From long waits to instant support: How our AI agent Amy changed affiliate engagement

Inside Deriv’s AI partner support for web chat and WhatsApp across 190+ countries and 57 languages.

117,000+ partners. 190+ countries. Dozens of languages. One question: How do you make every single partner feel supported, instantly, at any hour?

You don’t hire more people. You build Amy.

The maths was never going to work

Thousands of active partners spread across 190+ countries, from Kenya to Pakistan, from Bahrain to Brazil, each running their own business, each needing support, answers, and guidance to grow. Thousands more aspiring partners waiting to join.

For years, a dedicated Global Partnership (GP) team of 90+ people worked tirelessly to make it work. They answered questions on WhatsApp, handled web chat, educated partners about new products, explained commission structures, and tried to maintain consistent knowledge across dozens of regions and languages. They did remarkable work.

But the maths was never going to work.

90 people. 117,000+ partners. 190+ countries. Multiple languages. Multiple channels. 24 hours a day, 7 days a week.

The scale had outgrown what any team could manually manage. Partners were waiting hours for replies. New partners sat in an onboarding queue with no one to guide them, quietly losing momentum. Something had to change. And we knew it had to be us who changed it.

And it was costing everyone

When Deriv launched new products or updated its offerings, knowledge had to travel through the GP team before it could reach partners, creating lag and sometimes inconsistency across regions. A partner in Africa might get a slightly different answer than one in Southeast Asia. Not because anyone was careless, but because human knowledge transfer at scale is imperfect by nature.

Meanwhile, the Customer Experience (CX) team was handling a separate stream of partner queries coming through web chat, operating in their own silo, with no bridge between the two worlds. Partners who needed help from CX were on WhatsApp. Partners who needed their country manager were on web chat. Misdirected, frustrated, and sometimes giving up entirely.

And the GP team, hired to drive partner growth, identify opportunities, and build relationships, was spending the majority of their time answering the same repetitive questions, day after day.

The data tells the story plainly. Before Amy, median first response times from country managers ranged anywhere from several hours to multiple days. Partners were waiting. And waiting.

Amy’s median first response time since launch? Consistently under 25 seconds.

That gap is not a minor improvement. It’s a structural transformation.

Meet Amy

Amy is Deriv’s AI partner engagement agent, the always-on face of the partner programme. She talks to partners, replies to them, helps them get onboarded, educates them about the programme, explains commission structures, encourages them when they’re stuck, and supports them through every stage of their partnership journey with Deriv.

What makes her different is the scope. Amy carries the weight of two teams, Global Partnership and Customer Experience, operating across WhatsApp and web chat simultaneously, in any language, at any hour, without a queue.

She was built not to replace the people behind those teams, but to do what was structurally impossible for them to do alone: be present for every partner, everywhere, all the time.

What changed when Amy arrived

The impact of Amy shows up in two places: in how partners experience Deriv, and in what the business is now able to do.

For partners, engagement has transformed.

Consider what it used to look like. A partner in Bahrain sends a WhatsApp message on a Friday evening asking about their commission structure. It’s the weekend. Nobody is available. They wait. By Monday, the moment has passed, the motivation, the question, the momentum.

Now that same partner sends a voice note at 11 pm on a Friday. Amy listens, understands, and responds in seconds, in their language, with their actual data, with a chart showing exactly what they’ve earned. The partner doesn’t wait. They don’t repeat themselves. They move forward.

That shift, from waiting to instant, from generic to personal, is what’s driving the measurable lift in partner engagement we’re seeing across markets.

For the business, it’s been a structural unlock.

Two teams’ worth of partner-facing work handled by one agent means the GP team is now focused on exactly the work they were always meant to do:

• Identifying churn risks and early warning signals

• Spotting growth opportunities across markets

• Gathering market intelligence

• Driving onboarding and activation

• Improving operational efficiency

• Monitoring revenue health and sustainability

How Amy actually works

Amy’s impact isn’t accidental. It’s the result of choices we made deliberately, about what to build, how to build it, and what we refused to compromise on. We engineered her to handle the full complexity of human communication, not in a controlled, structured environment, but in the wild unpredictability of real partner conversations happening across WhatsApp and web chat, in dozens of languages, at all hours.

Here’s what that looks like in practice.

Pillar 1 — Human-level communication intelligence

A partner in Bahrain sends five WhatsApp messages in 90 seconds. Three questions, two voice notes, one screenshot, all before Amy has responded to the first. This is normal. This is how people actually communicate.

Amy handles all of it.

She stitches together fragmented multi-message conversations, consolidates everything being asked, and responds once, clearly, completely, and in context. She listens to voice notes, reads screenshots, and processes text simultaneously. If a partner communicates entirely in their native language, Amy responds in that language without being asked. She is currently fluent in 57 languages, not translated, but contextually fluent. She doesn’t convert words from one language to another. She understands the cultural and conversational context of what’s being said, and responds in kind.

This is the baseline. Every partner, every message, every format, every language. But understanding how a partner communicates is only half the story; the other half is knowing who they are.

Pillar 2 — Personalised partner intelligence

A partner in Kenya types: ‘How much have I earned in the last 6 months?’ A generic bot sends a link to a dashboard. Amy does something different.

She doesn’t point them anywhere. She fetches their actual commission data, composes a response, generates a trend chart, and delivers it directly in the conversation. Personal, visual, instant.

When a partner asks for their referral links, Amy retrieves their specific links so they can go and market Deriv immediately, no friction, no redirects.

When a new partner wants to understand the commission plans or learn how to promote Deriv effectively, Amy doesn’t just explain; she pulls relevant video tutorials from Deriv Academy and shares them in context. Every new partner gets a guided onboarding experience tailored to where they are and what they need to know next.

Pillar 3 — Campaign intelligence pipeline

One thing we kept hearing internally was this: campaign knowledge was being created but never reaching the right partners fast enough. So we built a pipeline directly into Amy. Marketing and content teams now run personalised campaigns targeting specific partner segments, and Amy is the delivery engine. But the way it works internally is what makes it remarkable.

Teams across the company feed campaign knowledge directly into a dedicated Slack channel called amy_office. They specify the campaign content, the target partners, and the context Amy needs to respond appropriately. Amy ingests this knowledge and, from that point, delivers campaign information only to the partners it’s intended for. A partner outside the target group who asks about the campaign gets no hint it exists.

No complex integrations. No separate campaign tool. Just a Slack message, and Amy knows exactly what to say, to whom, and when.

This is just the beginning

Amy is live. She is operating globally, engaging partners across 190+ countries, in 57 languages, across two channels, around the clock. She is doing what was structurally impossible before, being present for every partner, at every stage, without a queue.

The vision is simple: every partner, everywhere, feels like they have a dedicated relationship manager available at all times. Not a bot. Not a queue. A presence.

But this is version one.

In the next post, we go under the hood. How did we architect an agent that works across two completely different platforms without being rebuilt for each one? What happens when a frustrated partner switches from WhatsApp to web chat mid-conversation, and how does Amy carry every word of that context across without asking them to repeat a single thing? And what does it take to engineer a response in under 25 seconds, at scale, without degrading quality?

We’re just getting started. Subscribe to Deriv<ed> to follow the Amy series, and if you’re building something similar or want to discuss, we’d love to hear from you in the comments.

Mahantesh Ambi is an AI Engineering Manager at Deriv.

Follow our official LinkedIn page for company updates and upcoming events.

Explore Deriv careers.