Pricing

Documentation

Sign in

Get started

Book a demo

Get started

Get started

+115k queries answered
+115k queries answered
+24k tickets deflected
+24k tickets deflected
537 sources indexed
537 sources indexed
CUSTOMER STORY / CRYPTO
CUSTOMER STORY / CRYPTO

How VALR deflects 70% of support tickets when money is in motion

“In crypto, a support ticket is rarely just a simple question. Someone’s money is in motion, and they need an answer now, not tomorrow. Delivering on that promise at scale, especially inside a regulated business, is of utmost importance to us.”

Gianluca Sacco

COO, VALR

Challenge

Most support AI handles static questions like fee structures and password resets. VALR’s ticket volume looks nothing like that: deposits that have not reflected because an EFT was missing its bank reference, withdrawals held for compliance review, users stuck mid-way through KYC, accounts locked after a change to account details. Routing those to a human queue worked until 1.9 million registered users pushed the support team to capacity.

Solution

VALR unified its knowledge ecosystems into a single Kapa knowledge base grounded in 537 internal sources, then deployed that one engine across three surfaces: a deflector on the support.valr.com form, a widget on the website and help centre, and an internal assistant for the support team. Every answer arrives with inline citations, and when Kapa is uncertain it says so and hands the conversation to an agent with the full interaction history attached.

Results

  • ~70% deflection rate on support form submissions

  • 10,000+ conversations resolved in a single month

  • 7,000+ unique users served across web, form, and internal channels

  • ~4,000 tickets deflected per month, ~48,000 projected annually

  • 660+ support hours saved every month

  • 4 full-time equivalents redirected to complex cases

THE CHALLENGE

High-stakes support at scale

Most agentic AI support simple, static queries like fee structures or password resets. VALR faced a different set of challenges: high-volume and complex troubleshooting where funds are actively moving.

The most frequent ticket clusters included:

  • Unreflected Deposits: Missing bank reference numbers on incoming EFTs.

  • Withdrawal Delays: Compliance and security hold-ups.

  • Account Verification: Users stuck mid-way through KYC or onboarding.

  • Account Access Locks: Security suspensions triggered by changes to account details.

Standard FAQ bots route these requests to human monitored queues. As user growth surpassed 1.9 million, reliance on human agents reached capacity, making instant, precise AI resolution critical to maintaining trust.

By unifying its knowledge ecosystems through Kapa, VALR deployed a single AI knowledge engine across three primary surfaces, deflecting nearly 70% of support form tickets and saving over 600 support hours per month without adding support headcount.

Raspberry Pi is a full-stack engineering company. It designs its own silicon, boards, and operating system, and it serves three very different audiences:
Industrial and embedded customers
Enthusiasts and educators
Semiconductor buyers
Each group comes to the documentation site with different use-cases and questions.
The core documentation at raspberrypi.com/documentation is hosted on a website, but a large share of the most valuable technical content is not. The Product Information Portal (PIP) holds the product briefs, datasheets, and whitepapers that industrial customers rely on, almost all of those are PDFs. Raspberry Pis Official Magazine publishes issues that run to hundreds of pages each, also as PDFs. Some tooling, such as rpi-image-gen, documents itself inside its GitHub repository in code.
Keyword search on the docs site only surfaced the core documentation, and as Gordon put it in the announcement post: “with keyword search you have to guess the author's vocabulary.”
Gordon had already prototyped a retrieval-augmented generation (RAG) system of his own, so he understood what the technology could do and what it would take to run one in production. The question was whether to keep building or to buy, and if buying, from which vendor.

THE SOLUTION

One engine across multiple surfaces

VALR integrated Kapa as a centralised knowledge layer grounded in 537 internal sources—spanning help centre documentation, educational blog posts, trading competition guidelines, and product announcements.

How the Support Form Deflector Works:

  1. Contextual Processing: When a user completes a form on support.valr.com, Kapa analyses the query in real-time.

  2. Cited Answers: Kapa supplies an immediate, accurate response with inline sources. If uncertain, Kapa safely admits when it has limited knowledge.

  3. Seamless Transfer: If the issue requires human intervention, the user submits the form as normal. The entire Kapa interaction history transfers directly to the agent, eliminating repetitive back-and-forth.

Raspberry Pi is a full-stack engineering company. It designs its own silicon, boards, and operating system, and it serves three very different audiences:
Industrial and embedded customers
Enthusiasts and educators
Semiconductor buyers
Each group comes to the documentation site with different use-cases and questions.
The core documentation at raspberrypi.com/documentation is hosted on a website, but a large share of the most valuable technical content is not. The Product Information Portal (PIP) holds the product briefs, datasheets, and whitepapers that industrial customers rely on, almost all of those are PDFs. Raspberry Pis Official Magazine publishes issues that run to hundreds of pages each, also as PDFs. Some tooling, such as rpi-image-gen, documents itself inside its GitHub repository in code.
Keyword search on the docs site only surfaced the core documentation, and as Gordon put it in the announcement post: “with keyword search you have to guess the author's vocabulary.”
Gordon had already prototyped a retrieval-augmented generation (RAG) system of his own, so he understood what the technology could do and what it would take to run one in production. The question was whether to keep building or to buy, and if buying, from which vendor.

“What gave us confidence in Kapa is its ability to use our knowledge base to provide real-time, accurate solutions while highlighting our coverage gaps. It answers confidently on the vast majority of questions and gracefully steps aside on the rest. Because of this, we're never left guessing about what our customers need or how we should be helping them.”

Gianluca Sacco, COO @ VALR

Raspberry Pi is a full-stack engineering company. It designs its own silicon, boards, and operating system, and it serves three very different audiences:
Industrial and embedded customers
Enthusiasts and educators
Semiconductor buyers
Each group comes to the documentation site with different use-cases and questions.
The core documentation at raspberrypi.com/documentation is hosted on a website, but a large share of the most valuable technical content is not. The Product Information Portal (PIP) holds the product briefs, datasheets, and whitepapers that industrial customers rely on, almost all of those are PDFs. Raspberry Pis Official Magazine publishes issues that run to hundreds of pages each, also as PDFs. Some tooling, such as rpi-image-gen, documents itself inside its GitHub repository in code.
Keyword search on the docs site only surfaced the core documentation, and as Gordon put it in the announcement post: “with keyword search you have to guess the author's vocabulary.”
Gordon had already prototyped a retrieval-augmented generation (RAG) system of his own, so he understood what the technology could do and what it would take to run one in production. The question was whether to keep building or to buy, and if buying, from which vendor.

Performance metrics

Key results and impact

  • ~70% Ticket Deflection Rate: Achieved on support form submissions for account-specific troubleshooting traffic.

  • 10,000+ Conversations Resolved: Grounded resolutions delivered to over 7,000 unique users across web, form, and internal channels.

  • ~4,000 Tickets Deflected / Month: Projected at ~48,000 tickets annually.

  • 660+ Hours Saved / Month: Based on an estimated 10 minutes saved per deflected ticket.

  • 4 Full-Time Capacity Equivalent: Equivalent support capacity redirected from routine triage toward complex, high-touch user cases.

Raspberry Pi is a full-stack engineering company. It designs its own silicon, boards, and operating system, and it serves three very different audiences:
Industrial and embedded customers
Enthusiasts and educators
Semiconductor buyers
Each group comes to the documentation site with different use-cases and questions.
The core documentation at raspberrypi.com/documentation is hosted on a website, but a large share of the most valuable technical content is not. The Product Information Portal (PIP) holds the product briefs, datasheets, and whitepapers that industrial customers rely on, almost all of those are PDFs. Raspberry Pis Official Magazine publishes issues that run to hundreds of pages each, also as PDFs. Some tooling, such as rpi-image-gen, documents itself inside its GitHub repository in code.
Keyword search on the docs site only surfaced the core documentation, and as Gordon put it in the announcement post: “with keyword search you have to guess the author's vocabulary.”
Gordon had already prototyped a retrieval-augmented generation (RAG) system of his own, so he understood what the technology could do and what it would take to run one in production. The question was whether to keep building or to buy, and if buying, from which vendor.

What's next

By anchoring its help centre articles, the website, internal workspaces and Kapa AI Agent to a single AI layer, VALR created a support system that scales seamlessly with platform growth. VALR support agents remain focused on complex user inquiries, while Kapa independently handles routine troubleshooting with accuracy and speed.

Raspberry Pi is a full-stack engineering company. It designs its own silicon, boards, and operating system, and it serves three very different audiences:
Industrial and embedded customers
Enthusiasts and educators
Semiconductor buyers
Each group comes to the documentation site with different use-cases and questions.
The core documentation at raspberrypi.com/documentation is hosted on a website, but a large share of the most valuable technical content is not. The Product Information Portal (PIP) holds the product briefs, datasheets, and whitepapers that industrial customers rely on, almost all of those are PDFs. Raspberry Pis Official Magazine publishes issues that run to hundreds of pages each, also as PDFs. Some tooling, such as rpi-image-gen, documents itself inside its GitHub repository in code.
Keyword search on the docs site only surfaced the core documentation, and as Gordon put it in the announcement post: “with keyword search you have to guess the author's vocabulary.”
Gordon had already prototyped a retrieval-augmented generation (RAG) system of his own, so he understood what the technology could do and what it would take to run one in production. The question was whether to keep building or to buy, and if buying, from which vendor.

“Preparing for VALR’s next ten million users doesn't mean just hiring more agents anymore. Thanks to Kapa, we can automate repetitive inquiries and equip our team with a reliable AI knowledge base, allowing us to scale exceptional customer service without compromising on trust.”

Gianluca Sacco

COO, VALR

Everything you need to give your agents accurate retrieval, ready in minutes.

Get started for free. Connect your sources and get an API key or hosted MCP server to give your agents the context they need to do reliable work at scale.

Get started