Mapbox reduces monthly support tickets by 20% with Kapa.ai

Mapbox provides customizable maps and location services for applications, enabling developers to integrate maps, navigation, and location-based features. Developers use Mapbox worldwide, with nearly 40 percent of Fortune 500 companies as customers, including other prominent global brands like Strava, BMW, and Wolt.

“We've seen fantastic results with Kapa.ai, recently achieving a 20% monthly reduction in support tickets and significant productivity gains for our Technical Support Engineers. Kapa.ai now handles over 11,000 questions monthly from developers via our public documentation, in-app assistance, and Discord community.”

————

KEY RESULTS (February to March 2024)

20% fewer support tickets

from paid Mapbox users

26% more questions to kapa.ai

with 10% of questions answered in non-English

2500+ support hours boosted

from 11,000+ monthly questions answered by kapa.ai**

————

THE SITUATION

Meeting the need for developer support

Supporting developers and improving the user experience is a top priority for Mapbox. As of May 2024, around 4 million developers use Mapbox, nearly 500,000 visitors access Mapbox’s documentation site monthly, and Mapbox has more than 11,000 members in their Discord community. 

“Knowing all the variations and product use cases of Mapbox takes time and is quite complex. Both for our users of Mapbox and our internal technical support engineers”

Mapbox is a collection of products and technologies with a corresponding large knowledge base. Developers are not expected to know all answers, and even as technical support engineers, it takes time to fully know all details. Developers often spend time searching for answers or waiting to get one, and if they cannot find what they need, they submit a support request. However, for many support requests, the answers already sit in the documentation or other sources.

————

THE SOLUTION

How Mapbox uses Kapa.ai to optimize developer support

Recognizing the potential of AI, Mapbox expanded its support capabilities by deploying Kapa.ai both externally and internally.

Mapbox and kapa partnered up to create a custom LLM that has (A) access to the extensive public and private knowledge universe curated by the Mapbox's team and community, and is (B) deployed where end user have questions across web, Discord, and Zendesk support tickets.

The kapa bot proactively resolves support tickets before they occur, and with the internal Zendesk assistent, the support tickets received can now be handled with greater speed and accuracy.

1. Knowledge Sources

Kapa has access to the following Mapbox resources with all sources kept up to date on a weekly basis:

Public sources (External project)

• Hundreds pages of developer documentation, SDKs and blog posts

• Extensive view of API references

• Over 3,000 accepted answers from Stackoverflow

• Close to 100 Youtube tutorials and guides

• Other Custom Answers and FAQ

Private sources (Internal project)

• Thousands of past Zendesk tickets

• Internal FAQ

• Selected Slack channels

B. Deployment methods

Kapa is available for end users across 1) Mapbox websites and application and 2) Discord community. In addition, Mapbox's internal project, used by technical support engineers, is deployed as an in-app 3) Zendesk assistant.

1.1 Mapbox documentation site

Online visitors and users can now click on the “Ask AI” button on the bottom right of the Mapbox docs and ask kapa any technical or other related questions about Mapbox's products. The Mapbox team used the customizable off-the-shelf widget solution from kapa.ai, making it simple and easy to configure and deploy.

1.2 Developer AI Assistant

Mapbox has introduced a direct Q&A setup, making it easier for developers to engage in immersive conversations with Kapa and bookmark the site for future reference. This site was done by creating an iframe and embedding the off-the-shelf widget.

1.3 Mapbox Application

Members can also ask questions to kapa directly in Mapbox's logged-in Account page. Similarly deployed like the docs with an "Ask AI" widget in the bottom right corner.

2. Mapbox Discord Community

Mapbox hosts a self-serve Discord forum called "ask-ai" in Discord where users can ask questions to kapa and interact with kapa in the created thread.

3. Zendesk assistant

Mapbox has installed the kapa.ai Zendesk Agent App. Kapa helps reduce the time support agents need to answer tickets by immediately suggesting a reply based on all Mapbox's knowledge sources as well as the full ticket history.

“Kapa is like your know-it-all friend. It dramatically improves the speed and accuracy of finding the answers to your tasks. For example, when an issue comes in, Kapa cuts the research process in half by identifying the topic, drafting an answer and referencing relevant sources that streamlines the support process in Zendesk.”

As a best practice, Mapbox also suggests users in their auto-response after contacting Mapbox support to check out the 'Developer AI Assistant' for immediate help with any questions. Mapbox even fill in the question and shares the response as a giphy.

————

THE FUTURE

Optimizing developer experience and enabling new use cases

Using AI to enhance developer engagement and support will remain a strategic priority for Mapbox as they continue their growth journey, bolstered by their recent $280 million Series E investment from Softbank.

“The role of our support engineers has become more powerful with Kapa. It can take years for a support engineer to know the ins and outs of Mapbox, but Kapa now helps develop support engineers into more knowledgeable support specialists in less time.”

A next step in enhancing support for Mapbox’s developers and onboarding of technical support engineers involves integrating outstanding internal sources such as Jira and Confluence. This will enable internal support agents to draw upon all of Mapbox’s internal resources, expanding Kapa.ai’s breadth of answers and depth of knowledge.  

In the near future, Mapbox is exploring how kapa can support the process of identifying and supporting sales leads, drawing from the hundreds of conversations it has directly with developers each day. Starting from an understanding of what developers are building and what questions they have is valuable information for identifying high-intent leads for sales conversations.

**Assuming each question asked to kapa.ai would take a person (e.g., technical support engineer) ~15 minutes to resolve.