Each year, thousands of the world’s most influential leaders—CEOs, heads of state, and changemakers—gather in Davos, Switzerland, for the World Economic Forum’s five-day Annual Meeting. These discussions drive global action on urgent issues, from preparing billions of people for the jobs of the future to restoring biodiversity in areas like the Amazon Basin. But with more than 430 sessions, five marquee themes, and personal meetings to prioritize, setting a Davos itinerary is a high-stakes logistical puzzle. It’s not unusual for attendees to dedicate entire teams to researching and planning their Davos agenda.
These aren’t just any attendees—they’re among the most influential people on the planet. From heads of state and Fortune 500 CEOs to civil leaders and young change-makers, the leaders at Davos influence economies, industries, and policies worldwide. Nowhere else in the world does such a diverse group of global change-makers gather to discuss pressing international issues like sustainability and geopolitics.
As with any event, the World Economic Forum’s Annual Meeting attendees had questions to answer and problems to solve. They needed a fast, seamless way to navigate the expansive venue and make the most of their packed schedules, quickly answering questions such as finding out where someone is speaking, finding sessions on specific topics, or locating shuttles.
Scaling high-touch concierge at scale using Heroku, Salesforce, and Agentforce
Due to the curated guest list, most attendees couldn’t bring personal assistants. Instead, they relied on primary engagement managers (PEMs) as trusted advisors to provide on-site support at the event. PEMs answered pressing questions like “What ecology sessions can I attend on Thursday afternoon?”
But to answer these questions, PEMs had to find time during their busy days to search across separate resources for session details, venue information, and shuttle schedules. Digging for answers limited their ability to apply their deep expertise in industry strategy and policy—a key reason they were selected for the role.
Plus, with eight or more companies assigned to each PEM—many with multiple attendees—their days could fill up quickly. If a PEM wasn’t available, the World Economic Forum offered five service kiosks, but neither provided the 24/7 support necessary during such a busy event.
Building the World Economic Forums’s AI concierge for world leaders
Rather than give attendees yet another information resource, Annual Meeting planners were given the mission to build an AI-powered concierge directly into the World Economic Forum’s event app. Thanks to Heroku and close collaboration between Salesforce and the World Economic Forum, they were able to have the first version deployed within a month, with testing, enhancements, patching gaps, updating UI, and other tasks already underway.
I sent a message in the shared external channel, and I was like, ‘Hey guys, I need a new app on Heroku to host this chat.’ And the World Economic Forum’s head of engineering responded, ‘All right, I just shared access. I’ve already spun one up for you.’ No, ‘What’s the solution? How are we going to share credentials?’ It was literally, ‘I spun it up, you have access.’ We deployed in two days, and that was such a good feeling.
They were able to accomplish this feat by leveraging the existing Ask Astro solution, building the app in React, and deploying on a Heroku Dyno. Salesforce acted as a single source of truth, and Heroku Connect synchronized data from Data 360, Agentforce Sales, Agentforce Marketing, and internal knowledge articles into their Heroku ecosystem. This allowed the agent to provide on-demand answers to a broad variety of questions about everything from session details and transportation to venue logistics and customer service.
Because the World Economic Forum had already created this single source of truth in Salesforce by ingesting their attendee support data into Data 360, they were able to give context to the support actions they configured in Agent Builder and launch Agentforce in just five weeks. The custom chat interface, hosted on Heroku, integrated directly with Agentforce via API, which also paved the way for a smooth implementation.
We have Salesforce as a single source of truth, and we use Heroku to build our APIs on. So we are heavy users of Heroku Connect to synchronize the data from Salesforce to our Heroku ecosystem, and we’ve built a microservice architecture with many languages which allow our team to use the best technologies for what they’re good at.
Heroku also opened the door to more opportunities, enabling the team to develop a microservice architecture using multiple programming languages, allowing them to select the most suitable technologies for each task. For example, they use Python for machine learning and Ruby for lightweight microservices. This approach has allowed them to scale delivery teams quickly and consistently deliver value whenever a new product is proposed for their constituents. By taking care of platform concerns, Heroku allows the team to focus directly on building applications that generate meaningful impact.
Heroku and Slack fill agent knowledge gaps
A concierge—whether digital or human—is only as effective as the insights it provides. When a conversation stalls, it often means the agent couldn’t answer the user’s question, making inactivity a valuable metric for evaluating agent performance. To track this, the team used a Heroku scheduler to compile inactive conversations and run them through an LLM to assess whether the user’s question had been resolved. If not, the system automatically alerts the web team via Slack, prompting them to create new knowledge articles and close the content gap. This automation ensured the app could respond to unexpected user queries.
Keeping the agent secure
With such a high-profile event, the security and reliability of Agentforce was paramount. Salesforce partnered with the World Economic Forum to rigorously stress-test the model. The Office of Ethical and Humane Use even tried to “trick” Agentforce into providing incorrect responses or straying from preset instructions, but wasn’t able to: No matter the prompt, Agentforce responded with compliant, courteous answers.
Agentforce helps attendees find answers quickly
The convenience of instant answers helped attendees stay informed, on schedule, and connected without the hassle of searching through emails, navigating multiple platforms, or waiting for busy PEMs to respond. All attendees needed to do was open the Forum Live event app and start a chat with Agentforce.
With each active user asking roughly five questions a day, the agent resolved 75% of attendee inquiries at Davos 2025 with accurate, automated answers. Plus, with Agentforce, attendees were able to surface sessions based on topic or day and scroll to learn more.
Our goal is to create the right environment—physically and digitally—for change-makers to connect. Agentforce helps them get answers in seconds, so they can focus on what matters: relationships and dialogues that drive real-world impact.
What’s next for the World Economic Forum?
As adoption grows, the World Economic Forum plans to use Agentforce to enhance constituent and employee experiences across all events, delivering intelligent, personalized interactions at every touchpoint—even past the annual meeting. With Agentforce, the World Economic Forum is paving the way for the future of global events, where technology and human ingenuity work together to create experiences that inspire change.
Launch your next project with Heroku
How the World Economic Forum did it
- Agentforce provided the agentic layer to build and deploy AI agents
- Ask Astro provided the blueprint for the new app
- React was used to build the app
- Heroku Dynos were used to host the app
- Heroku Scheduler was used to automatically identify agent knowledge gaps and prompt the web team to resolve via Slack
- Data 360, Sales Cloud, and Marketing Cloud provided information to power on-demand answers
- Heroku Connect synchronized the data from Salesforce into the Heroku ecosystem
