Unified CCaaS platform Voice, chat, QA, and workforce management — unified. Local Phone #
Global DID New Local numbers, 150+ countriesChoose Your Model
Enhance Your Team
Answering Service
ecommerce
AI agents handle order tracking, returns, and product questions in 65+ languages, while trained specialists manage escalations and complex issues.
The short version: we bill per resolved issue from $1, and you pay nothing on the ones that are not resolved. If you would rather buy capacity than outcomes, AI voice is $0.15 a minute, web chat $0.05 a session, and a live shared team starts at $300 a month. We work inside Amazon through the SP-API, Shopify and eBay, and WhatsApp, LINE and TikTok arrive in the same queue as email and chat.
how it plays out
Order placed 5 days ago, stuck in transit
Order not yet dispatched from warehouse
Color received differs from product photos
Pre-purchase question from social channel
the team behind it
real members of the global team Ecommerce-Trained Specialists
65+ Languages
Follow-the-Sun Coverage
the part nobody prices in
Every support vendor sells you a price per minute or per ticket, which quietly assumes the interesting number is how much a contact costs. In ecommerce it is not. The interesting number is what share of contacts end without the customer needing to come back — because the ones that do come back arrive a second time, angrier, and often arrive as a refund, a chargeback or a marketplace claim instead of a message.
That is why deflection metrics read so well and change so little. A bot that closes 60% of conversations has not necessarily resolved 60% of problems; it has ended 60% of conversations. The customer who was told to check the tracking page is still a customer who does not know where the parcel is.
So we are willing to bill on the outcome. Per resolved issue, from $1, and nothing on the ones that are not resolved. It is the only pricing model in this category where our incentive and yours point the same direction — and the reason we can offer it is that the same company runs the automation and the people it escalates to, so there is nobody to blame in the handoff.
The second half of the argument is what happens when volume moves. Support contacts scale with orders, and headcount does not scale smoothly. You hire in whole people, weeks in advance, against a forecast. Peak season, a viral product or a carrier failure all break that forecast in the same direction. Per-minute and per-resolution both absorb it without a hiring decision; a payroll does not.
who takes which call
Most vendors sell you either an AI receptionist or a live answering service and leave the allocation to you. Since we run both, here is the split we would actually recommend — including the calls we think should never touch automation.
| The call | Route to | Cost of getting it wrong |
|---|---|---|
| "Where is my order?" — first contact | AI | Low. The tracking record answers it, and the answer is the same at 3am. |
| "Where is my order?" — second contact | Live agent | High. This is the last conversation before a chargeback. |
| Return eligibility and label generation | AI | Low. Your policy is a rule, and rules are what automation is for. |
| Refund not received, or charged twice | Live agent | High. Money moving the wrong way is never a routine ticket. |
| Pre-purchase product questions | AI, escalating on intent | Medium. Fast beats perfect, until the question implies a large order. |
| Marketplace claims with a policy clock running | Live agent | High. A missed A-to-z window is not recoverable by apologising well. |
| Damaged, unsafe or wrong item received | Live agent | High. Reputational, and sometimes legal. |
what it costs
Most vendors in this category quote by form. These are our published packages — not a quote, since language mix, hours and volume all move the number. The per-minute figures are simply the monthly price divided by the included minutes.
| Option | Published price | Per minute | What it's for |
|---|---|---|---|
| Live shared team | from $650 / month | includes 1,500 min → ≈ $0.43/min | Higher volume, where the same shared team costs about $0.43 a minute instead of $0.60 — the arithmetic that matters as orders grow |
| AI voice | $0.15 | per minute | Order status, delivery windows and returns policy questions by voice — the volume that spikes with your shipping problems, not your sales |
| Web chat | $0.05 | per session | Pre-purchase questions on the product page, where an unanswered one is a cart that never becomes an order |
| Live shared team | from $300 / month | includes 500 min → ≈ $0.60/min | A live team across email, chat and social, for stores where contacts are steady enough to plan around |
| Dedicated agent | from $5.80 | per hour, per agent | A dedicated agent who learns your catalogue, your policy edges and your marketplace accounts, and stays with them |
Anything involving a person is a starting price. Two things move it: which languages you need, and how senior an agent the work actually requires — a bilingual agent handling refund authority is not the same hire as one taking bookings. AI voice and web chat are flat, because there is no hiring in them. And every model above is open to every business on this page: which one fits is a question about the shape of your volume, not about your industry.
Minimum engagement is half a seat — not a five-seat block. We also bill per resolved issue, from $1, where you pay nothing if it is not resolved; that model fits businesses where a customer problem actually gets solved on the call, not businesses where the call is intake.
do the arithmetic yourself
Pull last month's contacts and count how many were the second or later contact about the same order. Call that share R. That is your repeat-contact rate, and it is the number this whole decision turns on — not the price per minute, not deflection.
Here is why. A contact costs you roughly the same whoever handles it. A contact that comes back costs you that twice, plus whatever the customer does in between — a marketplace claim, a chargeback, a one-star review, or simply not ordering again. None of those appear on a support invoice, which is exactly why they get left out of the comparison.
So run the comparison properly: total contacts × cost per contact × (1 + R), against the same sum with a lower R. A vendor who is 20% cheaper per contact and 10 points worse on R is not cheaper. We are willing to bill on resolution precisely because that is the number we are asking you to judge us on.
setup
Five days is our normal window to staff a live account — that is five days to find and place an agent who fits your business, not five days of software setup. AI voice can go live sooner because there is no hiring step.
your existing software
Amazon through the SP-API, two-way with Shopify, plus eBay — so an agent handling a case can see the order, the shipment and the return without asking the customer to repeat the number. WhatsApp, LINE and TikTok come into the same queue as email and chat. If a platform is missing from that list, ask before you assume: we will tell you straight whether we connect today or whether it needs building.
your existing software
Amazon 
Shopify 
eBay 
WhatsApp 
LINE 
TikTok Running something else? Tell us what it is and we will say straight whether we connect to it today or whether it needs building.
what comes off the call
Most of what makes ecommerce support slow is not language or skill — it is an agent who has to ask the customer for information the store already has. We work inside your systems, so the person handling the case starts with:
Amazon runs through the SP-API, Shopify is two-way, and eBay is connected. WhatsApp, LINE and TikTok land in the same queue as email and chat. If your platform is not on that list, ask rather than assume — we will tell you straight whether we connect to it today or whether it would need building.
where AI should not be
The second contact about the same order
The first "where is my order" is a lookup. The second one is a customer deciding whether to trust you, and it is the last cheap moment before it becomes a chargeback or a marketplace claim. It goes to a person, every time.
Anything with money moving the wrong way
Duplicate charges, a refund that has not landed, a subscription the customer says they cancelled. Automation is good at explaining refund timelines and bad at recognising that this particular one is genuinely the store's fault.
Safety, injury or a product that arrived damaged in a dangerous way
A bot that responds to "it caught fire" with a returns label is a screenshot. These route to a person immediately and to you, on whatever channel you tell us to use.
High-value orders and accounts you have flagged
You set the threshold — a dollar amount, a customer tier, a SKU. Above it, no automation touches the conversation, regardless of how routine the question looks.
a model we're not selling you
Per-minute and per-session pricing is clean, cheap and easy to compare, and for a store doing steady volume with simple questions it is a perfectly good answer. What it quietly does is make volume your problem and not ours. If a contact takes three attempts, we bill three times, and nothing in the arrangement pushes us to fix that.
That is why we put per resolved issue, from $1, nothing if it is not resolved at the top of the ecommerce page and nowhere else. It is the only line in our price list where a bad month for you is a bad month for us too. If we cannot agree on a definition of "resolved" for your store, that is a signal the model does not fit — and we would rather find that out in the first conversation.
And the reverse case is real: if your contacts are genuinely simple and repetitive, per-minute is the cheaper honest answer and we will tell you so. Every model in the table above stays open. We are recommending an order, not narrowing the menu.
when we're the wrong choice
questions
AI voice is $0.15 per minute and web chat is $0.05 per session. Live agents start at $300 per month including 500 minutes — about $0.60 per minute — or from $5.80 per hour for a dedicated agent working only your account. The minimum engagement is half a seat, not a five-seat block.
Five days is our normal window to staff a live account — that is five days to find and place an agent who fits your business, not five days of software setup. AI voice can go live sooner because there is no hiring step.
Yes, on the same number, with agents in our Bolivia and Nicaragua operation centres — native Spanish, not machine translation. English and Spanish share one queue, and we cover 60+ more languages if you need them.
That is the normal setup. You decide which call types go to AI and which reach a person, and you can change the split as your volume changes. Nothing about that decision is locked in at signup.
We agree the definition before we start, in writing, and it is specific to your store — typically that the customer's stated issue is closed and they do not contact again about the same order within an agreed window. If they come back, it was not resolved and it is not billed. We would rather argue about the definition up front than about an invoice later.
Inside them. Amazon runs through the SP-API, Shopify is two-way and eBay is connected, so an agent sees the order, the shipment and the return without asking the customer to repeat a number. If you run a platform outside that list, tell us which one and we will say straight whether we connect to it today or whether it needs building.
Yes — that is the normal setup. WhatsApp, LINE and TikTok come into the same queue as email and web chat, and the same agent sees what was already said on another channel. That matters more than it sounds: most repeat contacts in ecommerce are the same customer trying a second channel because the first one went quiet.
Per-minute and per-resolution billing absorb a volume spike without a hiring decision on your side. If you want dedicated people through peak rather than shared capacity, five days is our normal window to staff a live account — that is five days to find and place an agent who fits your business, not five days of software setup.
No, and most stores should not. The useful split is by the cost of getting it wrong, not by channel or by hour — the table above is the split we would actually recommend. You set where the line sits, and you can move it as volume changes.
start
We will walk you through live scenarios using your actual product data.