Introduction

The speaker introduces the topic of artificial intelligence and its importance.

Importance of AI

• AI is one of the most important technologies of our lifetime.

• AI has forever changed our world, but we need to ask more from it and build trust into every experience.

• Salesforce has developed a new platform that connects to customers in a whole new way.

Welcome Speech

The CEO thanks the audience for their support and introduces a customer who has inspired them with their use of generative AI.

Thanking the Audience

• The CEO thanks each and every person in attendance for their support.

• He expresses gratitude for the continued relationship between Salesforce and its customers.

Customer Inspiration

• A representative from Gucci is introduced as an inspiration for generative AI.

• Gucci's journey with Salesforce is discussed, including how they tested using AI to augment their advisors' capabilities.

• Positive results were seen, making Gucci "business case Zero" in luxury arena.

Augmenting Capabilities with Generative AI

The speaker discusses how generative AI can augment human capabilities, using Gucci's experience as an example.

Augmenting Service Agents' Capabilities

• Service agents become augmented through generative AI, gaining more capabilities such as marketing, sales, and commerce.

• This could be one of the great promises of generative AI: augmenting human capabilities.

Introduction

The CEO of Salesforce introduces the speakers and talks about the growth of Salesforce, their core values, and their commitment to giving back to the community.

Speakers

• Clara Shih, CEO of Salesforce AI

• Patrick Stokes, EVP of Product and Industries Marketing

• Kathy Baxter, Principal Architect of Ethical AI

• Srini, President and Chief Engineering Officer

• Julie Sweet, Chief Executive Officer of Accenture

Growth and Giving Back

• Salesforce has been in business for over 24 years.

• The company has given back over $621 million in grants to local non-profits and NGOs.

• Employees have volunteered over 8.1 million hours with non-profits and NGOs.

• Over 54,000 non-profits and NGOs run for free on Salesforce's service.

• 17,000 other companies have joined Salesforce in their 1/1/1 model.

Core Values

• Trust is critical when it comes to generative AI.

• Innovation is a key focus at Salesforce.

Einstein: AI for CRM

Srini walks through what they've been doing since they first introduced Einstein.

Overview

• Einstein does about a trillion trans predictions every week.

Benefits

-[](0:12:26 t:746 s): If you use our lead scoring app Einstein lead scoring you can convert your leads 3x faster just in Thanksgiving week if you're a Commerce.

Building a World-Class AI Team

In this section, Srini discusses how they built a world-class AI team and solved fundamental problems in predictive AI.

Investing in a World-Class AI Team

• Built a world-class AI team consisting of researchers, data scientists, and data engineers.

• Invited new techniques like Auto feature engineering, Auto feature selection, and Auto model selection to solve fundamental problems in predictive AI.

• Invested deeply with customers to learn new use cases.

Learning from LLM Technologies

• Learned that some LLM technologies are auto-regressive techniques that can be used for protein generations.

• Published many protein generation papers in Nature magazine.

Bringing Technologies to Customers

• Absorbed all the technology so that customers can get business wins.

• Excited about bringing generative AI to the future.

The Reality of Generative AI

In this section, Mark talks about the reality of generative AI and how it is not as simple as just putting corporate data into an LM.

Misconceptions About Generative AI

• CEOs think they can take all their corporate data and put it into an LM to have instant intelligent company.

• Public models vacuum up all available data off the internet which is not possible for regulated industries.

Trusting Generative AI

• Onus is on Salesforce to give customers next-generation generative Ai because trust is their number one value.

• Responsibility of LMs is to give you the best case with their generative AI, not to lie.

• Large banks want to use LMs to become more productive in mortgages and account service but it's not that simple.

Sharing Model

• Salesforce came up with a sharing model where every person in the company who gets Salesforce access also gets access around what data they can see or not see or use.

Introduction

In this section, the speaker introduces the concept of generative AI and explains how it differs from traditional AI models. The speaker also discusses the importance of data privacy and security in generative AI.

Generative AI and Data Privacy

• Generative AI models take all available data to generate intelligence.

• Cell-based security is not present in current generative models.

• Salesforce has developed a GPT trust layer that allows for generative AI without sacrificing data privacy or security.

Responsible Use of AI

• Responsible use of AI is critical due to societal implications.

• Salesforce has an ethics team that focuses on responsible use of AI.

• Kathy Baxter speaks about Salesforce's trusted AI principles, which include responsibility, accuracy, safety, transparency, empowerment, and sustainability.

Importance of Trust

• Every transaction and conversation at Salesforce begins and ends with trust.

• Industries such as banking, healthcare, and media require auditability of their data.

Deep Learning Principles

In this section, the speaker discusses deep learning principles and how they have expanded into generative AI.

Expansion of Neural Networks

• Generative AI expands upon deep learning principles by expanding neural networks.

• As the network expands, so does the ability to generate insights.

Salesforce's AI Strategy

In this section, the speaker discusses how Salesforce is already the number one company in customer relationship management with artificial intelligence. They introduce their product Data Cloud and explain why it is becoming an important cloud for customers.

Introduction to Data Cloud

• Data Cloud has become Salesforce's fastest-growing cloud ever.

• Creating a data cloud is essential for every customer preparing for generative AI.

• Customers may be creating data clouds outside of the Salesforce ecosystem, which is why they introduced Data Cloud as an intelligent, real-time, automated, and hyperscale product.

Importance of Data Cloud

• Salesforce is already delivering 30 trillion transactions per month with Data Cloud.

• Ford uses Data Cloud to provide the next-generation sales experience for their new component cars and trucks.

• The data cloud sets the stage at the beginning of every customer's AI journey.

Augmenting Customer Experience with Generative AI

• Salesforce aims to provide a customer 360 experience from sales to service to marketing to Commerce to Tableau to slack augmented by generative AI.

• Generative AI is changing everything and has seen huge growth. However, there are limits that need addressing when it comes to enterprise use cases.

• Every company needs an AI strategy now because there's a pretty big gap between productivity and trust when it comes to enterprise use cases.

Closing the Trust Gap

• There's a need for trust when it comes to privacy, hallucinations, data control, bias, toxicity in models used in enterprise AI.

• Salesforce aims to close the gap between productivity and trust with their first-generation trust layer.

Introduction to AI Cloud and Einstein GPT Trust Layer

In this section, Mark Benioff introduces the AI Cloud and explains how it is built for CRM. Patrick Stokes then discusses the problem of trusting generative AI and introduces the Einstein GPT Trust Layer.

Introducing the AI Cloud

• The AI Cloud is a trusted Enterprise AI built for CRM.

• It is built for Salesforce and its customers.

• It implements key technologies critical for using Einstein GPT trust layer.

Problem with Generative AI

• Large language models learn data instead of storing it.

• Access controls cannot be put on top of learned data like in databases.

• This creates a problem of trusting generative AI.

Solution: Einstein GPT Trust Layer

• The solution starts with a prompt, which provides detailed context and instructions.

• Grounding technique can be used to continue using prompts without training large language models.

• The Einstein GPT Trust Layer allows access controls to be put on top of learned data.

Understanding Generative AI

In this section, Patrick Stokes explains how generative AI works and why it's different from traditional databases.

How Traditional Databases Work

• Data is stored in databases with inherent location concepts.

• Access controls can be put on top of locations to control access to data.

How Generative AI Works

• Large language models learn data instead of storing it.

• Knowledge about something comes from identifying certain properties over time.

• Properties combined give knowledge about something but cannot control how recall comes out.

Using Prompts for Contextual Information

In this section, Patrick Stokes explains how prompts are more than just questions and can be used to provide detailed context and instructions.

What is a Prompt?

• A prompt is an entire canvas to provide detailed context and instructions.

• It can be used to ask for more information about a business or customer.

Grounding Technique

• Grounding technique can be used to continue using prompts without training large language models.

• Within the prompt, information about the business or customer can be included.

Importance of Contextual Information

• Generative AI without contextual information is unusable.

• Including contextual information in prompts allows generative AI to provide personalized and relevant responses.

Energy Investment Opportunity

In this section, the speaker talks about how they can improve their generation by including information that may or may not be true at any given time. They also discuss how they can protect sensitive data while using prompts to generate better results.

Improving Generation with Prompts

• The speaker explains that they can use prompts to generate immediately usable results without having to download and cut and paste data.

• They mention that they can add context from all of the customer data Salesforce has about a business, including sales, service, commerce, marketing, and telemetry data.

• The speaker acknowledges that there is a problem with using customer data in prompts because it contains sensitive information like personally identifiable information (PII).

• They explain that the Einstein GPT trust layer creates separation between corporate enterprise data stored in CRM databases and the large language model (LLM), allowing for responsible grounding of prompts in data without ever leaving Salesforce.

Protecting Sensitive Data with Einstein GPT Trust Layer

• The speaker describes several methods used by the Einstein GPT trust layer to protect sensitive data while still allowing for better generation results.

• They mention secure data retrieval, dynamic grounding, toxicity detection, zero retention, and data masking as some of these methods.

• The speaker demonstrates how they can use data masking to hide PII like Lauren Bailey's name from a prompt before generating results.

• They explain that after generating results with a prompt and adding context about a business, they can erase the prompt so none of the sensitive information is learned by the LLM.

Trusted AI Cloud Architecture

In this section, the speaker explains how their Trusted AI Cloud architecture works. They describe different layers of their architecture and how each layer contributes to ensuring trust in their system.

Layers of Trusted AI Cloud Architecture

• The speaker explains that the bottommost layer of their architecture is the trusted infrastructure layer, which provides data residency, compliance, security, and net zero.

• They mention that on top of this layer is the data cloud layer, which allows for lake house petabyte scale real-time data architecture built natively into the platform.

• The speaker describes open models as the next layer in their architecture. They explain that they will run a model tournament to give customers the best use cases and solve problems for them.

• They mention that customers can also bring their own models if they have big data science and AI teams.

Trust in Data

• The speaker emphasizes that trust is at the core of their architecture. They never look at customer data or share it with other customers.

• They explain that all learning happens within a customer's trusted boundary for their company.

• The speaker mentions that they abstract away complexity by optimizing models for security, compliance, and performance so customers don't have to figure it out themselves.

Dynamic Grounding and GPT Trust Layer

In this section, the speaker explains the importance of dynamic grounding in data masking and auditing. They also discuss how the GPT trust layer works in Salesforce's CRM apps.

Dynamic Grounding

• Dynamic grounding is required for toxicity detection or data masking.

• People want to know what all things happen, and they want to know what is their audit trail.

GPT Trust Layer

• The GPT trust layer provides an assistant for all applications.

• If you are a sales cloud user, you will get a sales cloud assistant that helps you close deals faster.

• If you are a service cloud agent, you will have an assistant that helps you close cases faster.

• Slack is going to be the UI for AI; it's that entire interface slack is going to wake up and allow users to access Enterprise knowledge.

• Prompt builders are available for Trailblazers who can build new generative AI apps using app builder tools.

• ISV partners can use this stack to implement and generate more value for customers.

Data Flows

• From the CRM apps, prompts are combined with company data in secure data retrieval.

• Dynamic grounding is done if any data masking is required before it goes through a secure gateway that allows communication with other models.

• Salesforce hosted models or external models with shared trust boundaries can be used without retaining any data in the llm (zero retention).

• Toxicity build filters bias filters are applied before returning results back to CRM app.

Investment in Salesforce LLMs

In this section, the speaker talks about Salesforce's investment in AI research and development.

• Salesforce has a world-class AI research team that has published more than 200 peer-reviewed journals and 200 patents in this area.

• Salesforce invests in SOTA models, open-source models, and partner models to provide the best model for specific tasks.

• Salesforce abstracts all the complexity of running models and runs a model tournament to pick the best and cheapest model for the job.

• The llm is protected with future preview, ensuring that customers are always up-to-date with the latest technology.

Formula One's Partnership with Salesforce

In this section, Mark Gallagher talks about the importance of having the customer at the center and how their partnership with Salesforce helps them design a unique experience for fans. They use AI, data, and CRM to create personalized experiences for fans.

Importance of Customer-Centric Approach

• Having the customer at the center is imperative.

• Only 1% of their fan base gets to attend a race.

• They want to engage the other 99% of fans.

Personalized Fan Experience

• Formula One has seen a shift in their fan base towards younger and more female fans.

• They use AI, data, and CRM to create digital experiences that resonate with individual fans.

• With personalized experiences, they can bring fans closer to the action than ever before.

Importance of Data

• Data is key to creating personalized experiences.

• Formula One collects data across all touchpoints.

• Salesforce helps visualize data faster using Data Cloud.

Innovation through Partnership

• Formula One sees themselves as a fan company and works with Salesforce to bring new experiences to life.

• As they grow their fan base and data, they find new ways to innovate.

Using AI Data and CRM for Customer Experiences

In this section, Sanjana Peralekar talks about how Formula One uses AI, data, and CRM seamlessly behind the scenes to provide amazing customer experiences. She explains how connecting all relevant data sources using Data Cloud makes it simple for personalization. Einstein GPT helps marketers quickly create landing pages that are personalized based on past performance.

Connecting Relevant Data Sources

• Connecting all relevant data sources is important for providing end personalization.

• Mulesoft can be used to connect any external or legacy system.

Harmonizing Data

• Harmonizing data into one consistent format is important for personalization.

• Data Cloud can be used to harmonize data from different systems.

Personalization with AI

• Einstein GPT helps marketers create personalized landing pages quickly.

• Content on the landing page is based on past performance.

• Interactive maps can be added for customers and fans attending events.

Developer Environment

• The developer environment takes over once the marketer's job ends.

The Role of Generative AI in the Future of Work

In this section, the speaker discusses how generative AI is changing the future of work and making roles more efficient. Formula One (F1) uses Apex and Einstein GPT to personalize their customer's experience.

Personalization with Generative AI

• Generative AI is changing the future of work by making roles more efficient.

• F1 uses Apex and Einstein GPT to personalize their customer's experience.

• F1 treats their end fans as Hospitality reps, building a trusted relationship with them.

• Einstein GPT helps scale this relationship by updating account descriptions and surfacing relevant contacts for upcoming events.

Using Outcomes-Based Email Generation with Einstein GPT

In this section, the speaker explains how Einstein GPT generates emails based on outcomes rather than past actions.

Outcomes-Based Email Generation

• Einstein GPT generates emails based on outcomes rather than past actions.

• Emails are generated based on what has been most highly performant with customers in the past.

Using Slack as a Centerpiece for Productivity Across Channels

In this section, the speaker discusses how F1 uses Slack as a centerpiece for productivity across channels.

Slack as a Centerpiece for Productivity

• F1 uses Slack as a centerpiece for productivity across channels.

• The updated sales homepage for all F1 Hospitality reps is enriched with all the amazing data they have in CRM about their customer.

• Einstein GPT helps personalize the slack canvas by identifying the best experiences at events.

Using Slack Huddles for Quick Questions

In this section, the speaker explains how F1 uses Slack huddles for quick questions.

Slack Huddles

• F1 uses Slack huddles for quick questions that don't require a whole phone call or conversation.

• After the call is done, a helpful summary is provided in the channel that can be used later.

Bringing Generative AI to Every Application

In this section, Clara Shi, CEO of Salesforce AI, talks about bringing generative AI to every application.

Bringing Generative AI to Every Application

• Salesforce is bringing generative AI to every application and platform in a trusted way.

• Generative AI is being shipped today and will change how we work.

Salesforce AI Cloud: The Power of Generative AI

In this section, the speaker talks about how companies without a data science team or machine learning engineers can take advantage of generative AI. They introduce Einstein GPT and its capabilities to write code, generate flows, sales emails, service responses, and marketing landing pages.

Introduction to Einstein GPT

• Companies without a data science team or machine learning engineers can take advantage of generative AI.

• Einstein GPT can write code for you and generate flows with flow GPT.

• It can also write sales emails, generate service responses, marketing landing pages and more.

• Einstein GPT is capable of crafting perfect emails by collecting all the context about that customer's interactions with marketing and service support issues.

Advantages of Einstein GPT

• With Einstein GPT, every seller in your organization, every service agent in your organization, every marketing manager in your organization can craft perfect emails instantly.

• Salesforce has launched Converse GPT and Marketing GPT at Connections last week. Slack GPT was launched last month at their New York World Tour. Tableau GPT was launched before that at their event in Las Vegas.

• Sales GPT and Service GPT are being added to the mix bringing generative AI into the flow of work where knowledge workers are already spending time.

Focus on Operational Challenges

• To succeed with Sales GPT we need context from Marketing and Service departments.

• Focusing on areas where Service agents struggle such as looking up product documentation for customers waiting to hear back from them on the phone or chat.

• In Sales, automating the process of piecing together a perfect email to get a prospect to meet with sellers so that they can focus on connecting and driving value.

• AI Cloud drives productivity and revenue for every workflow, user, department across every industry and segment around the world.

AAA's Perspective

• AAA is looking at generative AI in three different areas: customer service, support processing, and devops.

• They are leveraging generative AI while sticking to what's core to AAA with safety and trust in mind.

Partnering with Salesforce and Other Partners

In this section, the speaker talks about their approach to partnering with other companies and why it's important to find partners that can go shoulder-to-shoulder with them.

Importance of Finding the Right Partners

• Just about anybody is knocking down their door right now saying they have the best model or AI capability.

• They believe in partnering with companies like Salesforce and other partners to fold capabilities into the fabric of their platforms.

• They want to partner with companies that have the wherewithal to go shoulder-to-shoulder with them.

Reinforcement Learning in AI

In this section, the speaker discusses reinforcement learning in AI and how it continuously improves over time.

Two Aspects of Reinforcement Learning

• There are two aspects of reinforcement learning: reinforcement learning from human feedback and reinforcement learning based on objective business outcomes.

• Reinforcement learning from human feedback involves analyzing how workers use generated sales emails or service agent responses.

• Reinforcement learning based on objective business outcomes involves analyzing whether a sales email caused a sales opportunity to advance a stage or if a marketing landing page led to customer conversion.

• These two aspects ensure that every customer using AI Cloud will develop the best models for every use case specific to their industry, task, and department.

Customer Story: Rosenyol Using AI Cloud

In this section, the speaker shares a customer story about Rosenyol, a company based in France that uses AI cloud combined with data and CRM to personalize every interaction with their customers.

Personalizing Every Interaction

• Rosenyol wants to inspire people to spend more time in the mountains and create movements of sustainability and human potential.

• To foster loyalty without consumers, they need to personalize every interaction.

• Customer 360 is about understanding the desires and needs of their audiences so that they can deliver a better experience.

• With AI, Salesforce helps Rosenyol engage with their consumers seamlessly across seasons by analyzing data from all touchpoints.

Introduction

In this section, the speaker talks about how City Sports can help bikers around the world to adapt and strive in an ever-evolving world.

City Sports Can Help Bikers Adapt

• The speaker believes that City Sports can help bikers around the world to be agile when facing new challenges.

• The speaker introduces an incredible story about Sanchez.

Rossignol's AI Cloud Demo

In this section, the speaker demonstrates how Rossignol is using AI Cloud to deliver personalized recommendations and experiences for their customers.

Launching a New Product with Commerce GPT

• The demo starts with Commerce GPT helping launch a new product checklist for Rossignol.

• Commerce GPT helps generate product descriptions automatically, saving time and effort.

• Einstein GPT helps translate these descriptions into Spanish, which is relevant to their customer base.

Visualizing KPIs with Tableau Pulse

• Powerful visualizations are available for business users to keep track of their business as it happens using Tableau Pulse.

• Einstein GPT generates additive visualizations based on natural language questions asked by business users.

Improving Customer Service with Einstein GPT

• Einstein GPT provides helpful recommendations for service reps in real-time.

• Service managers can accept, edit or decline those recommendations while keeping a human in the loop.

• Conversations can be summarized and shared as knowledge articles amongst the whole team.

Getting Started with AI Cloud

In this section, the speaker talks about how to get started with AI Cloud and mentions some of their customers who are already using it.

Learning from Best-in-Class Companies

• The speaker mentions that they talked to Gucci, AAA, Russ and y'all, Royal Bank of Canada and other leading companies to learn how they use AI Cloud.

• The possibilities are endless when you bring AI data and CRM together.

Salesforce AI Summer Launch Keynote

In this section, the speaker talks about how Best in Class customers are using Slack for collaboration and how organizations are using Tableau to understand data. The speaker also emphasizes the importance of data and change management exercise.

Importance of Collaboration Tools

• Best in Class customers use Slack for collaboration.

• Teams come together to improve landing pages, commerce descriptions, etc.

• Organizations use Tableau to dig deeper into data used for training and fine-tuning AI models.

Importance of Data Management

• Data is important from both cloud and on-premise sources.

• Mulesoft is used to bring data together.

• Data Cloud is used to unify and harmonize data.

Change Management Exercise

• Change management exercise is crucial for successful implementation of AI technology.

• Salesforce Services helps with skill development and strategic planning.

• Partnerships with Accenture, Deloitte PWC, etc., help customers get started with AI technology.

Trailhead Courses

• Trailhead offers free online courses on over 35 AI badges ranging from basics to advanced courses on prompt grounding, tuning, etc.

Future of AI

In this section, the speaker talks about the potential of generative AI and LLM's (Language Model Models). The speaker also emphasizes the importance of trust and responsibility going forward.

Generative AI Capabilities

• Computers are more powerful than ever before due to generative AI.

• Sam believes that computers will rewrite applications to take full advantage of AI capabilities.

Exploring Next Generation AI

• Customers are exploring the next generation of AI to take productivity to a new level.

• Neural networks help customers understand what is possible with AI technology.

Conclusion

In this section, the speaker concludes by talking about how Salesforce is shipping various AI features and mobilizing its entire team around AI. The speaker also emphasizes the importance of reskilling employees for successful implementation of AI technology.

Shipping Various AI Features

• Salesforce is shipping various AI features such as sales email generation, service reply recommendations, summarization, etc.

Reskilling Employees

• Trailhead offers free online courses on over 35 AI badges for employees to reskill and learn new capabilities.

• Successful implementation of AI technology requires reskilling and learning new capabilities.

AI and Enterprise Technologies

Julie Sweet, CEO of Accenture, discusses the potential for AI in enterprise technologies. She emphasizes the importance of trust and security being built into these technologies from the beginning.

Potential for AI in Enterprise Technologies

• Trust and security are top of mind with all clients.

• Generative AI will transform how clients work and engage with customers.

• Clients should start by fully utilizing the AI they have already paid for.

• Using business criteria to evaluate use cases is important.

Accenture's Use of AI

• Accenture uses diagnostic and predictive AI across its business.

• Predictive capabilities were not being fully utilized until a sales officer made sure they were being used.

• Adding generative AI to existing processes can draw more connections and get insights faster.

Trust and Security in Generative AI

• Trust is a top concern when using generative AI systems.

• Companies need to consider the fundamental security of their data when using these systems.

Importance of Responsible AI

In this section, Julie Sweet emphasizes the importance of responsible AI and provides practical advice for companies to ensure they are using AI responsibly.

Responsible AI Checklist

• Companies should have a compliance program overseen by an audit committee.

• When leaving the room, one should be able to call someone in their company and ask where AI is being used, what the risks are, how they are being mitigated, how they are monitored, and who is accountable.

• It's important to understand every time you're using generative AI where the risk lies. Partnering with companies like Salesforce can help you understand if it's low risk or not.

• Make sure you're clear on what the risks are, how to mitigate them, and how to monitor them.

Conclusion

Julie Sweet stresses that responsible use of AI is crucial for companies. She advises companies to have a compliance program overseen by an audit committee and provides practical advice for ensuring responsible use of AI.