# Bots That Understand Your Customers

> Deploy conversational AI bots with natural language understanding, context retention, and multi-channel support. Handle unlimited conversations simultaneously with seamless escalation to humans when needed.

Canonical: https://bots.do/

## In code

```ts
import { Bot } from 'bots.do'

// Handle customer questions with AI
on.Message.received($ => {
  const answer = $.Bot.search.Knowledge({ query: $.data.text })

  if (answer.confidence > 0.8) {
    $.Bot.send.Reply({ to: $.user, text: answer.response })
  } else {
    $.Bot.escalate.toHuman({ reason: 'low_confidence' })
  }
})

// Deploy across channels
Bot.deploy({ slack: true, discord: true, webChat: true })
```

## Frequently asked questions

### How accurate are bot responses compared to human support?

Bots achieve 85-90% first-contact resolution for common questions vs 95%+ for humans, but handle unlimited conversations simultaneously. Average accuracy: 88% for factual questions (docs, pricing, features), 75% for troubleshooting (multi-step problems), 60% for empathy-requiring situations (complaints, refunds). Bot+human hybrid achieves 98% resolution: bots handle routine queries, escalate complex cases. Response time: <500ms (bot) vs 2-5 minutes (human).

### Can bots handle conversations in multiple languages?

Yes. Auto-detect language and respond in 95+ languages using multilingual models (GPT, Claude). Translation accuracy: 95%+ for common languages (Spanish, French, German), 85%+ for others. Maintain conversation context across language switches. Localized personality—formal tone for German, casual for English. Support regional dialects and slang. Custom training for domain-specific terminology in each language.

### How do bots learn from conversations?

Continuous learning through reinforcement learning from human feedback (RLHF), A/B testing responses, and supervised fine-tuning on successful conversations. Track metrics: resolution rate, user satisfaction, escalation rate, response time. AI identifies patterns—which answers work, common failure modes, emerging topics. Human review of flagged conversations (low confidence, negative feedback) trains models weekly. Improvement rate: 10-15% resolution improvement over 3 months.

### What happens when the bot doesn't know the answer?

Multi-layered fallback strategy: (1) Search extended knowledge sources (external docs, Stack Overflow, community forums), (2) Ask clarifying questions to narrow intent, (3) Provide related articles that might help, (4) Escalate to human with full context. Never make up answers—confidence thresholds prevent hallucinations. Transparency: 'I'm not sure, but here's what I found' or 'Let me connect you with a human specialist.' Average escalation rate: 15-20% of conversations.

### Can bots integrate with existing support tools?

Yes. Native integrations with Zendesk, Intercom, Freshdesk, HubSpot, Salesforce Service Cloud, and 50+ platforms. Sync conversation history, customer data, and ticket status bidirectionally. Bots can create tickets, update customer profiles, check order status, process refunds (with approval), and schedule calls. Embedded web chat widgets, Slack app, Discord bot, WhatsApp Business API, Twilio SMS. Full API for custom integrations.

## Part of Platform.do

bots.do is one of the primitives of Platform.do, in the Agents layer (reason, research and act). Every .do shares one grammar, Business-as-Code, and one SDK.

- What is Business-as-Code? https://platform.do/business-as-code
- Platform.do: https://platform.do/
- Docs: https://docs.platform.do