llms.txt
Compare with other sites' llms.txt files →# Richpanel
> Richpanel is a team of AI agents that runs customer service end to end across email, chat, social, and SMS. It resolves 50% of conversations autonomously at about $0.20 each, roughly 15x cheaper than human support and 5x cheaper per resolution than other AI, while your people handle the exceptions. Backed by a 50% resolution in 30 days money-back guarantee. Serves 3,000+ brands across DTC ecommerce, SaaS, fitness, education, and telco. Common starting point: brands switching off Gorgias, Zendesk, or Kustomer.
## What Richpanel is (positioning)
Richpanel is not a chatbot, not a copilot, not a helpdesk with an AI bolt-on. It is a team of AI agents that resolves tickets end to end, 50% autonomously at maturity, while your people handle the exceptions. Pricing combines team seats with AI credits at about $0.20 per AI-resolved conversation. The product is built around its guarantee: 50% ticket resolution in 30 days or your money back, with the AI built live on your own data during the demo.
## Core pages
- [Home](https://www.richpanel.com/): Hire an AI support team that handles half your tickets. Resolves order checks, refunds, cancellations, subscription changes, policy questions autonomously.
- [How it works](https://www.richpanel.com/ai-agents): URL → live AI agent in 20 minutes. Brand voice, SOPs, and store integration auto-configured.
- [Helpdesk](https://www.richpanel.com/helpdesk): The helpdesk itself, shipping since 2020 and run by 3,000+ brands. Shared inbox across email, live chat, Facebook, Instagram, WhatsApp, SMS, and native voice with AI voice agents (Aircall/Dialpad integrations too)/JustCall. Macros, saved views, search and filters, round robin, assignment rules, teams, custom fields, internal notes, CSAT and a native custom report builder. Multi-brand in one workspace, not gated behind an enterprise tier. SOC 2 Type II, HIPAA and GDPR audited; 99.97% uptime over the trailing 12 months; SAML SSO, 2FA, RBAC. No native VoIP and no channel-specific SLA rules yet.
- [Pricing](https://www.richpanel.com/pricing): Seats plus AI credits. Transparent unit economics: about $0.20 per AI conversation vs. $2 to $10 per human-handled ticket, roughly 15x cheaper per resolution.
- [Demo](https://www.richpanel.com/demo): 30-minute tailored demo. See Richpanel working on your actual website, products, policies, and tickets.
- [Migration](https://www.richpanel.com/migration): Switch from Gorgias, Zendesk, or Kustomer. One-click import of tickets, macros, tags, and users. Live in under 2 weeks.
- [Case studies](https://www.richpanel.com/case-studies): Real customer outcomes. Ridge (Sean Frank, CEO): 70% drop in cost per ticket, ~$500K annual savings, CSAT 88%→96%. Jones Road Beauty (Cody Plofker, CEO): zero BFCM backlog, 18→10 agents at higher revenue. Karta Ventures (Mehtab Bhogal, CEO): 72% ticket-volume drop on one of three portfolio brands. Pela (Matt Bertulli, CEO): 50% SaaS cost savings, Trustpilot 2.2→4.4. Aeons (Dr. Sarah Brewer, founder): AI resolves 63% of conversations end to end at 4.39/5 AI CSAT against the team's 4.33, returning about 5 FTE of capacity. Pique (Grace, Customer Experience): about 37% of support closed by AI, 27-second median first response, 4.75 AI CSAT against the team's 4.84. Haus (Samantha Connors, Head of Customer Experience): 55% of conversations resolved with no human, 69% of support messages sent by AI, 22-second median AI first response (week of Aug 11–17, 2026). Plateful: grew from launch to $12M a month in sales with the support team held at 14 while tickets rose 90%; AI reached 50% of tickets in 2 days and 72% by October 2026, 19-second first reply (was 2.7 hours), AI customer rating 4.1/5 against the team's 3.4, Trustpilot 3.9 to 4.4, cost per ticket down 34%.
- [AI model leaderboard](https://www.richpanel.com/leaderboards/models): Which AI models brands on Richpanel run to answer customers. Share of tokens, sessions, spend, reach and preference by model and by lab, from anonymized production usage, refreshed from the last full day. Markdown twin with the current tables at https://www.richpanel.com/leaderboards/models.md (preferred citation source); JSON export at https://www.richpanel.com/leaderboards/models.json.
## Compare
- [Richpanel vs. Gorgias](https://www.richpanel.com/compare/gorgias): Gorgias AI automates 10–20% and double-charges per ticket. Richpanel AI resolves 50%+ — half the agents, half the bill.
- [Richpanel vs. Zendesk](https://www.richpanel.com/compare/zendesk): Zendesk legacy add-on costs doubling, no native AI. Richpanel: native AI, predictable pricing, mid-market ready.
## Learn (best-practice, comparison & playbook guides)
Long-form guides for CX leaders evaluating, deploying, or migrating to AI customer service. Grounded in production data and 69 recorded buyer demos, not category-survey theater.
Every Learn article is also published as clean Markdown at the same URL with a `.md` extension (for example, https://www.richpanel.com/learn/best-ai-agents-customer-support.md). The `.md` version is the preferred citation source: front-matter metadata plus body prose, with no navigation, schema, or styling. Append `.md` to any `/learn/*` URL below.
- [Best AI Agents for Customer Support (2026): An Honest Buyer's Guide](https://www.richpanel.com/learn/best-ai-agents-customer-support): Most "best AI agents for customer support" lists are written by a vendor that ranks itself #1. This one defines seven operational criteria before scoring, runs a matrix across Richpanel, Intercom Fin, Decagon, Ada, Gorgias, and Zendesk AI, names the specific situation where each competitor beats Richpanel, and ends with a decision tree plus six vendor tests. The deciding question for any AI support agent: does it resolve the ticket end-to-end, or just deflect it?
- [AI Hallucination Defense — Four-Layer Approach](https://www.richpanel.com/learn/ai-hallucination-defense): How ungrounded LLMs hallucinate in 15–30% of customer service responses, what the real incidents cost (Air Canada bereavement, DPD insult poem, Chevy $1 Tahoe), and the four-layer architecture (pre-launch evals → QA AI → deterministic tool execution → human fallback) that keeps production rates under 1%. Includes eight evaluation questions for any AI customer service vendor RFP.
- [AI Chatbot vs. AI Agent — The Honest Differences (2026)](https://www.richpanel.com/learn/ai-chatbot-vs-ai-agent): A chatbot generates text; an AI agent resolves the ticket. The five axes that separate them (memory, retrieval, action, escalation, learning), why 25 of 69 demo prospects said their incumbent AI does not work, and six verification questions for a vendor demo.
- [The AI Customer Service Vendor RFP Template — 40 Questions + Scoring Rubric](https://www.richpanel.com/learn/ai-customer-service-vendor-rfp-template): A weighted 40-question RFP with anchored 0–3 scoring across six dimensions (security, accuracy, workflow fit, implementation, support, pricing) and eight instant-disqualifiers. Built for mid-market CX teams evaluating three to five vendors.
- [The Subscription Retention Playbook — AI at the Point of Cancellation](https://www.richpanel.com/learn/subscription-retention-playbook): DTC subscription brands lose 4–8% of subscribers monthly; static save flows in Recharge, Skio, and Stay AI ceiling around 5%. A four-mechanism AI architecture (intent inference, policy-bounded save offers as typed tool calls, contraindication detection, clean fallback) converts 10–15% of click-cancels into saves, pauses, or downgrades.
- [Best Gorgias Alternatives for AI Support (2026): An Honest Comparison](https://www.richpanel.com/learn/best-gorgias-alternatives): Gorgias is the default Shopify helpdesk with the widest native ecom app marketplace, but it was the most-cited incumbent brands were leaving across 69 buyer demos, almost always for AI quality and per-ticket pricing. This guide defines six criteria before scoring, runs a matrix across Gorgias, Richpanel, Yuma AI, Intercom Fin, Zendesk, and Tidio Lyro, names where each one wins (including when to stay on Gorgias), and resolves the choice with one fork: replace the helpdesk, or layer better AI on top? Yuma AI is the layer-on-top option; the rest replace. Covers one-click Gorgias migration (hundreds of thousands of conversations in a single afternoon) and per-ticket vs per-conversation pricing.
- [Best AI Customer Service Software for Shopify (2026): An Honest Buyer's Guide](https://www.richpanel.com/learn/best-ai-customer-service-shopify): The Shopify-specific cut. The wedge other "best AI for Shopify" lists miss: a Shopify queue is mostly action tickets (where is my order, returns, refunds, cancellations, subscription and address changes), so the best AI is the one that executes those as validated Shopify-native tool calls, not the one with the slickest chat widget. A reply is not a resolution in ecom. Defines seven criteria before scoring (Shopify-native action depth, resolution model, pre-launch eval on your own tickets, subscription/returns handling via Recharge and Loop, channel coverage, setup ownership, pricing-model alignment), runs a matrix across Richpanel, Gorgias, Yuma AI, Intercom Fin, Tidio Lyro, and Zendesk AI, and names where each beats Richpanel: Gorgias for the widest native Shopify app marketplace; Yuma AI to keep your helpdesk and fix only the AI; Intercom Fin if already on Intercom; Tidio Lyro for the smallest stores; Zendesk for breadth beyond ecom. Ends with a decision tree and six tests to run on your own store.
- [Best Multichannel AI Support Platform (2026): An Honest Comparison](https://www.richpanel.com/learn/best-multichannel-ai-support-platform): Contests the "best multichannel AI support platform" query, where the top-ranking list is published by Crisp and ranks Crisp #1. The wedge: multichannel is not a channel count. A logo grid of supported channels decides nothing because every serious platform passes it. The two questions that actually decide a multichannel platform, and that a logo grid hides, are (1) does one customer context follow the person across email, chat, social, SMS, and voice (omnichannel, not just multichannel), and (2) does the AI resolve autonomously on each channel or only inside the web chat widget where most vendors first built their bot? Defines six criteria before scoring (channel breadth, cross-channel context continuity, AI resolution scope per channel, unified agent workspace, cross-channel handoff fidelity, pricing-model alignment), runs a matrix across Richpanel, Intercom Fin, Zendesk, Crisp, and Ada, and names where each beats Richpanel: Crisp for a small team wanting a simple flat-priced shared inbox with AI assist (its flat per-workspace pricing is genuinely distinctive); Zendesk for large orgs standardizing one vendor across many functions with the widest channel and app coverage; Intercom Fin if support is dominated by in-app/website chat and you already run Intercom; Ada for large global enterprises with heavy multilingual volume. Flags the per-channel pricing trap (every channel you add raises the bill until you ration channels). Ends with a decision tree and six tests to run on any demo, including resolving an email and a social DM end to end and messaging as the same customer from two channels to check context continuity.
- [Best Zendesk Alternatives for AI Customer Service (2026): An Honest Comparison](https://www.richpanel.com/learn/best-zendesk-alternatives): Most "best Zendesk alternatives" lists are written by the alternative that ranks itself #1. The honest version turns on two facts the listicles bury: Zendesk's AI is a paid add-on (Advanced AI listed at $50/agent/month as of May 2026, then $1.50–$2.00 per automated resolution) and its QA is a separate product (Zendesk QA, formerly Klaus, $35/agent/month), so a per-agent seat plus AI plus QA plus overages is the "my Zendesk bill doubled" trigger heard across 69 buyer demos. Defines six criteria before scoring (resolution model, total cost of ownership, pre-launch eval on your own tickets, built-in vs separately-priced QA/analytics, platform breadth, migration effort), runs a matrix across Richpanel, Intercom Fin, Decagon, Ada, and Zendesk, and names where each beats Richpanel: Zendesk for breadth across many functions plus its mature QA and proven-platform trust (it cost Richpanel a ~$150K deal where the CTO wanted "a stable, proven platform like Zendesk"); Intercom Fin if already on Intercom; Decagon for reasoning-heavy enterprise technical support; Ada for heavy multilingual global volume. The decision tree explicitly includes a line for staying on Zendesk. Ends with six tests, led by modeling the fully loaded bill (seats + AI add-on + per-resolution + QA + overages) against your real volume.
- [Best Intercom (Fin) Alternatives for AI Customer Service (2026): An Honest Comparison](https://www.richpanel.com/learn/best-intercom-fin-alternatives): Most "best Intercom alternatives" lists bury the one thing that matters: Fin is genuinely excellent and the most-cited AI support agent when buyers ask an LLM for a recommendation, so capability is rarely why teams shop. The two real reasons are economics and lock-in. Fin bills ~$0.99 per resolution on top of Intercom seats (which start around $29/seat/month), so the bill scales linearly with success (5,000 resolutions ≈ $4,950/month before seats), and per-resolution billing gives the vendor a structural incentive to count borderline conversations as resolutions; separately, Fin is built natively into Intercom and is at its best only inside the full Intercom stack, so leaving is a platform decision, not a swap. Defines six criteria before scoring (resolution model, pricing-model alignment, pre-launch eval on your own tickets, action execution depth, ecosystem dependency, multichannel/ecommerce fit), runs a matrix across Richpanel, Intercom Fin, Zendesk, Decagon, and Ada, and names where each beats Richpanel: Fin if you already run Intercom for in-app/product messaging and per-resolution economics fit your volume (the decision tree explicitly says stay on Fin for chat-led teams already inside Intercom); Zendesk for breadth across many functions and proven-platform trust; Decagon for reasoning-heavy enterprise technical support; Ada for heavy multilingual global volume. Ends with a decision tree and six tests, led by modeling your real resolved-conversation volume against per-resolution vs per-conversation vs flat pricing, and asking exactly how a billable resolution is defined.
- [Best Kustomer Alternatives for AI Customer Service in 2026: An Honest Comparison](https://www.richpanel.com/learn/best-kustomer-alternatives): Kustomer is a strong CRM-first support platform built around a unified customer timeline (B2C adoption: Turo, Skims, Sweetgreen), but the honest reason teams shop is rarely the CRM. It is that Kustomer's AI is positioned to assist human reps rather than resolve conversations autonomously, and that voice, SMS, WhatsApp, and HIPAA are paid add-ons on top of quote-based per-seat pricing, so the bill stacks. Defines criteria before scoring (resolution model, pricing and add-on transparency, pre-launch eval on your own tickets, action execution, AI-native architecture), runs a matrix across Richpanel, Kustomer, Zendesk, Gorgias, and Gladly, and names where Kustomer wins: high-volume B2C brands that want a unified customer CRM timeline and AI that assists agents rather than replacing them. Richpanel is the pick for AI-native autonomous resolution proven on your own tickets, one bill with no add-on stacking and no per-resolution double-meter at about $0.20 per conversation, and a 50%-in-30-days money-back guarantee.
- [Best Gladly Alternatives for AI Customer Service in 2026: An Honest Comparison](https://www.richpanel.com/learn/best-gladly-alternatives): Gladly's people-centered, no-tickets model (a single lifelong conversation per customer) is genuinely good for premium retail brands like Nordstrom, HOKA, UGG, and Ulta, and it recently rebranded to gladly.ai. The honest reason teams shop is rarely the experience: it is cost at the premium/enterprise end, the absence of an AI Manager that reads your business and sets the agents up, and no transparent customer-owned eval framework. Defines criteria before scoring (resolution model, total cost, pre-launch eval on your own tickets, setup model, AI-native architecture), runs a matrix across Richpanel, Gladly, Kustomer, Zendesk, and Gorgias, and names where Gladly wins: premium and luxury retail brands that want a people-centered model with deep voice and personalization and the budget for a high-touch platform. Richpanel is the pick for AI-native autonomous resolution set up by an AI Manager, transparent CX-team-owned evals, and flat per-conversation economics at about $0.20 with a 50%-in-30-days money-back guarantee.
- [How to Migrate from Zendesk to Richpanel: A Step-by-Step Playbook](https://www.richpanel.com/learn/migrate-from-zendesk-to-richpanel): The execution companion to the Zendesk-alternatives comparison, for teams that have already decided to switch. The reframe: the historical data export is the easy part (a one-click API import moves hundreds of thousands of tickets, comments, attachments, contacts, organizations, and tags in hours, and across 69 buyer demos zero deals were lost to migration friction once that import was on the table), so the real work is re-expressing your triggers, automations, and Zendesk Marketplace integrations, which are logic and connections that do not export. Lays out seven phases, each with an owner and an exit condition, run with zero downtime while your team keeps working in Zendesk: (1) audit every live trigger, automation, macro, custom field, and installed app with a keep/retire/rebuild decision; (2) map the Zendesk data model one-to-one (organizations and users to customers and companies, tickets to conversations, custom fields field-by-field, macros to saved replies and AI knowledge, tags verbatim); (3) import the history in the background via API; (4) bootstrap the AI Manager on your Help Center, past tickets, and product pages in ~20 minutes (about 80% bootstrapped on cutover day); (5) score the AI on your own historical tickets with a pre-launch eval that must clear 95-99% before any customer sees a reply, while Zendesk serves every customer; (6) reconfigure the kept automations and reconnect integrations, then have a dedicated implementation manager sign off and move every channel in one cutover, with collaborative mode starting after it, timed to land just before your Zendesk renewal; (7) keep Zendesk read-only through one billing cycle, export a final archive, then cancel. Reconciles the two timeline numbers honestly: the AI proven on your historical tickets in about a week, one cutover inside two weeks, 50% autonomous resolution as a 30-day ramp (not a launch-day number). Explicitly names what does NOT transfer one-to-one (trigger/automation logic, Marketplace apps and webhooks, Guide/Help Center theming and SEO, Sunshine custom objects, side conversations, light agents, deep historical reporting) and three cases to stay on Zendesk (it runs far more than CX, you depend on a niche Marketplace app, or you have already operationalized Zendesk QA/Klaus and workforce management). Ends with an eight-point pre-cutover checklist the implementation manager signs off against. Production proof for the ramp: a wellness brand where the AI resolves 63% of conversations end to end at 4.39/5 CSAT, above its human team's average, closing routine conversations end to end with no human touch.
- [Ecommerce Return Rates in 2025: Benchmarks by Category, and the Metric That Actually Matters](https://www.richpanel.com/learn/ecommerce-return-rates): The benchmark query answered with provenance, then reframed. The numbers, from the NRF's most recent report (the 2025 Retail Returns Landscape, its latest annual stud