Consumer Chatbot Technology

The Reality Behind AI Chatbots and the Path to Autonomous AI

In the rush to adopt the latest Consumer Chatbot Technology, it’s easy to overlook a fundamental reality: consumer chatbot technology isn’t ready for enterprise use—and it likely never will be. The reason is simple: AI assistants are only as effective as the data that powers them. Most large language models (LLMs) are trained on data from public websites, which lack the specific business and customer data that enterprises need. This means consumer bots can’t adequately assist employees in selling products, marketing merchandise, or improving productivity, as they lack the necessary personalization and business context.

To achieve the vision of AI that goes beyond simple chatbots performing basic tasks—like drafting emails, essays, blogs, or graphics—to a more advanced role where AI acts autonomously and addresses business-critical needs, a different approach is needed. This vision involves AI taking action with minimal human intervention, using digital agents to identify and respond to these needs.

At Salesforce, we are pursuing a clear path to AI that not only takes action but also automates routine tasks, all while adhering to established business rules, permissions, and context. Instead of relying solely on LLMs, which primarily focus on generating human-like text, future AI assistants will depend on large action models (LAMs) that integrate decision-making and action-taking capabilities.

The Journey Toward AI Autonomy

Our journey towards this vision began with the Salesforce Data Cloud, a robust data engine built on the Einstein 1 Platform. This platform integrates data from across the enterprise and third-party repositories, enabling companies to activate their data, automate workflows, personalize customer interactions, and develop smarter AI solutions.

Recognizing the shift from generative AI to autonomous AI, Salesforce introduced Einstein Copilot, the industry’s first conversational, enterprise-class AI assistant. Integrated across the Salesforce ecosystem, Einstein Copilot utilizes an organization’s data, whether it’s behind a firewall or in an external data lake, to act as a reasoning engine. It interprets user intents, interacts with the most suitable AI model, solves problems, generates relevant content, and provides decision-making support.

Expanding the Role of AI in Business

Since its launch in February 2024, Salesforce has been expanding Einstein Copilot’s library of actions to meet specific business needs in sales, service, marketing, data analysis, and industries like ecommerce, financial services, healthcare, and education. These “actions” are akin to LEGO blocks—discrete tasks that can be assembled to achieve desired project outcomes. For example, a sales representative might use Einstein Copilot to generate a personalized close plan, gain insights into why a deal may not close, or review whether pricing was discussed in a recent call. Einstein Copilot then orchestrates these tasks, provides recommendations, and compiles everything into a detailed report.

The ultimate goal is for AI not only to gather and organize information but also to take proactive action. Imagine a sales representative instructing their digital agent to set up meetings with top prospects in a specific territory. The AI could not only identify suitable contacts but also suggest meeting times, plan travel schedules, draft emails, and even create talking points—all of which it could execute autonomously with the representative’s approval.

Tectonic dreams of the day AI is smart enough to interpret our search engine typos and produce the results for what we were actually looking for!

The Future of AI Autonomy

The possibilities for semi-autonomous or fully autonomous AI are vast. As we continue to develop and refine these technologies, the potential for AI to transform business processes and decision-making becomes increasingly tangible. At Salesforce, they are committed to leading this charge, ensuring that our AI solutions not only meet but exceed the expectations of enterprises worldwide.

Salesforce is in a strong position to deliver on all of them because of the volume and breadth of data housed in Data Cloud, the heavy workflow traffic in our Customer 360 CRM, and the fact we’ve delivered an enterprise-class copilot that is rapidly expanding its library of actions. It will not happen overnight. The technology needs to advance, organizations and people have to be able to trust AI and be trained to use it in the right ways, and more work will need to be done to ensure the right balance between human involvement and AI autonomy. But with our continued investment in CRM, data, and trusted AI, we will achieve that vision before too long.

Salesforce is in a strong position to deliver on all of them because of the volume and breadth of data housed in Data Cloud, the heavy workflow traffic in our Customer 360 CRM, and the fact we’ve delivered an enterprise-class copilot that is rapidly expanding its library of actions.
Jayesh Govindarajan, Senior Vice President, Salesforce AI