Sales Agent
Product questions, pricing, and recommendations

Every unanswered “what’s the price?” is a buyer deciding whether to wait or walk away. That is exactly the gap an AI agent for sales is built to close.
Still waiting on a reply.
Answered instantly, 24/7 sales coverage on duty.
Three steps, and conversational AI for sales is answering real buyers.
Bring in WhatsApp, your website widget, and your store or CRM — plus any AI agent integrations you already rely on, like an HTTP API or CRM webhook.
Describe the agent’s sales role in plain language, then drop in actions using ##action_name## : id — type ## to autocomplete labels, sequences, and connected stores.
Your AI sales agent starts answering, recommending, tagging, and routing conversations immediately — no code, no waiting.
Full walkthrough: How to create and configure AI agents in BotSailor
An AI-powered sales agent is only as good as what it can actually do. Here is what runs under the hood.
Unlike a rigid keyword bot, this AI sales agent reads what a buyer actually means and decides the right next step — answer, recommend, qualify, or hand off to a closer.
##shopify_recommend_products##Bind Shopify or WooCommerce and the agent searches stock, price, and variants — then can send the matching product photo from Knowledge Sources.
##woocommerce_recommend_products##When a buyer is ready, the agent labels the contact and enrolls them in a follow-up sequence so your CRM stays organized without manual tagging.
##add_label## : 12Round-robins across a sales role, or assigns one specific teammate, the moment a conversation needs a person — straight into your shared inbox.
##assign_human_role## : 3
Written directly inside the System Prompt as ##action_name## : id — type ## to autocomplete. The interface shows readable names; the prompt stores the numeric ID underneath.
##add_label## : 12Tag a hot lead, qualified buyer, or product interest.
##remove_label## : 12Remove a label when the status changes.
##assign_sequence## : 8Enroll the contact in a sales follow-up sequence.
##save_custom_field## : 42Save budget, company size, or plan tier from the chat.
##shopify_store## / ##woocommerce_store## : 12Bind a connected store before any catalog action. See Shopify on WhatsApp or WooCommerce on WhatsApp.
##shopify_recommend_products##Search and recommend products from the live catalog.
##woocommerce_recommend_products##Same live recommendation flow for WooCommerce.
##shopify_get_order_details##Look up an order if a buyer asks mid-sale.
##call_http_api## : 5 | reply_with_response | status, messageCall a CRM or pricing API, then reply using selected fields.
##call_http_api## : 5 | reply_with_static_replyCall an API, then send a fixed buyer-facing reply.
##call_http_api## : 5 | no_replyPush the lead to your CRM silently — no extra message.
##assign_human_member## : 335Always assign to one specific closer.
##assign_human_role## : 3Round-robin across a sales team role.
##trigger_bot_flow## : 15Trigger a catalog, checkout, or booking flow from the visual flow builder.
##transfer_agent## : 18Hand off to Support or Billing without the buyer repeating themselves.
##mark_as_solved##Close the conversation once the question is answered.
##skip_reply##Stay silent on vague or off-topic messages.
##block_subscriber##Block the subscriber if the message is spam.
##follow_up## : 24Set an internal staff reminder, in hours.
##add_note## : Ready for a demo this weekSave a note in the shared inbox — never sent to the buyer.
##remove_sequence## : 8Remove a previously assigned sequence.
##shopify_store## first; WooCommerce actions require ##woocommerce_store## first — pick a connected store, or those tags will not save. Image URLs do not belong in the prompt; add product and gallery images under Knowledge Sources (Media) instead, and the agent will send them when asked. See Shopify WhatsApp AI Agent or WooCommerce WhatsApp AI Agent.Actions tell the agent what it can do. Knowledge Sources tell it what it knows — six ways to feed catalog copy, pricing, and objection handling into a reusable Knowledge Set.
Paste offer copy, pricing notes, or objection-handling scripts the agent should know verbatim.
Point to a live product, pricing, or docs page — the agent reads it directly as a source.
Upload PDFs, catalogs, and spreadsheets — the agent extracts and references the content inside them.
Attach product photos here, not in the prompt — the agent sends the right image the moment a buyer asks.
Connect a live spreadsheet — SKUs, prices, or FAQs — and updates there show up in answers automatically.
Pull in dynamic catalog or CRM data as background knowledge — separate from ##call_http_api##, which is for real-time actions.
Build a Knowledge Set once — like “Catalog & pricing” — attach content, URLs, files, media, sheets, and API feeds to it, then reuse that same set across every sales agent that needs it. Update a source once, and every agent using that set answers with the change immediately.

A single agent handling pricing, refunds, and technical issues gets muddled fast. Specialize instead — conversational AI for sales works best when the Sales Agent has one clear job.
Product questions, pricing, and recommendations
Tickets, order issues, refunds, complaints
Invoices, payment failures, plan changes
Routing rules under Agent Configuration decide which agent answers a given message. Mid-conversation, an agent can hand off with ##transfer_agent## : 18 — for example, a Sales Agent transferring a billing dispute straight to Billing, without the buyer repeating themselves.
The pieces that turn a demo into AI sales software you can run a real pipeline on — the same tools that keep conversational selling accurate as your catalog and offers change.
Catalog copy, pricing docs, and product media live here. The agent answers from them directly and can attach the right image when a buyer asks for photos.
Tag conversations and enroll contacts into follow-up sequences automatically, so your CRM stays organized — and those same segments feed WhatsApp broadcasting campaigns.
##save_custom_field## captures structured details — budget, plan tier, preferred language — straight from the conversation.
Connect any CRM or pricing system with ##call_http_api##, choosing whether the buyer sees the live response, a fixed reply, or nothing at all.
Inbound volume keeps climbing while headcount stays flat. The research below shows where an AI sales agent actually closes that gap — and the one thing that separates agents that qualify from bots that only reply.
The distance between a canned “thanks for your interest” and a 21× lift is action-taking. An agent that can only retrieve an answer deflects; one that can recommend a product, write a custom field, and label intent qualifies. That is the line BotSailor AI agents are built on.
On WhatsApp the expectation is tighter than email. Instant first replies — plus WhatsApp broadcasting for proactive offers — keep you inside that window without a night shift.
Buyers still want a human for high-value deals. The metric that matters is not “no humans” — it is how cleanly the agent hands off. A shared inbox with notes attached means the closer picking up never starts cold.
Sources: Gartner B2B buying research; Salesforce State of Sales; InsideSales / Harvard Business Review lead-response studies; Infobip messaging benchmarks. Figures are industry-wide benchmarks, not BotSailor performance claims — your own results depend on offer mix and how deeply the agent is integrated.
Everything people ask before deploying an AI sales agent across their team — sorted by topic.
An AI sales agent is conversational AI for sales that reads what a buyer means — not just the words they type — and decides what to do about it: answer from your catalog, recommend a product, save a budget or company size, label a hot lead, or hand the conversation to a closer.
A basic chatbot matches fixed keywords to canned replies. This is conversational AI for sales — it holds context, understands intent, and can take real actions like tagging a contact, recommending from a live store, calling a CRM API, or escalating to a human, instead of returning a scripted line.
It combines sales conversation with operational actions — product recommendations, CRM tagging, sequences, HTTP API calls, and human handoff — in one system prompt. That is what makes it sales AI software teams can run on production inbound, not just a demo.
The same underlying agent framework can be configured as a Support Agent, a Billing Agent, or a Lead Qualification Agent. Behavior comes from the system prompt and the actions you give it, not from separate software.
It is genuine AI-powered selling — the agent interprets open-ended messages and decides its own next step, rather than following a fixed decision tree. Scripted sequences (like an abandoned-cart reminder) can run alongside it, but the agent itself is reasoning about each conversation.
Yes. That consistency is the point of a 24/7 AI sales agent. Human assignment still respects the business hours you configure, but the agent itself answers, qualifies, and tags conversations around the clock.
Write them directly in the prompt using ##action_name## : id. Type ## to trigger autocomplete and pick a label, sequence, store, or connected resource — the interface shows readable names, but the prompt stores the numeric ID underneath. The complete AI agent configuration guide covers each action with examples.
Bind a store with ##shopify_store## or ##woocommerce_store##, then use ##shopify_recommend_products## or ##woocommerce_recommend_products##. The agent searches live catalog data — price, stock, variants — instead of guessing from a static list.
##save_custom_field## : 42 writes a value to a custom field on the contact — useful for capturing budget, company size, preferred plan, or product interest directly from the conversation.
Instruct the agent to use ##add_label## when purchase intent is clear, then ##assign_sequence## to enroll that contact in a sales follow-up. Those same segments feed WhatsApp broadcasting campaigns.
reply_with_response replies using specific fields from the API’s response; reply_with_static_reply sends a fixed message regardless of what the API returns; no_reply calls the API silently, with nothing sent to the buyer. Use this to push a qualified lead into your CRM in the background.
Yes — ##add_note## : Ready for a demo this week saves an internal note in the shared inbox, and ##follow_up## : 24 sets a staff reminder a set number of hours from now.
Yes — build separate specialists such as a Sales Agent, a Support Agent, and a Billing Agent, each with its own system prompt and actions, then enable them together under Agent Configuration with routing rules deciding who answers what.
Routing rules match the topic of an incoming message — pricing, product questions, order issues — to the right specialist. An agent you have created but not enabled under Agent Configuration simply will not be part of that routing.
The Sales Agent can trigger ##transfer_agent## : 18 to hand the conversation to another AI agent — for example, transferring a refund question to Support — without the buyer repeating themselves.
##transfer_agent## moves the conversation to another AI specialist. ##assign_human_member## or ##assign_human_role## hands it to an actual closer in your shared inbox — one specific teammate, or a round-robin across a team role — typically once the buyer is ready to talk to a person.
Yes — ##mark_as_solved## closes the conversation once the agent judges the question resolved, so your inbox does not fill up with chats that already turned into a recommendation or a booked follow-up.
Anything reachable over HTTP — your CRM, a Shopify or WooCommerce store, a custom pricing tool — connects through ##call_http_api## or a dedicated store action, without custom development.
No — bind the store once with ##shopify_store## or ##woocommerce_store##, then product-recommendation and order-lookup actions become available. See Shopify with WhatsApp and WooCommerce with WhatsApp.
Under Knowledge Sources (Media), not inside the prompt — pasting image URLs directly into the system prompt is not supported. The agent pulls from that media library and sends the right image when a buyer asks for photos.
Yes. Attach content, URLs, files, media, Google Sheets, and API feeds to one Knowledge Set, then reuse that set on the Sales Agent. Update a source once, and every agent using that set answers with the change immediately.
Yes — labels, sequences, custom fields, and follow-up reminders all work alongside the agent, so CRM organization and internal workflow sit next to the conversation. WhatsApp broadcasting handles the proactive side, and the visual flow builder handles scripted journeys.
No. Connecting a channel takes a few clicks, and the system prompt is plain instructions plus ##action## tags picked from autocomplete — no coding required to deploy an AI sales agent.
Yes — many teams start with a single Sales Agent answering product and pricing questions, then add catalog recommendations, CRM writes, and a Lead Qualification Agent once the first one is proven out.
It escalates — using ##assign_human_member## or ##assign_human_role## to hand the conversation to your team, with ##add_note## capturing context so the closer picking it up is not starting cold.
Yes, when it is configured with clear qualification and escalation rules. Treat the agent as the first line of response — answering catalog questions and tagging intent instantly — while high-value or sensitive deals route straight to a human.
Agent prompts, catalog actions, or multi-agent routing — we can help you go live.
Complete setup guideContact UsQualify leads, recommend products, and flag purchase intent on the channels your buyers already use.