<\/span><\/h2>\n\n\n\nThe key design decision behind Liam is that every channel feeds the same core. Instead of five disconnected bots, you deploy one employee with three shared layers.<\/p>\n\n\n\n
1. Customer-facing touchpoints.<\/strong> Website Chat, WhatsApp, support email, inbound voice calls, and SMS are all entry points to the same employee.<\/p>\n\n\n\n2. A shared engine.<\/strong> Behind every channel sits one core that maintains customer memory, applies your policies, and coordinates tool use. It has four parts:<\/p>\n\n\n\n\nCompany Brain:<\/strong> your approved knowledge, including manuals, FAQs, and help-center content, retrieved through a hybrid search approach.<\/li>\n\n\n\nCustomer Context:<\/strong> cross-channel identity matching, so a phone number, email address, and order ID can all be tied to the same person.<\/li>\n\n\n\nSkills and Workflows:<\/strong> ticket triage and orchestration across your business systems.<\/li>\n\n\n\nGuardrails and Approvals:<\/strong> human-in-the-loop thresholds delivered through Slack or Microsoft Teams.<\/li>\n<\/ul>\n\n\n\n3. Internal business tools.<\/strong> This is where Liam actually works: Zendesk, Freshdesk, and Intercom for tickets; Shopify and Stripe for orders and refunds; Slack and Teams for manager approvals; and scheduled reports for leadership.<\/p>\n\n\n\nThe separation matters. Customers only ever see the front-line channels. Behind the scenes, Liam is operating inside your internal systems, and human managers stay in the loop wherever money or risk is involved.<\/p>\n\n\n\n
<\/span>What Liam does on each channel<\/span><\/h2>\n\n\n\nEvery channel shares the same brain, but each has its own tone, formatting, and setup needs.<\/p>\n\n\n\n
<\/span>Website Chat<\/span><\/h3>\n\n\n\nLiam answers pre-sales and technical questions, searches your knowledge base, captures leads, and performs instant account lookups. Setup covers widget styling, your custom domain, allowed origins, and a welcome message. This is a recommended launch channel.<\/p>\n\n\n\n
<\/span>WhatsApp<\/span><\/h3>\n\n\n\nUsing the WhatsApp Business API, Liam answers questions, provides order updates, processes returns, and can send proactive notifications. You need a verified Meta number and approved message templates. This is the second recommended launch channel.<\/p>\n\n\n\n
<\/span>Email<\/span><\/h3>\n\n\n\nLiam monitors shared support mailboxes, reads multi-turn threads, drafts responses, and resolves tickets. Connection is through Gmail or Outlook OAuth, or through Zendesk email forwarding. You can add this at any time.<\/p>\n\n\n\n
<\/span>Voice<\/span><\/h3>\n\n\n\nFor inbound phone calls, Liam handles spoken queries with low latency, books appointments, and transfers to a human when needed, with no robotic phone menus. Setup involves a dedicated number, a language persona, and a telephony SIP bridge.<\/p>\n\n\n\n
<\/span>SMS<\/span><\/h3>\n\n\n\nLiam sends transactional reminders, dispatch confirmations, and shipping updates, and handles short two-way replies. You connect an SMS provider such as Twilio, a sender ID, and complete 10DLC compliance opt-in.<\/p>\n\n\n\n
Because the brain is shared, turning on a new channel later does not mean retraining the employee or re-uploading documents. You switch the channel on and the same knowledge, policies, and customer history apply.<\/p>\n\n\n\n
<\/span>A real scenario: one customer, multiple channels, zero repeated questions<\/span><\/h2>\n\n\n\nThe clearest way to see the difference is to follow a single customer through a day. Here is the sequence RhinoAgents uses to illustrate Liam, from the live product page.<\/p>\n\n\n\n
Before: the fragmented experience<\/strong><\/p>\n\n\n\nAlex asks about order #12345 on the website chat. A bot replies with a tracking link. Later, Alex messages on WhatsApp to cancel part of the order. A different bot, with no record of the website chat, asks for the order number again, then for the email address, then for the reason. Alex asks for a human. The human agent opens the helpdesk, finds nothing from the website chat, and asks the same questions a third time. A refund request goes to a manager through email, where it sits until the next morning.<\/p>\n\n\n\n
After: with Liam<\/strong><\/p>\n\n\n\n\n10:00 AM, Website Chat.<\/strong> Alex asks where order #12345 is. Liam matches the customer’s email to Shopify and carrier records, verifies the shipping address, and confirms the package is out for delivery today.<\/li>\n\n\n\n11:30 AM, WhatsApp.<\/strong> Alex writes: “I asked about my order earlier on your site. Can I cancel it?” Liam matches the incoming phone number to the existing session, recalls the 10:00 AM conversation, and explains the cancellation policy without asking Alex to re-explain anything.<\/li>\n\n\n\n11:32 AM, Zendesk.<\/strong> Alex confirms they want a partial refund for the split-shipped item. Liam creates Zendesk ticket #10429 and attaches the full transcript from both the website and WhatsApp.<\/li>\n\n\n\n11:33 AM, Slack.<\/strong> Company policy says refunds over $25 need supervisor approval. Liam posts an interactive card in the #support-approvals<\/code> channel showing the customer, the $45.00 amount, and the policy check, with Approve and Reject buttons.<\/li>\n\n\n\n12:00 PM, resolution.<\/strong> The support lead clicks Approve. Liam triggers the Stripe refund, confirms the credit to Alex on WhatsApp with a receipt, marks the Zendesk ticket solved, and asks for a CSAT rating.<\/li>\n<\/ul>\n\n\n\nThe ticket was solved in about 30 minutes, and Alex never repeated a single detail. The human on the team spent one click, not one investigation.<\/p>\n\n\n\n
<\/span>Human approval for refunds and cancellations<\/span><\/h2>\n\n\n\nThe fastest way to lose trust in AI-driven support is for it to hand out money it should not have. Liam is built around the opposite principle: it can act, but financial actions are gated by your rules.<\/p>\n\n\n\n
Here is how the refund flow works:<\/p>\n\n\n\n
\nA customer asks for a refund or cancellation.<\/li>\n\n\n\n Liam checks your company policy and verifies the order.<\/li>\n\n\n\n Liam creates a Zendesk ticket with the full context.<\/li>\n\n\n\n If the amount is above your threshold, Liam sends a one-click approval card to a manager in Slack or Microsoft Teams.<\/li>\n\n\n\n Only after the manager approves does Liam execute the refund in Stripe or the cancellation in Shopify, then notify the customer.<\/li>\n<\/ol>\n\n\n\nYou set the dollar thresholds. Small, routine refunds can be handled without interruption, while larger ones always get a human decision. Every step lands in an audit log, so finance and support leads can see what was requested, who approved it, and when.<\/p>\n\n\n\n
The approach removes the two failure modes at once: refunds are no longer stuck waiting for someone to find the right ticket, and an automation is never able to issue unauthorized credits.<\/p>\n\n\n\n
<\/span>Answers grounded in your own knowledge base<\/span><\/h2>\n\n\n\nA support employee is only as good as the information it is allowed to use. Liam connects to your Notion workspace, help center, PDF runbooks, Google Docs, and URLs, and uses a hybrid retrieval approach that combines dense vector search with BM25 keyword matching. In practice, that means it can find the right passage whether a customer uses your exact terminology or describes the issue in their own words.<\/p>\n\n\n\n
Two things make this practical for a live support operation:<\/p>\n\n\n\n
\nAuto-sync.<\/strong> When your product documentation or return policy changes, Liam picks up the updates, so customers get current guidance and not last quarter’s rules.<\/li>\n\n\n\nVerified answers.<\/strong> Responses are drawn from approved content, which is how RhinoAgents addresses the hallucination risk that makes many teams nervous about automating support.<\/li>\n<\/ul>\n\n\n\nWhen the knowledge base does not cover a question, the right behavior is escalation, not guesswork. Liam hands off to your human team with the conversation log and context attached, so the customer does not start over with a person either.<\/p>\n\n\n\n
<\/span>Beyond answering: sentiment, identity, and daily reporting<\/span><\/h2>\n\n\n\nThree capabilities round out the role and are worth calling out because they are the ones human teams usually struggle to do consistently.<\/p>\n\n\n\n
Sentiment and frustration handling.<\/strong> Liam monitors customer tone in real time and adjusts how it responds. When a conversation signals churn risk, it can route the account to a designated customer success manager through a warm escalation, with the full history attached.<\/p>\n\n\n\nCross-channel identity matching.<\/strong> Liam links profiles using verified identifiers such as email, phone number, and order number. That is what lets a WhatsApp message pick up a website conversation without extra verification steps.<\/p>\n\n\n\nDaily 9:00 AM briefing.<\/strong> Every morning Liam delivers a summary to your Slack channel and email: total volume, autonomous resolution rate, CSAT, SLA risks, trending customer questions and bugs, and any pending refund approvals. Support leads start the day with the numbers already compiled.<\/p>\n\n\n\n<\/span>One dashboard for every channel<\/span><\/h2>\n\n\n\nWith five separate bots, reporting means opening five tools and reconciling them in a spreadsheet. With Liam, there is one support dashboard with channel filters and live telemetry.<\/p>\n\n\n\n
The sample view RhinoAgents shows for a 30-day period illustrates what leaders see: total conversations, the share resolved autonomously, Zendesk tickets created, human approvals requested, a breakdown by channel, and a live stream of recent resolutions. In that example, website chat and WhatsApp account for the large majority of volume, with email, voice, and SMS making up the rest. That split is exactly why RhinoAgents recommends starting with those two channels.<\/p>\n\n\n\n
The point of a single pane of glass is not aesthetics. It is that you can finally answer basic operational questions: which channel is growing, where autonomous resolution is strongest, and how many decisions are landing on managers.<\/p>\n\n\n\n
<\/span>Traditional chatbots vs. an AI support employee<\/span><\/h2>\n\n\n\n<\/th> Traditional chatbot<\/th> Liam, the AI Customer Support Employee<\/th><\/tr> Logic<\/td> Static decision trees and keyword matching<\/td> Reasoning grounded in your Company Brain and approved policies<\/td><\/tr> Channels<\/td> Separate bots with no shared memory<\/td> One employee across web, WhatsApp, email, voice, and SMS<\/td><\/tr> Resolution<\/td> Pastes FAQ links or deflects to a human queue<\/td> Creates tickets, checks orders, and triggers authorized refunds<\/td><\/tr> Financial risk<\/td> Blind to it<\/td> Policy thresholds with one-click Slack or Teams approval<\/td><\/tr> Reporting<\/td> Disjointed logs per channel<\/td> One dashboard with channel attribution and daily briefings<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nThe difference is not that Liam is a smarter chatbot. It is that Liam is scoped as a role, with the access and accountability of one.<\/p>\n\n\n\n
<\/span>Results RhinoAgents reports<\/span><\/h2>\n\n\n\nOn its product page, RhinoAgents cites these headline figures for Liam: first response in under 30 seconds, CSAT of 94% or higher, and 78% of tickets resolved autonomously, with 24\/7 coverage across channels. Treat them as benchmarks for what a well-configured deployment aims for. Your own numbers will depend on the quality of your knowledge base, the channels you activate, and how your refund and escalation rules are set.<\/p>\n\n\n\n
The practical gains tend to show up in the same places:<\/p>\n\n\n\n
\nSpeed.<\/strong> Customers get an answer within seconds on every channel, at any hour, including outside business hours.<\/li>\n\n\n\nFewer repeated questions.<\/strong> Shared history means conversations continue where they left off.<\/li>\n\n\n\nLess manual work for the team.<\/strong> Order lookups, ticket creation, and routine refunds stop consuming human time.<\/li>\n\n\n\nBetter use of human judgment.<\/strong> Managers decide on the exceptions that matter, in one click, instead of reconstructing context.<\/li>\n<\/ul>\n\n\n\n<\/span>A simple launch plan: start with Website Chat and WhatsApp<\/span><\/h2>\n\n\n\nYou do not have to switch on all five channels at once. RhinoAgents recommends launching with Website Chat and WhatsApp on day one, since together they capture the large majority of routine inbound volume while proving the value of shared memory. A sensible rollout looks like this:<\/p>\n\n\n\n
\nHire Liam and pick your channels.<\/strong> Select Website Chat and WhatsApp. Connections use secure OAuth, so no code is needed. Selecting a channel does not bind credentials until you authorize it.<\/li>\n\n\n\nComplete the setup checklist.<\/strong> Configure widget branding and allowed web origins, verify your WhatsApp Business number, and connect your knowledge base and refund threshold rules.<\/li>\n\n\n\nConnect your tools.<\/strong> Link your helpdesk (Zendesk, Freshdesk, or Intercom), your commerce stack (Shopify and Stripe), and Slack or Teams for approvals.<\/li>\n\n\n\nReview the first week.<\/strong> Use the daily briefing and dashboard to see what Liam resolved, what it escalated, and what your knowledge base is missing.<\/li>\n\n\n\nAdd channels later.<\/strong> Turn on email, voice, and SMS when you are ready. No retraining or re-uploading is required.<\/li>\n<\/ol>\n\n\n\n<\/span>Pricing<\/span><\/h2>\n\n\n\nLiam is priced as a straightforward monthly plan, with no per-seat human license penalty and unlimited team members who can collaborate with the employee.<\/p>\n\n\n\n
\nStarter, $20\/month:<\/strong> up to 3 AI employee roles, Website Chat and knowledge base ingestion, 10,000 monthly execution credits, and Zendesk and Freshdesk sync.<\/li>\n\n\n\nPro Omnichannel, $49\/month (most popular):<\/strong> up to 5 roles, Website Chat plus WhatsApp Business, Slack one-click refund approvals, 50,000 monthly execution credits, and the unified dashboard with 9 AM briefs.<\/li>\n\n\n\nBusiness, $99\/month:<\/strong> up to 10 roles, all five channels including Voice, 150,000 monthly execution credits, 400+ enterprise connectors including Salesforce and SAP, and a dedicated SLA and support engineer.<\/li>\n<\/ul>\n\n\n\nA free 14-day trial is available with no credit card required, so you can test Liam against real conversations before committing.<\/p>\n\n\n\n
<\/span>Security and data privacy<\/span><\/h2>\n\n\n\nSupport conversations contain personal and financial information, so security cannot be an afterthought. RhinoAgents lists SOC 2 Type II certification for Liam, zero training of public models on your customer data, automatic PII redaction, and encrypted audit logs. The product also supports GDPR and HIPAA-oriented requirements. Combined with approval thresholds for money-related actions, this gives security and finance teams something concrete to review.<\/p>\n\n\n\n
<\/span>Who is it for?<\/span><\/h2>\n\n\n\nLiam fits teams that field a steady flow of repeat questions and transactional requests across more than one channel:<\/p>\n\n\n\n
\nDirect-to-consumer and ecommerce brands<\/strong> handling order status, returns, and refunds on Shopify and WhatsApp.<\/li>\n\n\n\nB2B SaaS companies<\/strong> where customers move between in-app chat and email, and where fast, accurate knowledge base answers matter.<\/li>\n\n\n\nGrowing support teams<\/strong> that want to cover nights and weekends and absorb volume spikes without hiring ahead of demand.<\/li>\n\n\n\nSupport leaders<\/strong> who want one source of truth for performance across channels.<\/li>\n<\/ul>\n\n\n\nIf your team is small and your channels are few, Starter gives you a low-cost way to begin. If customers already reach you on WhatsApp, Pro Omnichannel is the natural fit.<\/p>\n\n\n\n
<\/span>Frequently asked questions<\/span><\/h2>\n\n\n\n<\/span>Is this one AI employee or multiple separate chatbots?<\/span><\/h3>\n\n\n\nIt is one unified employee. Liam uses a single Company Brain, one customer history, and one action engine across Website Chat, WhatsApp, Email, Voice, and SMS. A customer who starts on your website and follows up on WhatsApp is recognized, and the earlier conversation carries over.<\/p>\n\n\n\n
<\/span>How does Liam handle refund requests safely?<\/span><\/h3>\n\n\n\nLiam follows human-in-the-loop financial rules. It verifies your policy, creates a Zendesk ticket, and sends a one-click approval card to your support manager in Slack or Microsoft Teams. Only after approval does it process the refund through Stripe or the cancellation through Shopify and notify the customer.<\/p>\n\n\n\n
<\/span>How does cross-channel recognition work between the website and WhatsApp?<\/span><\/h3>\n\n\n\nLiam links profiles using verified identifiers such as email address, phone number, or order number. If a customer asks about an order on the website and later messages on WhatsApp, Liam matches the order record, recalls the context, and continues the conversation.<\/p>\n\n\n\n
<\/span>Can we launch with just Website Chat and WhatsApp first?<\/span><\/h3>\n\n\n\nYes, and RhinoAgents recommends it. Those two channels cover most routine volume. You can add Email, Voice, and SMS later with a single toggle, without retraining the employee or re-uploading documents.<\/p>\n\n\n\n
<\/span>What reports does Liam provide?<\/span><\/h3>\n\n\n\nEvery morning at 9:00 AM, Liam sends an automated briefing to your Slack channel and email. It covers ticket volume, autonomous resolution rate, CSAT, trending customer questions, and pending refund approvals.<\/p>\n\n\n\n
<\/span>How does Liam search our knowledge base?<\/span><\/h3>\n\n\n\nLiam connects to your Notion workspace, help center, PDF runbooks, or Google Docs and uses hybrid dense vector plus BM25 retrieval. When your documentation or return policy changes, Liam syncs the updates so customers get accurate, current answers.<\/p>\n\n\n\n
<\/span>Conclusion: stop deploying channels, start deploying an employee<\/span><\/h2>\n\n\n\n