Payments for online sellers, with a lifeguard team on WhatsApp
AbacatePay is a payments platform for businesses that sell online: Pix, boleto and credit card via API and checkout, with self-serve sign-up. Many of its customers are developers and SaaS founders integrating payments into their own products.
It's a product and engineering company, and it runs customer service the same way: the support team (the "lifeguard team") and the sales team work in the same Nuvia workspace, and the team itself builds and improves the funnels.
When the product scales, the support WhatsApp becomes the door to everything
As the customer base grew, the support number started receiving everything: integration questions, stuck sign-ups, limit increase requests and, in the middle of it all, people ready to buy. A customer with thousands of transactions a day asking for a better rate landed in the same queue as someone who couldn't set up a webhook.
A support analyst isn't a salesperson, and a salesperson can't answer webhook questions. Moving a customer from support to sales, and from sales to the technical side, was manual and got lost along the way.
When you start scaling the product, lots of people start reaching out on WhatsApp for support, for sales, for all kinds of things. Sometimes a guy shows up and says: "I already do 3,000 transactions a day, can you get me a better Pix rate?". That landed on the support number, and the support person sometimes doesn't know how to sell.
Tools that solved one piece of the funnel
AbacatePay had already tried other tools and attempted an in-house solution. Each one solved one part, but none connected the whole funnel. And there was a clear concern: every experience with AI handling customers felt robotic.
We had tried several tools and even tried building it in-house. When you look at the whole funnel, it's really hard to take someone from support and get them to sales and to the technical side. That's a bit of what Nuvia managed to do: integrating customer service across the whole funnel.
One funnel for each case, all talking to each other
AbacatePay organized its operation into separate funnels by case, in the same workspace and the same inbox:
- 01Support and triageReceives everything that reaches the support WhatsApp, understands the request and solves or routes it. AI is on.
- 02SalesReceives anyone who wants to sign up, negotiate rates or activate cards. Run by the team, with AI turned off at every stage.
- 03Activation and expansionCampaigns from the sales number to people who created an account but didn't send documents, and to those who can activate credit card. Replies land in the sales funnel.
The customer doesn't know there are funnels. They're just talking to AbacatePay.
Smart FunnelSee how one funnel per type of conversation worksThe conversation switches funnels when the case changes
A support question that turns into interest in credit card or a rate negotiation leaves the support funnel and enters sales right away. The AI makes the switch itself and tells the customer the sales team will take it from there. The other way around, a lead that came in through the sales number but was actually a support case goes back to the support funnel in one click.
In practice, 82% of switches between support and sales are made by the AI itself, with no one on the team having to move the conversation. The team only sends back to support what wasn't a sale.
Every experience I've ever had with AI handling customers, for sales or anything else, I thought was awful. Nuvia's I thought was really good.
Every team on the same contact, with history and data
Support, campaigns and sales are all recorded on the same contact. The data the AI collects in support (tax ID, email, website and business description) is saved and used by sales. When the customer comes back, with another issue or to activate cards, the team reads the history and picks up where it left off.
Campaigns use that context too: activation and card messages reach customers the team already knows from support, with the full history at hand.
ConversationsSee the inbox with each contact's full historyPeople step in at the right time, with the data in hand
The AI solves what's in the knowledge base, including technical integration questions. Anything that needs account access or a human decision, like a rejected sign-up, a webhook failure or a limit increase, goes to an analyst. Before handing off, the AI checks what's already on the contact, asks only for what's missing and assigns the conversation to a person.
After the handoff, the AI steps aside: it only reassures the customer, without touching the conversation. Tier-2 support is moved only by the team. Outside business hours, the case is already assigned and queued for first thing in the morning.
This interaction startled me. We never put this into the AI, but it learned it from some past ticket.
Built by the team itself, through Nuvia's MCP
AbacatePay doesn't open a ticket to change its customer service. The team connects Claude, Cursor or Claude Code to Nuvia's MCP and builds right from there: creates and edits funnels, writes the rules for each stage, creates custom fields and simulates conversations before publishing. A new rule, like "anything about cards goes to sales", goes live in minutes.
MCP isn't just an engineering tool. The CEO, the Head of Sales, the support analyst and the engineering team all use it: to tune the funnel, look up conversations, review campaigns and understand what customers are asking for.
- Reception and Service stages updated
- Conversation simulated before publishing
- New funnel version published
Integrated into the product through the API and webhooks
As an API company, AbacatePay plugs Nuvia into its own stack. Through the Nuvia API, data from the payments platform reaches contacts and conversations. Through webhooks, Nuvia events trigger actions in AbacatePay's systems. Customer service is part of the product, not an isolated tool.
API and WebhooksSee the Smart Funnel integrations and APIOne conversation, many teams
- 01Arrives on WhatsAppOften straight from the AbacatePay platform, with a question, a problem or interest in buying.
- 02Goes to the right funnelThe AI understands the request and decides: solves it in support, hands it to an analyst or takes it to sales.
- 03Switches teams without starting overIf the case changes, the conversation switches funnels with its history and data.
- 04Comes back and expandsWhen the customer returns, the team sees everything that happened. A support conversation can turn into a referral or a card activation.
One customer, many teams, one conversation
- JulySupport · AI → teamThe customer asks for help taking their account to production. The AI collects the tax ID and confirms the sign-up approval through the API.
- JulySupport · teamAsks for a higher withdrawal limit for the operation. The AI hands off to an analyst, who approves the increase the same day.
- AugustSupport · AI → teamPix failing on a payment link. The AI asks for details and a screenshot and hands off to technical support, where the analyst already has all the data, explains what happened and fixes it.
- OctoberActivation · campaignReceives a campaign from the sales number.
- OctoberSupport → salesComes back referring a new customer. The team reads the history, sorts out the referral's sign-up and offers commercial terms and card payments.
Thousands of tickets with a lean team, without sounding robotic
In less than three months, the Smart Funnel took over organizing AbacatePay's customer conversations on WhatsApp, across support, sales, activation and expansion. The AI handles more than a third of them on its own, and in more than half, AI and the team work together in the same conversation: big-team productivity with a lean team.
- 37%of conversations handled by AI alone, without the team having to step in
- 55%with AI and the team together: the AI triages and collects the data, the team decides
- 82%of funnel switches made by the AI itself
- 0tickets to the development team for each change to customer service
Nuvia was a game changer for us here at Abacate. We use it for support and sales. We managed to handle thousands of tickets with a super small team, without making everything robotic and stiff. Best of all, we didn't lose the human touch with our customers, and that made all the difference.
Every conversation becomes an insight
With support, sales and campaigns in one place, customer service became a source of data. The CEO, the Head of Sales and support look up conversations through MCP to understand what customers are asking for, review campaign results and decide the next tweak to the funnel and the product.
Smart FunnelSee the analysis of every conversation: summary, sentiment and qualificationSupport that generates revenue
People who want to buy don't wait in the support queue. Card requests, proposals and rate negotiations reach sales right away, and activation and card campaigns reach people who already know AbacatePay. Support stopped being just a cost and started opening opportunities.
CampaignsSee how to keep your base active with engagement campaignsContinuous improvement without a development queue
When a new type of request shows up, like the return of credit card or a recurring question, the team tunes the funnel through MCP the same day. Customer service keeps pace with the product.