Salesforce the customer management giant
Advertisement
Salesforce stands for software that can be used to manage all of a company's customer contact. The second largest software group in the world also includes providers such as the communication platform Slack and Tableau, which provides analytical software. Artificial intelligence didn't just become an issue for Salesforce with the advent of ChatGPT and the associated hype. Liesel Klokkers, Senior Director Solution Engineering, reports on her own AI applications in this deep dive. Above all, she explains why AI agents will play an important role in the entire working world in the future and how Salesforce is trying to prevent hallucinations.
From predictive AI to generative AI
At Salesforce, the use of AI began with so-called predictive AI, long before ChatGPT. “It’s about making predictions from existing amounts of data,” says Klokkers. This means that the sales planning was validated. Generative AI initially moved into the company itself in the form of a chatbot that can be accessed via Slack. To ensure that all internals remain internal, Salesforce has developed its own Einstein app, i.e. its own chatbot. New are the AI agents, which are offered via a platform called Agentforce. “Customers are adapting this incredibly quickly because the speed with which software solutions can be brought live is placed on a completely new basis.” Just a few weeks after the platform was released, Salesforce was able to look back on more than 200 major customers, as shown in the quarterly figures .
AI agents instead of automation
According to Klokkers, there were three major waves of AI: “Predictive AI. I have a lot of data and can make predictions based on it and derive next big actions.” Then the co-pilots came, “who helped summarize things, create texts, collect content in the company.” And now, says Klokkers, they come third wave, the AI agents, “who can not only create and summarize information, but can act independently based on this information and can define processes independently.” The promise of Salesforce and other AI providers: Agents take over the annoying tasks, Employees can rely on the focus on creative and exciting processes. Agents should be able to complete repetitive tasks independently, but always hand in hand with people. “Do we have to take care of jobs?” Klokkers says that customers are more likely to report that they had not previously found any employees for the tasks that agents are now taking on.
How intelligent is artificial intelligence? What consequences does generative AI have for our work, our leisure time and society? In Heise's “AI update” we, together with The Decoder, bring you updates on the most important AI developments every weekday. On Fridays we look at the different aspects of the AI revolution with experts.
Klokkers uses the example of a chatbot used in customer contact to explain the difference between automation and an AI agent. With an old chatbot, you previously had to “manually anticipate what questions a customer would ask when contacting this chatbot and had to program out all of these eventualities.” Defined if-then relationships were always set in advance for all possible cases. So-called large action models can “independently analyze what the task is and also create an execution plan from given actions and decide independently how I have to complete a task.” According to Klokkers, this makes AI agents much “more powerful” and above all “more flexible”. However, you also need data – and some of it must be structured, i.e. already present in a system, but it can also be unstructured, for example in the form of operating instructions, which are available as PDFs. At Salesforce, the data is summarized in the cloud in real time. At the same time, there are options for action available to the AI agent in order to satisfactorily trust customers. In the end, he should leave the chat happier than before.
(emw)
