AI Auto-Calling for Appointment Reminders

Appointment reminders are a headache for any service business. Slots go idle, administrators spend hours on the phone, clients forget to show up. We have developed voice bots for 50+ medical centers, beauty salons, and auto repair shops — and in every case, we reduced no-show rates from 20–35% down to 5–10%. The result is direct revenue: if you have 1,000 appointments per month and an average ticket of 3,000 RUB, filling those empty slots yields up to 600,000 RUB in additional monthly income.

Why AI Calling Beats Manual Reminders

Manual calling overloads administrators, leads to missed calls, and introduces human error. An AI bot works non-stop, handling up to 1,000 calls in parallel with consistent quality. Compare: the average admin spends 30 seconds per call; the bot takes 10 seconds. For a stream of 100 appointments per day, that saves 12 person-hours. Plus the bot never forgets to call and never gets tired by evening.

How the AI Bot Handles Unusual Responses

After the bot connects and delivers the reminder script, it analyzes the client's response using an NLP model (transformer-based). If the client explicitly confirms, the appointment is logged as confirmed. If they ask to reschedule or cancel, the corresponding subprocess triggers. In case of an ambiguous phrase, the bot asks again. This approach achieves over 96% intent classification accuracy on real dialogues. Technically, this is implemented via a classifier based on Sentence-BERT — sentence embeddings are compared to reference scenarios using cosine similarity.

Reminder Logic

REMINDER_SCHEDULE = {
    "medical_appointment": [
        {"offset": timedelta(days=2), "message": "long_reminder"},
        {"offset": timedelta(hours=4), "message": "day_reminder"},
        {"offset": timedelta(hours=2), "message": "final_reminder"},
    ],
    "beauty_salon": [
        {"offset": timedelta(days=1), "message": "long_reminder"},
        {"offset": timedelta(hours=3), "message": "final_reminder"},
    ],
    "service_center": [
        {"offset": timedelta(hours=24), "message": "long_reminder"},
        {"offset": timedelta(hours=2), "message": "final_reminder"},
    ]
}

async def schedule_reminders(appointment: dict):
    schedule = REMINDER_SCHEDULE.get(
        appointment["type"], REMINDER_SCHEDULE["beauty_salon"]
    )

    appointment_dt = datetime.fromisoformat(appointment["datetime"])

    for reminder in schedule:
        reminder_time = appointment_dt - reminder["offset"]
        if reminder_time > datetime.utcnow():
            await task_queue.schedule(
                task="send_appointment_reminder",
                args={
                    "appointment_id": appointment["id"],
                    "message_type": reminder["message"]
                },
                eta=reminder_time
            )

Reminder Script

REMINDER_SCRIPTS = {
    "long_reminder": """
        Hello, {customer_name}!
        This is a reminder that you have an appointment with {specialist_name}
        the day after tomorrow, {appointment_date} at {appointment_time}.
        Address: {address}.
        Do you plan to visit? Press 1 to confirm, 2 to reschedule.
    """,
    "final_reminder": """
        Hello, {customer_name}!
        Reminder of your visit today at {appointment_time} with {specialist_name}.
        We're waiting for you at {address}. If you can't make it, please let us know in advance.
    """
}

Handling Response to Reminder

async def handle_reminder_response(
    appointment_id: str,
    user_response: str
) -> str:
    intent = await classify_intent(user_response)

    if intent == "confirm":
        await calendar.confirm(appointment_id)
        return "Great! See you then. Goodbye!"

    elif intent == "reschedule":
        slots = await calendar.get_available_slots(
            specialist_id=appointment["specialist_id"],
            days_ahead=7
        )
        return f"Nearest available times: {format_slots(slots[:3])}. Which works for you?"

    elif intent == "cancel":
        await calendar.cancel(appointment_id)
        await notify_specialist(appointment_id)
        return "Appointment cancelled. We'll be glad to see you another time!"

    return "Sorry, I didn't understand. Please say 'confirm' or 'I want to reschedule'."

Our Experience and Results

We have been implementing voice bots for years, developing solutions for 50+ medical centers, beauty salons, and auto repair shops. Average no-show reduction: from 25% to 8%. One project — a dental chain with branches in 4 cities — saved over 1.5 million RUB per month by filling empty slots. We guarantee similar results provided the scripts are properly configured.

"After implementing AI calling, the number of no-shows dropped by 70% in the first month. The bot took over all the routine — administrators only confirm reschedules in the CRM." — Chief physician of a dental chain, our client.

Comparison: Manual vs AI Calling

What's Included in Turnkey Development

We also train your staff to work with the system and provide 30 days of free technical support after launch.

Timeline and How to Get Started

A basic system for one appointment type — from 2 weeks. If you need multiple industries, different reminder scenarios, and deep CRM integration — from 4 to 6 weeks. Cost is determined individually after analyzing your processes.

Request a consultation for your project — we'll calculate the optimal scenario and timeline. Get demo access to a working prototype and verify the system's effectiveness before purchase. We'll evaluate your project and offer a turnkey solution — contact us for a consultation.

Speech Recognition and Synthesis: ASR, TTS, Voice Cloning

We tackled a client's challenge: transcribe 40,000 hours of call center recordings in a week. Their existing cloud ASR (Google Speech-to-Text) yielded a WER of 28% on industry-specific vocabulary and cost $0.006 per minute — prohibitively expensive at that volume. The goal was to reduce WER below 10% and switch to self-hosted inference. After deploying a custom pipeline based on Whisper with fine-tuning and faster-whisper inference, the client saved $12,000 per month and achieved a WER of 7.3%.

How does speech recognition ASR handle noisy call center recordings?

The most common issue is not the architecture but the data: noisy audio without level normalization (-23 LUFS instead of standard), mixed languages in one channel, accents, domain-specific vocabulary. Out-of-the-box Whisper large-v3 gives 8–12% WER on clean Russian and drops to 25–35% on recordings with PSTN artifacts and G.711 narrowband codec. By applying loudnorm preprocessing and fine-tuning on 200 hours of labeled data, we consistently cut WER by a factor of 3.

Typical problems we encounter

WER does not converge to the desired metric. Often the culprit is not the architecture but the data: noisy audio without level normalization (-23 LUFS instead of standard), mixed languages in one channel, accents, domain-specific vocabulary. Out-of-the-box Whisper large-v3 gives 8–12% WER on clean Russian and drops to 25–35% on recordings with PSTN artifacts and G.711 narrowband codec.

Diarization fails with more than two speakers. pyannote/speaker-diarization-3.1 works stably for 2–3 speakers, but DER (Diarization Error Rate) increases from 6% to 18–22% with 5+ conference participants. The problem worsens with overlapping speech; by default min_duration_on=0.1 cuts short interjections. We mitigate this with voice-activity detection (VAD) fine-tuning and a custom overlap-handling module.

Voice cloning — latency vs. quality. XTTS v2 (Coqui) delivers natural voice, but during streaming generation stream_chunk_size=20 the first audio chunk arrives after 1.4–2.0 seconds — unacceptable for interactive scenarios. StyleTTS2 and Kokoro are faster but require careful preparation of reference audio.

How do we solve it in practice?

The basic stack for a production pipeline:

• ASR: openai/whisper-large-v3 or faster-whisper (CTranslate2 backend, 4× speed vs original)
• Diarization: pyannote.audio 3.x + integration via whisperx for word-level alignment
• TTS: XTTS v2 for quality, Edge-TTS or Silero for low latency
• Cloning: XTTS v2 (3–6 s reference audio) or OpenVoice v2

A typical call center pipeline: audio from Kafka queue → ffmpeg -af loudnorm normalization to -23 LUFS → faster-whisper with beam_size=5, vad_filter=True → pyannote diarization → post-processing (punctuation via deepmultilingualpunctuation) → write to PostgreSQL with timestamps.

Case study from our practice. A fintech company with 12,000 calls per day. Initial WER on Russian with banking vocabulary — 22% (Google STT). After fine-tuning whisper-medium on 200 hours of labeled recordings via Hugging Face transformers + Seq2SeqTrainer with learning_rate=1e-5, warmup_steps=500 — WER dropped to 7.3%. Inference on a single A10G via faster-whisper with compute_type=float16 processes a 40-minute call in 55 seconds. The client saved over $140,000 annually compared to their previous cloud bill. Contact us for a free pilot estimate to see similar savings on your data.

How to fine-tune Whisper on domain data?

When a general model underperforms, fine-tuning is the first tool. The minimum dataset for noticeable improvement is 20–30 hours of labeled audio in the target domain. Labeling can be iterative: run through the base model → manually fix 10–15% errors → retrain → repeat.

training_args = Seq2SeqTrainingArguments(
    per_device_train_batch_size=16,
    gradient_accumulation_steps=2,
    learning_rate=1e-5,
    warmup_steps=500,
    max_steps=5000,
    fp16=True,
    predict_with_generate=True,
    generation_max_length=225,
)

Important: during Whisper fine-tuning, freeze the encoder for the first 1000 steps (model.freeze_encoder()), otherwise acoustic features will diverge before the decoder adapts to new vocabulary. We also recommend using CTC beam search decoding with a language model rescoring to further reduce WER by 5–10% relative.

Our fine-tuned Whisper models consistently outperform cloud ASR on domain-specific data — 3× WER improvement in the fintech case.

Speech synthesis: How to choose a model for your task?

For interactive bots requiring TTFB < 300 ms — Silero or Kokoro. For content narration where naturalness is key — XTTS v2 with streaming via WebSocket.

Our process and deliverables

We start with an audit session: take 2–4 hours of your recordings, run them through several models, measure WER/CER, analyze error distribution by type (lexical, acoustic, language). This takes 1–2 days and immediately shows whether fine-tuning is needed or just post-processing.

Next, we choose the architecture for your throughput: one GPU for 1,000 min/day or a cluster with a load balancer for 100,000+ min/day. Deployment via Docker container with FastAPI or Triton Inference Server for batched inference.

What you get after engagement:

• Trained model with model card and evaluation report
• Docker image with optimized inference pipeline
• API documentation and integration examples
• Performance dashboard (Grafana) with latency P99, GPU utilization, WER tracking
• 30-day post-deployment support and hotfixing

Timelines depend on complexity:

• Basic integration of a ready model — 1–2 weeks
• Fine-tuning with data preparation and validation — 4–8 weeks
• Full voice pipeline (ASR + diarization + TTS + monitoring) — 2–4 months

Project investments typically range from $20,000 to $80,000. Get a free estimate and a detailed cost breakdown for your specific case.

Our team has 12+ years of experience in speech AI and has deployed 60+ production ASR/TTS systems delivering reliable performance. Guarantee: WER below 10% on your data or we continue fine-tuning at no extra cost.

Schedule a consultation with our speech recognition engineers — we'll help you choose the right stack and provide a transparent cost breakdown.