• Agencies needing a conversation designer and persona writer for a client engagement.
• Support and ops teams wanting an after-hours FAQ line, appointment confirmation bot, or triage-and-route agent.
• Founders validating a voice-first idea who need a credible, testable demo before raising or hiring.

Typical scenarios: a clinic appointment scheduler, a restaurant reservation line, an order-status hotline, a real-estate qualification agent, or a friendly in-product onboarding voice.

FAQ

Q: Do you build the production system or just the design? I deliver a working prototype plus all design assets. Basic is a text-based demo; Standard and Premium include a voice prototype. Full production deployment, telephony provisioning, and long-term maintenance can be scoped as a separate engagement.

Q: Which platforms and tools do you work with? I commonly work with LLM-based voice stacks (for example Retell, Vapi, LiveKit, or a Twilio + LLM + TTS setup) and major TTS providers. If you already have a backend or preferred vendor, I design and build to fit it. We confirm the platform during discovery.

Q: Can you match a specific brand voice or an existing character? Yes. Share brand guidelines, sample copy, or reference recordings and I will tune the persona, vocabulary, and speaking style to match. I do not clone a specific real person's voice without clear rights and consent.

Q: What languages do you support? Conversation design and prompts are delivered in English by default. I can structure the agent to support additional languages where your chosen TTS and speech-recognition tools allow it; please mention target languages up front so I scope it correctly.

Q: Do you provide the voice audio or voice talent? I recommend and configure synthetic TTS voices and tune their delivery (rate, pauses, pronunciation). I do not record human voice-over; if you want a custom recorded or cloned voice, I will point you to the right path and design around it.

Q: How many revisions are included? Basic includes 1 revision, Standard 2, and Premium 3, within the agreed scope. Revisions cover wording, flow, and persona adjustments; brand-new intents or a platform change beyond the original brief are handled as an add-on.

Q: What do you need from me to start? Your goal and key use cases, brand/tone references, example questions your users ask (or call transcripts if you have them), and details of any existing platform or backend. The more real examples you share, the more accurate the agent.

Q: How do you handle data and privacy? I work with sample/anonymized data where possible and follow any handling rules you specify. I do not store or share your transcripts beyond what is needed to build and test the agent, and I can sign an NDA on request.