Features
Answers from your documents. A person ready when it can't.
The assistant answers from your uploaded docs, website and FAQs, on your website widget and on WhatsApp. When it can't find a supported answer, it says so and hands the chat to your team.
Every question. The same checks.
Before a reply is sent, each question passes through the same steps. That is why the assistant answers from your facts, and hands over when it does not have them.
01 Work out what the customer wants
The question is sorted by topic and sent to the right agent: support, sales, billing, or straight to your team.
02 Find the facts in your documents
It searches the files, manuals and FAQs you uploaded for the passages that match the question.
03 Check the answer is supported
If the match is below the confidence threshold, it says it could not find an answer and offers a person instead of guessing.
04 Run the safety rules
Server-side rules filter off-topic requests, prompt injection attempts and personal details before anything is sent.
05 Stream the reply
The answer is written to your customer word by word, with the source it came from.
For your technical team
- Answer flow
- Multi-step, with fallback paths
- When unsure
- Human handoff
- Knowledge index
- pgvector on PostgreSQL
- Keys, secrets, files
- AES-256-GCM
- Webhook signing
- HMAC-SHA256
- Safety rules
- Server-side
The right question. The right agent.
Send each question to the agent that knows the answer, with its own instructions and documents, and keep the whole conversation in view.
Agents per department
Sales, support and billing each get their own assistant.
Set up separate agents for each part of your business. Each one has its own documents, instructions and tone of voice, so a billing question is never answered from a sales brochure.
- Custom instructions per agent
- Keyword and intent routing
- Per-department knowledge bases
- Turn an agent on or off in one click
Intent triage
Incoming questions are sorted by topic and sent to the right agent or to the human queue.
- Intent detection
- Quick path for greetings
- Topic categories
- Fallback routing
Conversation context
The assistant keeps the conversation history in view, so customers do not have to repeat themselves.
- Conversation history in every reply
- Customer context across sessions
- Source tracking on answers
Your files are the source. No answer means no guess.
The assistant answers from what you upload, and says so when it cannot find an answer. Your customers get facts from your own policies.
Your documents as the source
Upload your files or paste your website link.
Add PDFs and Word files, or point it at your website. The assistant searches only your content when it answers, using semantic search over your documents stored in PostgreSQL.
- PDF, DOCX and web page indexing
- Semantic search over your content
- Per-agent or shared documents
Confidence thresholds
When the match is weak, it says so and offers a person.
If the match with your documents is too weak, the assistant does not guess. It tells the customer it could not find an answer and offers to bring in your team. Supported answers carry a citation to the page they came from.
- Answers only from your documents
- Source citations on answers
- Human handoff when unsure
The AI answers first. Your team steps in.
Replies stream as they are written, on your website and on WhatsApp. When a customer asks for a person, or the assistant cannot find a supported answer, your team takes over with the full chat history.
Human takeover
Claim a conversation and reply yourself.
The conversation is locked on the server and moved to your team console. Your agents can watch live chats, claim one in a click, reply directly, and hand it back to the AI when they are done.
- Claim a conversation in one click
- Live conversation viewer
- Hand the chat back to the AI
- Resolved conversations are logged
WhatsApp support
The same assistant, where your customers already chat.
Run the assistant on WhatsApp through the Meta Cloud API. Customers send a text message, the answer is found in your documents, and the reply comes back in the chat.
- Text messages
- Quick-reply buttons
- Human handoff on WhatsApp
Meta WhatsApp Cloud API
Streaming replies
Answers stream word by word, so customers see the reply while it is being written instead of waiting on a blank screen.
- Word-by-word streaming
- Typing indicators
- Works on mobile and desktop
Checked on the server. Before your customer sees it.
Safety rules run on the server before a reply reaches your customer. Off-topic requests, prompt injection attempts and personal details are filtered out, and stored secrets are encrypted.
Server-side guardrails
Rules run before and after the model answers.
Guardrails block prompt injection, filter off-topic requests, mask personal data and keep replies in your brand voice. When a customer asks for a person, the chat is handed to your team. Every check is written to an audit log you can review.
- Personal data masking
- Brand voice and safety rules
- Human handoff when the customer asks for a person
- Audit logs for review
Encryption and privacy
Stored API keys, secrets and uploaded knowledge files are encrypted with AES-256-GCM. Your data is never used to train public AI models.
- AES-256-GCM for keys, secrets and files
- Role-based access control
- Not used to train public models
Learning from your team
When a person on your team answers a question the assistant could not, that answer can be saved to the knowledge base for next time.
- Q&A saved from resolved chats
- Duplicate questions are skipped
- Shared across your agents
See how it is going. Plug it into your tools.
See how conversations go, pick the AI model you want to pay for, and send conversation events to the systems your team already uses.
Analytics and CSAT
Satisfaction, resolutions and reply times in one view.
Track customer satisfaction ratings, how many conversations were resolved and how quickly customers got their first reply, so you know which topics need a better document.
- CSAT ratings
- Resolution stats
- First response time
Your choice of model
On the Premium plan, bring your own API keys for OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Groq, or any OpenAI-compatible endpoint. Your keys are stored with AES-256-GCM encryption.
- OpenAI-compatible providers, including Groq and Mistral
- Bring your own keys (BYOK) on Premium
- Automatic fallback to another provider when one fails
- OpenAI
- Claude
- Gemini
- DeepSeek
- Groq
- Mistral
Webhooks and REST API
Send conversation events and escalations to your CRM, help desk, or any tool that accepts webhooks, as they happen. Each delivery is a signed HMAC-SHA256 HTTP POST your system can verify.
- Events: conversation created, message received, handoff, closed
- HMAC-SHA256 signed webhooks
- Works with any tool that accepts webhooks or a REST call
- Built-in test events and delivery stats
Away from your desk? The inbox comes with you.
LayBuild AI for iPhone and Android is in development. It brings the team inbox to your phone, so a chat the assistant hands off can be answered from anywhere.
What the apps will do
Claim handed-off chats
The same queue as the web inbox, with the full conversation and the reason the assistant stepped back.Reply to website and WhatsApp chats
Your answer goes out on the channel the customer wrote from.
Needs a human · 3
- Refund for damaged itemWhatsApp · 2mClaim
- Change billing emailWebsite · 6m
- Bulk order pricingWebsite · 11m
So sorry about that. Sending a replacement today.
Asked before you sign up.
Straight answers about accuracy, setup and your data. Anything else, our team is a message away.
How does LayBuild AI avoid made-up answers?
How does LayBuild AI avoid made-up answers?
LayBuild AI searches your uploaded files (PDF, DOCX, help center articles) and your website with semantic search over PostgreSQL vector embeddings, and answers from what it finds. If the information is missing or the match is below the confidence threshold, it says it could not find an answer and offers a human agent instead of guessing.
Can LayBuild AI process customer voice notes on WhatsApp?
Can LayBuild AI process customer voice notes on WhatsApp?
No. LayBuild answers text messages on WhatsApp. Voice notes, images and videos are not transcribed or read, so customers need to type their question.
How does live human agent handoff work?
How does live human agent handoff work?
When a customer asks for a person, or the assistant cannot find a supported answer, the conversation is locked on the server and moved to your team console. Your agents see the full chat history and can reply directly or hand the chat back to the AI.
Can we bring our own AI models (BYOM)?
Can we bring our own AI models (BYOM)?
Yes, on the Premium plan. LayBuild AI works with OpenAI, Anthropic Claude, Google Gemini, DeepSeek, and OpenAI-compatible API endpoints, using your own API keys. Stored keys are encrypted with AES-256-GCM.