AI Support Bot for Mobile Apps with RAG

Support teams handling mobile app inquiries often spend 70% of their time answering repetitive questions. We implement RAG-based bots that resolve up to 80% of typical queries in 2 seconds, cutting first-line load by 60%. Our approach relies on Retrieval-Augmented Generation (RAG), combining a vector knowledge base with an LLM. This yields accurate answers based on your documentation, avoiding hallucinations. Accuracy improves from 65% to 92% compared to pure search. With over 5 years of mobile development experience and 30+ AI support projects launched, we guarantee quality and transparency at every stage—from knowledge base audit to deployment on App Store and Google Play.

RAG outperforms simple keyword search by 3x in answer accuracy. Operators save up to 40 hours per week per 1,000 inquiries. Return on investment averages 3 months. According to our analysis, RAG reduces the cost per inquiry by $0.50, saving $5,000 monthly at 10,000 inquiries. For large clients handling 50,000 inquiries monthly, savings reach $25,000 per month.

How RAG Works in a Support Bot

The classic flow: the knowledge base (articles, documentation, resolved tickets) is split into semantic chunks, indexed in a vector database. On a new query, we retrieve the top-5 relevant chunks and pass them as context to the LLM. The model generates a response strictly based on that data.

Our stack, proven on projects with up to 10,000 daily requests:

Knowledge base (Confluence, Notion, MDX files)
    ↓ Chunking + Embedding (text-embedding-3-small / BGE-m3)
Qdrant / pgvector
    ↓ Semantic search (top-5 chunks)
GPT-4o mini / Claude 3.5 Haiku
    ↓ Answer generation with context
Mobile client

Critically: chunks must be semantic, not mechanical 500-character slices. Cutting an article mid-paragraph loses context. We use RecursiveCharacterTextSplitter with delimiters on headings and paragraphs.

What Happens When the Bot Doesn't Know?

The bot never feigns omniscience. If confidence drops below a threshold (e.g., 0.7), it escalates to a human operator. The handoff is transparent: the user sees a "Connecting to a specialist" message, and the dialogue history is forwarded via webhook.

Integrations for live chat: Zendesk Chat API, Intercom, AmoCRM, or a custom ticket system. A key pattern: the bot continues working in the background while waiting for an operator. If the queue is long, the bot retries finding an answer or suggests helpful articles.

Ticket Classification and Prioritization

If the bot cannot resolve an issue, it classifies the query before handoff:

CATEGORY_PROMPT = """
Classify the user's inquiry into a category:
- billing (payment, invoice, refund)
- technical (error, feature not working)
- account (account, password, access)
- other

Return only the category name, no explanation.
"""

async def classify_ticket(user_message: str) -> str:
    response = await openai_client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": CATEGORY_PROMPT},
            {"role": "user", "content": user_message}
        ],
        temperature=0
    )
    return response.choices[0].message.content.strip()

The category determines the queue and priority. temperature=0 ensures deterministic output.

UI Features for Support

Attachments. Users can upload screenshots and videos. On iOS—PHPickerViewController with media type restrictions; on Android—ActivityResultContracts.GetContent(). Files are uploaded to S3/Cloudinary, with a preview inserted into the chat.

Ticket status. If a ticket is created, the user sees its number and can track it through the same bot: "Status of my ticket #12345".

Response rating. After each bot response, a thumbs up/down prompt appears. Data flows into analytics and helps improve the knowledge base. Get a consultation on your project to assess automation potential.

What's Included

Timeline Estimates

Discuss your project with us to choose the optimal architecture. Contact us to order a turnkey support bot with quality guarantee. We guarantee transparency at every stage—from initial consultation to app store release.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.