Salesforce Unveils its Four Pillars of Enterprise AI Success

Salesforce today announced new research and outlined four key areas for enterprise AI success, setting the stage for its upcoming Dreamforce 2024 event. Clara Shih, CEO of Salesforce AI, emphasized the critical juncture we’re at with AI adoption: “This is like 1998 was for the Internet. There’s tremendous opportunity, but also a lot of hype. Our research focuses on separating signal from noise and finding high-impact enterprise use cases.”

Salesforce’s “Trends in AI” research report is a comprehensive analysis of the current state of AI adoption in the enterprise. The study, which surveyed thousands of business decision-makers, technical decision-makers, and members of the general population worldwide, explores the challenges, opportunities, and trends shaping AI implementation across various industries.

The AI FOMO Phenomenon

Shih highlighted a growing sense of urgency among business leaders regarding AI adoption, describing what can only be termed as “AI FOMO” (Fear of Missing Out). This sentiment is backed by hard data from Salesforce’s research.

“The vast majority of business executives, unsurprisingly, we found in our report, fear they’re missing out on the advantages of AI,” Shih noted. “And it’s a well-founded fear.”

This fear is not unfounded, as the potential impact of AI on business performance is significant. Shih drew a parallel to underscore the magnitude of the current AI revolution: “This today is like 1998 was for the Internet, for AI, and there’s tremendous opportunity, there’s also a lot of hype.”

The research reveals that 88% of business leaders fear they’re missing out on the advantages of AI. This anxiety underscores the urgency for clear, actionable strategies in AI implementation. Companies are caught in a difficult position – they recognize the transformative potential of AI but are also wary of the risks and challenges associated with its adoption.

Shih emphasized that this fear is driving rapid experimentation and adoption across industries: “I think everyone is feeling excitement and also pressure in a good way.” She added, “A lot of those use cases, sometimes it takes a few experimental cycles, and a few years to be able to uncover and I think we’re very much in that moment now.”

The Four Pillars

This AI FOMO is pushing companies to seek out solutions that can help them quickly leverage AI’s potential while navigating its complexities. It’s this dynamic that Salesforce aims to address with its four-pillar approach to enterprise AI success, providing a framework for companies to move beyond the fear of missing out and into concrete, value-driving AI implementations.

Trusted Enterprise-Grade Solutions

Shih highlighted the importance of trust in AI deployments. “The majority of people using AI for work find it difficult to get what they want out of base models, because so much more context and tuning is required to make a model really performant, reliable, and secure for workplace applications,” she explained.

Salesforce’s Einstein Trust Layer addresses this with features like citations, audit trails, and data masking. Shih elaborated, “Trust is our number one value, and it aligns with the top issue for buyers. Our Einstein Trust Layer covers citations, audit trail, data masking, permissioning – all the data sharing rules that companies have put into Salesforce get reflected and honored by Einstein.”

The research found that 74% of the general population is concerned about the unethical use of AI, highlighting the critical need for trust-building measures in enterprise AI solutions.

Data Quality and Integration

The research underscores the crucial role of data in AI success. “86% of analytics and IT leaders agree that AI outputs are only as good as data inputs,” Shih noted. She further explained the challenges of data integration: “Analysts estimate that over 80% of data in companies today is unstructured, in the form of emails, Slack transcripts, phone transcripts, videos, sales enablement materials, company policies, product documentation.”

Salesforce’s Data Cloud aims to unlock siloed enterprise data, connecting both structured and unstructured sources to power AI applications. Shih provided an example: “Many companies will log their customer and prospect website visit activity into a data warehouse or data lake. If a salesperson’s customer visits the company’s website and views certain product pages and registers for certain product webinars, that’s very material information that you want to let the salesperson know right away. You don’t want that to be a quarterly report that has to be manually pulled.”

Quick Time-to-Value with Turnkey Solutions

Salesforce is focusing on turnkey solutions to address the frustration many companies face with long implementation times. Shih emphasized, “It’s been over 18 months now since ChatGPT ignited the world’s imagination around the possibility of using AI for business and many other applications. And yet, a lot of companies are still struggling with build and they’re also struggling with buy because they don’t want to buy a solution that’s overly siloed and difficult to customize.”

She continued, “We’ve seen customers get up and running in a matter of minutes, then take days to customize, and they’re off to the races. They’re able to deploy into production and see real ROI in the form of reduced costs, reduced customer handle time, increased customer satisfaction, higher accuracy, more frequent first-time resolution, and higher sales and marketing conversions.”

Choice and Flexibility in AI Models

Recognizing the rapidly evolving AI landscape, Salesforce is adopting an LLM gateway approach. “Four out of five leaders say they’re already using multiple models,” Shih reported, explaining that different models excel at different tasks and come with varying cost implications.

She elaborated, “Different models are good for different tasks, there’s different size models. You can fine-tune or customize models for certain tasks like question answering or summarization or translation. And so rather than having to bring the whole kitchen sink every time, you start to see more specialized models that are really good at one or more of those tasks.”

Shih also touched on the economic considerations: “The costs really add up if you’re using the state-of-the-art, largest foundation model every time. It becomes pretty untenable, cost-wise, and it’s also overkill, not to mention these large models sometimes have high latency.”

Practical AI Use Cases: From Sales to Service

While discussing the potential of AI, Clara Shih provided concrete examples of how Salesforce’s AI solutions are already delivering value in real-world scenarios, particularly in sales and customer service contexts.

Sales Cloud Use Case: Comprehensive Account Overviews

Shih highlighted a powerful use case within Sales Cloud: “Being able to summarize a call and get an overview of everything that’s happening at an account right in B2B.” She explained the complexity of B2B sales environments:

“Think about the number of people at the vendor organization that interact with a number of people at the customer organization. You’ve got customer success, you’ve got legal, you might have finance, marketing, multiple people on the sales team, right, an account executive, a sales manager, a regional vice president, a sales engineer, professional services partner.”

In such complex scenarios, Salesforce’s AI solution provides a crucial advantage: “Our Einstein platform is used, we have an out-of-the-box template that makes it very easy for any member of the selling team or the account team to see a summary of what’s happening and be able to walk into that meeting fully informed and fully prepared to have the best possible meeting.”

This use case demonstrates how AI can synthesize vast amounts of data from multiple touchpoints, providing sales professionals with actionable insights and saving hours of preparation time.

Service Cloud Use Case: Intelligent Customer Service Replies

In the realm of customer service, Shih described their most popular deployed solution: “By far the most popular deployed out-of-the-box solution we have is the customer service reply recommendations. This is where call centers are just being inundated every day with calls, chats, messages, emails, customer inquiries that either require a prompt response, answer or prompt action.”

The AI solution addresses this challenge by automating the process of finding and compiling relevant information: “Instead of that poor customer service rep having to search through pages of product documentation and knowledge articles and match that up to exactly that particular customer’s warranty and configuration, Einstein is able to automate that with our new knowledge-grounded replies.”

This functionality extends beyond human agents to AI-powered chatbots: “That works both for the contact center rep, it also works within Einstein bots, which is the next generation of generative AI-powered conversation assistance for Customer Self Service.”

Where’s the Beef?

While Salesforce’s four-pillar approach addresses key challenges in enterprise AI adoption, the industry still faces significant hurdles. Despite the hype surrounding AI, widespread adoption remains elusive. Many organizations struggle with implementation, data integration, and realizing tangible benefits from their AI investments.

The pressure on companies to show ROI is mounting, as Shih acknowledged: “I think everyone is feeling excitement and also pressure in a good way. It actually reminds me a lot of the late 90s, where some people asked, is the internet overhyped? And the answer was yes, because in the hype cycle, there’s always a lot of investment, and not everything is going to pan out, but the areas that will pan out will be extraordinarily transformational and extraordinarily valuable.”

Low-Code AI Solutions

Salesforce’s history of delivering low-code solutions positions it well to bridge the gap between AI’s potential and practical business applications. However, the pressure is on for the company to translate its vision into widespread adoption. As Clara Shih aptly put it, ‘It’s still early.”

The remainder of 2024 and into 2025 will be a crucial time in determining whether Salesforce can leverage its CRM dominance to make AI truly accessible and valuable to business users, moving beyond the realm of data scientists and into the hands of everyday professionals. Shih emphasized this point: “What makes the models come to life are, number one, having that contextual data about who the end user is and who the customer is. The second critical requirement is the Einstein Trust Layer. The third learning that we’ve had is just how critical it is for the AI to be in the flow of where the employee is working.”

While the foundation is laid, the true test will be in the execution and the tangible results businesses achieve with these AI solutions. At Dreamforce, we hope to see more evidence of widespread usage outside of well-known early adopters. Salesforce’s approach, focusing on trust, data integration, rapid deployment, and model flexibility, appears well-aligned with enterprise needs. However, the AI landscape is evolving rapidly, and Salesforce will need to continue innovating to maintain its position as a leader.