Hire an AI team that finishes the job.

It reads the order, applies your policy, closes the ticket.

Book a demo See how it is kept safe

4.7 on G2 Shopify

4.8 on Shopify

Previous era

Humans manage the AI.

New era

An AI that manages the AI.

3,000+ brands run their support on Richpanel

Your AI answers. Your team does the rest.

It will start picking up answers that we didn’t give it.

I did my best to coach it. I don’t think it’s me. I think it’s them.

It handles 30 to 40% of tickets. Otherwise they do fine.

An AI that builds the AI. You only approve.

It reads your tickets and policies, then writes the instructions and wires the tools.

Book a demo

Richpanel's AI Team and Models screen: CX Manager above Customer AI, Agent AI and QA Agent, each with the model you choose

You write the policy. It handles the mess.

A four line subscription playbook beside three customer messages, each resolved by a different line of it, and the allergy question routed to a person

A closed ticket is not the outcome.

A deflection closes the ticket and loses the customer.

Live conversation AI resolving

Customer I want to cancel my subscription. I still have two bags left and they keep piling up.

Read the delivery schedule

Checked the plan: 2 bags a month

Pulled what is left unopened

Richpanel AI Your next delivery is due Thursday, and you are on the 2 bag plan. I can push it out four weeks so you finish what you have, or move you to the single bag every month. Which suits you better?

Subscription kept Delivery schedule read, plan changed, confirmation sent. No agent touched it.

A cancellation becomes a skip.

Cancellation link sent→Subscription kept

A return becomes an exchange.

Money leaves→Revenue retained

A slow reply becomes a chargeback.

Nine hours to reply→Answered in seconds

Richpanel tool permissions: order, shipping and product lookups set to always allow, while cancel, refund, address edit and discount actions require approval

Start collaborative. Hand over more when you trust it.

Lookups run free. Money and order edits wait for your approval.

See the safety architecture

When the instructions run out, it escalates. It never guesses.

Every model errs. Five things stand between the error and your customer.

Layer 1 10,000+ evaluations before a model ships.Quirks get corrected in the prompt layer. 10 to 15% more accurate off the shelf.

Layer 2 A second model reads risky drafts.A dollar amount or an order change routes to a reviewer.

Layer 3 Missing instructions escalate. Always.Where your policy is silent it hands over, every time.

Layer 4 Tool permissions cap the blast radius.No answer reaches an action you have not switched on.

Layer 5 QA AI reviews 100% of conversations.Misses ranked by frequency, so the fix targets the cause.

We will not claim it never errs. We claim the error does not reach your customer.

Stop QAing by hand. It finds its own gaps.

QA AI reads every closed conversation and traces the miss to its cause.

248escalations Traced to one missing toolGrouped automatically, not sampled by hand.

+9points Resolution rate, after it was connectedProjected before publish, measured after.

Richpanel Learning Insights: a missing subscription tool flagged after 248 escalations, with a one click fix and the projected gain in auto resolution

When a person is needed, nine steps become two.

It hands over with the history read, the order pulled, the reply drafted.

AI is only useful when the operating numbers move.

Aeons Ancient Roots olive oil on a Tuscan table

63% of customer messages now sent by AI 4.39 out of 5 against the team's 4.33, and five agents of capacity back. Read the case study

A Ridge brass wallet and key case on an olive Ridge backpack

70% lower cost per ticket About $500K back every year, and CSAT climbed from 88% to 96%. Read the case study

Iced Pique teas on a poolside ledge in sunlight

AI closes about 37% of all volume CSAT parity with the team, and a median first response of 27 seconds, not 6.8 hours. Read the case study

SOC 2 Type IIHIPAA auditedGDPR99.97% uptime3,000+ brands

One meter, not two. About $0.20 a resolution.

Per ticket helpdesks bill the ticket, then bill the AI that resolved it. We charge once.

$0.20 per AI resolution Annual rate. $200 a month minimum.

15x below human support A $1,500 to $2,000 agent runs near $3 a conversation.

5x below other AI Gorgias $0.90. Intercom Fin $0.99. Siena and Sierra $1 to $2.

Marginal cost per resolution. Blended cost falls by less, and we will show you both on your own numbers.

Book a demo Build your exact price

“We have a helpdesk, and switching is a pain.”

We move everything while your team works as usual, and buy out up to 100% of your current contract.

The Richpanel migration wizard importing an existing helpdesk, including tags, users, macros and conversations

Week 1

We move your data

Conversations, macros, tags, routing and SLAs. Hundreds of thousands import in an afternoon.

Week 2

The AI trains on your tickets

The agents run against your hardest recent conversations before a customer sees one.

Cutover

You approve, then we switch

Your current helpdesk serves until you sign off. Nothing is cancelled before that.

“We’ve had zero downside. The implementation was pretty immediate, which actually shocked me. It just went off without a hitch.”

Cody PlofkerCEO, Jones Road Beauty

Answered honestly.

Most helpdesk AI can answer and, once configured, act. The difference is who does the configuration, the QA and the improvement work. In Richpanel a CX Manager AI reads your ticket history, drafts the operating instructions, wires the tools and writes the QA rubric, then keeps tuning after launch. Your team approves rather than builds.

Every model is evaluated more than 10,000 times before it is offered, and the quirks found get corrected in the prompt layer, which measures 10 to 15% more accurate than the same model off the shelf. On top of that, a second model reviews any draft that states a dollar amount or touches an order, tool permissions cap what the AI can reach, and where your instructions are silent it escalates instead of guessing. We do not claim zero mistakes. We claim the mistake does not reach your customer.

It acts. Order lookups, tracking, address edits, exchanges, store credit, subscription skips, pauses and swaps, cancellations and returns run as real tool calls against Shopify, Recharge, Loop, Skio and the rest of your stack. Each action carries its own permission, so you decide which ones run automatically and which need a person to approve.

A 30 minute proof of concept is built live on your own data during the demo, then a 2 week pilot, then full deployment in about 4 weeks. Half your eligible conversations resolved by day 30, or every dollar back.

About $0.20 per AI resolved conversation on the annual committed rate, with a $200 per month minimum, plus $99 per seat for the humans. You pick the model and the token budget. There is one meter: a conversation the AI resolves is not billed again as a ticket.

Sometimes not on payroll alone, and we will say so. At $5 an hour the per ticket costs can land close. The value at that point is instant first response at 3am, peaks absorbed without overtime, the refunds turned into exchanges and the subscriptions saved, and the chargebacks that slow first responses cause. Bring your refund rate and chargeback volume to the demo and we will build the case on those numbers rather than on headcount.

Not natively. Richpanel integrates with Aircall, Dialpad and JustCall so calls and their notes sit beside the rest of the conversation, but we do not host the voice line itself. If a single native voice pane is a hard requirement, say so early.

Ask this before the automation promise, and we will ask it too. When an agent answers a tracking or cancellation question today, whichever system they open is the system the AI needs. If that is a custom OMS, an ERP or a supplier sheet rather than Shopify, the integration work gets scoped and priced before anyone commits to a resolution rate.

Guaranteed outcomes. Or every dollar back.

50% AI automation in 30 days

30% lower total bill In writing, from day one.

$0 risk to you Full refund if we miss.

Book a demo Miss either number, every dollar back.

Bring your hardest tickets to the call.

Twenty minutes. Your store, your policies, your conversations. No slideware.