Agentic AI Helps People with Disabilities at Big Events

This week, Salesforce showed how agentic AI can make big events more accessible and welcoming for people with disabilities. At the Dreamforce conference, Salesforce used Agentforce, which is an agentic AI tool, to work in a special help desk for people who needed extra support. The Agentforce Channel Expert was trained to answer questions that came in through Slack, which is a messaging tool many people use for work. Instead of making people wait a long time for a human to answer, this AI agent could give fast answers to regular questions like "Where is the quiet area in Moscone South?" or "What time does the accessible shuttle leave?". The results were really good. The AI agent answered 43 out of every 100 questions all by itself, which was even better than the 40% goal Salesforce set. This meant that the human helpers did not have to spend their time answering the same questions over and over again.

How the AI Helped Real People

The way Salesforce set up the AI was smart because it knew when to ask for help from a human. When a question was too hard or needed someone to really understand what a person needed, the AI would pass it to a human helper right away. This meant the people who worked at the help desk could focus on the really important work, like finding someone to guide a person who could not see, or helping someone get special help with seating or other things they needed. By splitting the work between the AI and humans, more people with disabilities got help faster. Salesforce trained the AI on lots of information about the event, and even tested it with different ways people might ask questions. The company used Google Gemini, which is a tool that helps make questions sound more real, to make sure the AI could understand people no matter how they asked for help. This kind of careful work shows that agentic AI can really help people when it is made the right way.

Workers Are Excited But Also Worried

While agentic AI is helping with things like accessibility at events, workers around the world have mixed feelings about it. According to a big survey by EY, which is a major company that studies business trends, 84% of workers in the United States said they were excited to use agentic AI in their jobs. These workers thought AI agents would make their work easier, faster, and more fun. A huge number—90% of workers who were already using AI agents—said the AI was helping their teams get more done. This is really encouraging news. However, the same survey showed a scary side. More than half of all workers, or 56%, were worried that agentic AI might make their job go away or become less important. Non-managers were even more scared, with 65% of them worried about their jobs, compared to 48% of managers. These workers wanted their leaders to talk with them about what was going to happen, but many bosses were not doing a good job of explaining how agentic AI would change their work.

The Problem with Not Talking and Not Training

Here is a big problem: companies are not teaching their workers how to use agentic AI properly. The EY survey found that 85% of workers were learning about agentic AI by themselves after work, and 83% said that most of what they knew about agentic AI they taught themselves. This is not good because workers should be getting help from their bosses and their companies. Only about half of senior leaders, or 52%, said their company had a full plan to train workers about agentic AI. When workers do not get training and their bosses do not talk to them about the future, workers get confused and worried. This makes it harder for companies to use agentic AI in the best way. The survey showed that companies that did a better job of explaining their agentic AI plans saw better results and workers felt more confident. This tells us that talking with workers and giving them training is super important, not just for making workers feel better, but for making AI work better too.

Safety Worries Around the World

In other parts of the world, people are also thinking about agentic AI, but they are worried about safety. In France and Germany, the people in charge of computer safety are saying that companies need to use zero trust architecture when they set up agentic AI. Zero trust means not automatically trusting anything, and checking everything carefully before letting it do something important. Experts say agentic AI can be tricked into doing bad things, and about 23% of IT workers have already seen AI agents get fooled into giving away passwords. There is also a real example of someone using agentic AI to steal information from at least 17 different companies. These kinds of problems show that agentic AI needs strong safeguards, which are like safety guards that keep bad things from happening. The good news is that companies are spending lots of money to make agentic AI safer. The market for AI safety is expected to grow from 20 billion dollars in 2023 to 141 billion dollars by 2032.

What This Means for the Future

The story of agentic AI and accessibility shows us something important: this technology can make the world more inclusive and help people, but only if it is done carefully. Salesforce proved that agentic AI can help people with disabilities get better and faster help at events. But at the same time, we see that workers need to understand how AI will change their jobs, and companies need to be honest and give them training. We also see that agentic AI needs to be kept safe and secure so bad people cannot use it to hurt others. Moving forward, the best approach is for companies to use agentic AI to help people do their jobs better, not to replace workers. When agentic AI works together with humans who know how to use it well, everyone can benefit. The challenge for organizations now is to move forward with agentic AI in a way that helps people with disabilities and other groups who need more support, while also taking care of workers' worries and keeping everything safe.