Your First AI Conversation — The Complete Botline Launch Manual

This manual walks you from creating your Botline account to your first live AI reply on WhatsApp. There are 15 chapters and 3 appendices. Every paragraph teaches one thing. Every step gives you the exact URL to click.

You can read it linearly or jump to any chapter. Bookmark this page — the URL never changes.

Botline runs entirely in the browser. There is no software to install and no server to configure.

1. Open botline.cc and click Get Started. You land on botline.cc/signup.
2. Enter your business email. Use an address you actually check — every notification (OTP, escalation alerts, weekly digests) is sent here.
3. Click Send code. A six-digit one-time code arrives by email. Botline does not use passwords — every login is OTP-based.
4. Enter your business name. This becomes the label customers may see on receipts and broadcasts. Editable later under Settings → Profile.
5. Pick a plan. Free is enough to complete this guide. Starter unlocks broadcasts, multiple agents, and Draft Review mode. Upgrade in one click later.
6. Land on /dashboard. Bookmark it — this is your home base.

Botline groups businesses into archetypes. Picking yours seeds your account with starter FAQs, a greeting menu, broadcast templates, sample products, payment methods, agent profiles, and workflows tuned for your category — before you have configured anything yourself.

1. Open /dashboard/onboarding.
2. Search by keyword — qurbani, salon, clinic, bakery, boutique, real estate.
3. Click your match. A confirmation modal lists exactly what changes.
4. Confirm country and currency.
5. Click Confirm. The seeding orchestrator runs in the background.

You can switch later. Picking is not a lock-in.

Even if you serve customers in a single country, set this correctly. Reminders fire at the right local time, prices show with the right symbol, and phone numbers validate.

1. Open /dashboard/settings → Profile tab.
2. Confirm Country, Currency, Timezone. Pakistan is Asia/Karachi, India is Asia/Kolkata, UAE is Asia/Dubai.
3. Pick your primary language. The AI defaults to it but always replies in whatever language the customer wrote.
4. Save.

A 2 PM appointment reminder set at the wrong timezone is a cancelled booking. Set this once, correctly.

Your AI agent is only as good as the information you give it. Botline uses RAG (Retrieval-Augmented Generation): every customer question pulls the most relevant chunks of your documents into context before the AI replies.

1. Open /dashboard/knowledge-base. Three tabs: Documents, FAQs, Media.
2. Documents: upload PDFs, paste URLs to crawl, or type directly. Up to 20MB per file.
3. FAQs: aim for 25+ short Q/A pairs — higher signal than long PDFs.
4. Media: upload product photos, brochures, certificates. The AI attaches these in replies when relevant.
5. Toggle Use for AI per document. Off = stored but never used in replies.

Sample shapes: a pricing PDF, 25 FAQs, an About page, individual policy docs (returns, refunds, cancellations, shipping), and a "Things to never say" doc for legally sensitive content the AI must avoid.

The greeting is the first thing every new customer sees. Botline can send a welcome message + numbered menu + optional banner image, video, or PDF with a caption.

1. Open /dashboard/whatsapp/menu.
2. Edit the Welcome message.
3. Add menu items: each has a number key, a label customers see, and a reply that fires when the customer types that number.
4. Use the emoji picker on each label.
5. For a banner image or brochure: upload to Knowledge Base → Media → toggle Send on greeting + add a caption.

Best practice: one banner image (1200×630px, <1MB) and a one-line caption is the sweet spot. A 50-page PDF on first contact gets you blocked.

Botline can run 24/7, but you may want it to behave differently when you are closed.

1. Open /dashboard/settings → Hours tab.
2. Set business hours per day. Enable Closed all day for off days.
3. Pick the out-of-hours behavior: AI replies normally, AI replies with closed message, or AI silent.
4. Customize the closed-message text.
5. Save.

Two providers: Baileys (default, links your existing number via QR, no Meta approval) or Cloud API (Pro+, official, required for high-scale broadcasts).

Baileys (recommended for starting):

1. Open /dashboard/whatsapp.
2. Click + Add WhatsApp Number → Baileys.
3. A QR code appears.
4. On the phone holding the WhatsApp number: WhatsApp → menu → Linked devices → Link a device → scan.
5. Wait 10–30 seconds. Badge flips to ✅ Connected.

If the badge later says Disconnected, click ↻ Reconnect and re-scan.

Cloud API: same page, choose Cloud API instead, sign in to Meta Business, verify the number by SMS, wait for provisioning. See the full Cloud API guide.

Two layers: the system prompt (account-wide brand voice and rules) and the agent prompt (per-agent narrowing).

1. Open /dashboard/settings/ai. In System prompt, write a paragraph: who you are, what you sell, your tone, what you absolutely will not do.
2. Open /dashboard/agents. Click an existing agent. Edit name, avatar, prompt, persona toggle.
3. Toggle the agent Active.
4. Back at /dashboard/settings/ai, toggle AI is replying to On.
5. Save.

Once you grow past a few conversations a day, one generic agent feels limiting. Botline lets you run multiple specialised agents and route conversations to the right one.

1. Open /dashboard/agents. Click + New agent.
2. Name, avatar, narrow prompt: Sales qualifies and closes; Support handles tracking and returns; Booking collects appointment details; Payment Verifier reads receipt screenshots.
3. Assign a knowledge-base subset — only the docs this agent should see.
4. Set the routing rule (keywords, intent, tag).
5. Save and toggle Active.

More agents do not cost more. Every reply uses one agent — routing chooses.

Three modes:

• Off — AI does not reply. Every message lands in your inbox.
• Draft Review (AI Beta) — AI generates a reply but does not send. The reply appears as a gold dashed bubble with Approve · Edit · Reject. Customer waits.
• Live — AI sends automatically. Confidence-based: low-confidence replies still queue as drafts.

Draft Review is the seatbelt. The AI does the work; you stay the final author. The customer never sees a wrong word.

1. /dashboard/settings/ai → Reply mode radio → pick Draft Review.
2. Slider sets confidence threshold (default 0.85). Below threshold, even Live mode queues drafts.

Day 1 (Mon): 8 drafts arrive. Approve 6 quickly, edit 1 to remove an outdated price, reject 1 that mentioned a competitor. Each takes 4 seconds.

Day 2: add a "Things to never say" doc. Reject rate drops to zero.

Days 3–4: trust grows. Raise threshold to 0.90. Glance and approve.

Day 5: move three intents to Live (greeting, hours, order tracking) via routing rules. Drafts arrive only for harder conversations.

End of week 1: 5–8 drafts a day, 5 minutes total. The AI handles the rest.

Anti-pattern: approving every draft without reading is worse than turning Draft Review off. Read each one.

Send "Hi" to your connected WhatsApp number from another phone. Watch /dashboard/conversations.

Test these scenarios:

• Greeting — first contact. Menu fires? Media attaches?
• Menu number — type "1". Right reply?
• Common question — ask about your bestseller.
• Curveball — ask something not in your KB. Does it gracefully say "I am not sure"?
• Rude message — "you are useless". Stays calm? Escalates?
• Competitor mention — redirects?
• Off-topic — pivots back?

If any fail, fix the prompt or KB before going Live.

If you have done chapters 1–12, this chapter is one click.

1. /dashboard/settings/ai → Reply mode → Live (or keep Draft Review another week).
2. Confirm AI is replying is On.
3. Save.

That is it. Every WhatsApp message that arrives is now read, KB-matched, agent-routed, and answered (or drafted) within 1–3 seconds.

For the first 24 hours, skim 5–10 random replies every couple of hours. If anything is off, hit Take over, type a fix, and update the KB or prompt at the end of the day.

The conversation page at /dashboard/conversations is where you spend most of your time once live.

Three columns: conversation list (left) · thread (middle) · customer info + actions (right).

The actions:

• Take over — pauses AI for this conversation only.
• Resume AI — re-enables it.
• Tags — vip, complaint, lead, paid, wholesale. Filter the inbox by tag.
• Lead score — auto 0–100, override manually.
• Move to — reroute to a different agent or staff member.
• Pin — keep at the top of the inbox.
• Unsend (Pro+) — recall any sent message within 48 hours.
• Drafts — gold dashed bubble → Approve / Edit / Reject.
• Internal note — private to staff.

You cannot stare at the inbox 24/7. Botline pings you when needed.

1. Open /dashboard/settings/notifications.
2. Add channels: Email (always on), WhatsApp (Pro+), Slack (Incoming Webhook), Telegram (link a bot).
3. Toggle events on/off:
   • AI escalated to human — anger, refund, low confidence.
   • Draft pending > 30 min.
   • VIP customer messaged.
   • Disconnected — WhatsApp link dropped.
   • Daily digest at 8 PM local time.

Multi-tier escalation: AI → Staff → Engineering / Ops (Pro plan, for billing or platform-level issues, via the Escalate to Engineering button).

Tick each item. When all are checked, your first AI conversation is live.

[ ] Sign up at botline.cc/signup with business email
[ ] Verify OTP, land on /dashboard
[ ] Pick business profile at /dashboard/onboarding
[ ] Confirm country, currency, timezone
[ ] Upload at least 1 PDF + 25 FAQs to /dashboard/knowledge-base
[ ] Edit greeting menu at /dashboard/whatsapp/menu
[ ] Upload banner image, toggle "Send on greeting"
[ ] Set business hours under /dashboard/settings → Hours
[ ] Connect WhatsApp at /dashboard/whatsapp (QR scan)
[ ] Write system prompt at /dashboard/settings/ai
[ ] Edit seeded agent's prompt at /dashboard/agents
[ ] Activate the agent
[ ] Set Reply Mode to "Draft Review"
[ ] Test by sending "Hi" from another phone
[ ] Review and approve / edit / reject 5 drafts
[ ] Configure escalation notifications
[ ] Once comfortable, set Reply Mode to "Live"

Welcome to Botline. You are now live.