Dreamforce 2024 has come to a close, leaving us with exciting announcements, demos and insights. Once again, it proved invaluable for anyone invested in the Salesforce ecosystem. Thanks to Salesforce+, I was able to watch many of the key sessions. Here’s my breakdown of the essential updates and must-know insights from this year’s event.
Note: Please bear in mind that many of the announcements from Salesforce are forward looking and subject to change.
Announcements & Updates
Artificial Intelligence
Agentforce: Humans + AI + Data + Action
"I think it's not just what is possible, it's what you are going to make possible because, as you're going to see, these agents, you can make your own. You can make them highly customised for you. You can make them do all these amazing things. But because only Salesforce does these three things: automate every single customer touch point, have the deeply integrated Data Cloud, have the complete Agent Builder, and deliver it as all one integrated platform, means that you don't need to DIY your AI."
You may remember Dreamforce 2023‘s mantra: ‘Data + AI + CRM + Trust’. This year, it’s evolved to ‘Humans + AI + Data + Action’; a significant shift.
Starting from October 25th 2024, Einstein Copilot will be upgraded to Agentforce. This marks Salesforce’s entry into autonomous agents. Unlike Copilot which requires human direction, Agentforce works independently. Agents can be triggered by events, such as conversations, data updates, automation, etc. They analyse data, formulate a plan and take action. Over the coming months, this will increasingly be powered by Agentforce’s new Atlas Reasoning Engine.
Agents are configured through the new Agent Builder. Each agent includes:
- Role. A natural language description of the topics and instructions the agent should follow.
- Topics: specific tasks the agent needs to perform.
- Instructions: guidelines of what the agent should do.
- Data. The data accessible to the agent (e.g. unstructured, structured, etc).
- Actions. What can the agent ‘do’. This can either be standard, out-the-box actions or custom actions created for your business’s specific needs (e.g. Apex, Flows, Prompts, etc).
- Guardrails. Clear definitions of what the agent should not do.
- Channel. What channels the agent can work with.
Prompt Builder
Salesforce also announced a number of enhancements to Prompt Builder. For example:
- Sensitivity-Based Data Masking. Mask data based on field classifications (e.g. Compliance Categories and Data Sensitivity Levels) in Object Manager. This works alongside pre-existing pattern-based masking; providing greater granularity and control.
- Prompt Performance Metrics. View aggregated performance metrics for Prompts (i.e. upvotes and downvotes) from Data Cloud in Prompt Builder.
- Record Snapshot. Pulls in data available from the running user’s page layout for the object to ground the prompt.
- Einstein Search Retriever. Easily configure a prompt’s search parameters and desired outputs from Data Cloud’s Data Model Objects (i.e. performing Retrieval Augmented Generation).
- 100+ Industry-Specific Prompts for Agentforce. Salesforce has released over 100 industry-specific prompts. These are customisable to suit the needs of your business. View the library here.
Free AI Training and Certification
With all these changes, Salesforce recognises it must empower the community to realise the potential of Agentforce and its wider AI solutions. This is not only to boost knowledge and confidence, but also help Salesforce professionals understand the art of the possible. A big announcement from Dreamforce is that Salesforce will be offering free AI training and exams. This applies to:
- AI Associate and AI Specialist certification exams from September 23 2024 to December 31st 2025.
- Retakes are NOT free. AI Associate will cost $50 USD and AI Specialist $100 USD + local taxes per retake.
- Virtual hands-on workshops from November.
- New 3-day instructor led course(s) will be available from late 2024.
- Instructor led courses will only be delivered in person (i.e. not virtual).
- They will be offered once per quarter in certain locations (e.g. San Francisco and London, etc).
Large Language Model (LLM) Benchmarking
A key decision for any business, whether new to AI or already using it, is picking the right LLM to power their solutions. After all, you would not want the engine of a tractor powering a sportscar. Each LLM has its strengths and weaknesses, making this choice critical.
Dreamforce showcased Salesforce’s freely available LLM Benchmark for CRM analysis. It is an interactive dashboard comparing various models for sales and service use cases across four key indicators:
- Accuracy. Measures factuality, completeness, conciseness of outputs and the quality of instruction following.
- Cost. Categorised as either high, medium or low, based on the use case.
- Speed. Time taken for a full response.
- Trust and Safety. Compares privacy, safety and truthfulness.
This helps businesses make informed choices. However, keep in mind choosing a model isn’t a one-off exercise. As new LLMs emerge, including Salesforce’s own offerings, today’s best fit could become obsolete.
Data Cloud
"Data Cloud is not just a simple data store. Data Cloud really is the backbone for everything we are going to be building on this platform for the next decade."
Dreamforce reiterated the critical role of data quality and availability in gaining the best value out of generative AI. This requires harnessing both structured and unstructured data safely. Salesforce is continuing to invest heavily in Data Cloud to help power Agentforce. Upcoming enhancements include new connectors, Data Cloud One, improved security and governance controls, hybrid search and sub-second real-time personalisation. Learn more here.
Salesforce Foundations
Starting September 17th 2024, Salesforce is offering a free upgrade for Sales and Service Cloud customers on Enterprise Edition and above. Salesforce Foundations combines Sales, Service, Marketing, Commerce and Data Cloud into one package. This uses Salesforce’s new Lightning User Interface (SLDS 2) (see True to the Core for more info).
Bear in mind:
- Agentforce Agents will be available later this year.
- Out-the-box Data Cloud entitlements are those provided at Dreamforce 2023.
- Email sends are limited to 2,000 per month.
- Some features (e.g. Marketing and Commerce) are subject to regional availability.
- Upgrading requires some setup. Refer to the Help Guide for more info.
To upgrade, contact your Salesforce Account Executive for more information.
True to the Core
True to the Core remains one of the most valuable Dreamforce sessions, allowing the community to directly speak with the Salesforce product team. Key themes this year included:
- Consumption based licensing. Salesforce is aiming to (in future) simplify its product licensing. This will likely lead towards consumption-based billing. For example, Agentforce prices starting at $2 per conversation.
- Generative Canvas for Lightning. Sales professionals can use generative AI to pull together and share client information with colleagues ahead of meetings. Learn more here.
- Lightning Page Loads. The first page load time is now 26% faster (benchmarked against March 2023). This is targeted to improve to 40% by Spring ’25.
- Lightning User Interface Updates. The new Lightning Design System (SLDS 2) offers a cleaner, and more accessible interface (WCAG 2.1AA compliant). It’s available for Starter & Pro Suite as well as new Sales Professional and Enterprise Edition orgs. A beta for existing orgs is targeted for Spring ’25, with general availability expected by Summer ’25. Learn more about SLDS 2, including steps to prepare here.
- List views. List views are now LWC-enabled. Winter ’25 introduces beta multi-column sort. Future updates include related list view multi-column sort, default list view sorting and related list view inline editing.
- LWC Local Dev. Developers can now see changes in real-time with LWC Local Dev.
- Roadmap Visibility. Increased visibility into roadmaps via the RoadmapExchange. View updates for specific products and join pilots. Roadmap Previews provides additional information.
- Salesforce Support. Salesforce is evaluating adding ‘Chat’ to Standard Support and, from the 11th of October, embedding Agentforce onto the Help and Training Portal.
Slack: The Conversational Interface for Work
Slack also featured heavily at this year’s Dreamforce, with two major updates:
- Salesforce Channels. Deeper integration between Slack Sales Elevate and Salesforce allows you to view and edit Salesforce records in Slack (available now). Slack Channels in Salesforce are targeted for February 2025.
- Agents in Slack. Slack integrates with Agentforce (October 2024) and partner Agents, turning it into an ‘AI hub’.
Additional Announcements and Updates
- Coming soon – native-low code integration for Flow. Targeted for Spring ’25, this includes connectors to 40+ systems (e.g. Jira, Zendesk & QuickBooks). Connections will allow you to trigger flows based on changes in external systems, easily map data and dynamically pull field definitions into Salesforce.
- Marketing Cloud and Commerce updates. New features and products such as Marketing Cloud Advanced Edition, Einstein Marketing Intelligence, enhancements for Unified Commerce and Point of Sale capabilities.
- AI Developer Tools. A number of new tools are on the horizon to boost productivity for Developers (e.g. Code Analyzer v5, DevOps Testing, Requirements to Code and Data Seeding). .
- Agentforce World Tour. The Agentforce World Tour will start in October 2024 and recap updates from Dreamforce. There will be demos and the opportunity to dive deeper into the new features.
- Reminder: Winter ’25. With updates to App Builder, User Management, Reports, Flow and more, now is a good time to get up to speed. Check out our Winter ’25 release summary for more info.
Artificial Intelligence: Updates from thought-leaders
"I think the progress that you're going to have with agents over the next year or two is going to be spectacular and surprising."
Dreamforce featured insightful AI sessions from top thought-leaders. Here’s a synthesis of my takeaways:
- AI Risks and Opportunities. AI has risks which should not be ignored. However, avoid pessimism or catastrophising. Focus on near-term challenges, be adaptable and seize the opportunities.
- Regulation. AI regulations vary globally. Assess, monitor and follow best practices to maximise the benefits of your AI.
- From Research to Product. Many businesses are starting to shift their efforts from proof of concepts to how best to operationalise generative AI. This shifts the onus from research to product-focused discussions. It must not be just a technical decision.
- Build or Buy. If your use case is core and unique to your business, consider building solutions. Otherwise, partnering with an existing solution may offer a better ROI. Choose carefully.
- AI Agents. Training autonomous AI agents is not a purely technical task. Think of training an autonomous agent as similar to training a new employee. Define the role and engage the relevant team to identify best practices and guardrails to help shape the agent’s behaviour.
- Differentiation. Effectively applying AI to business workflows can help drive differentiation. However, do not forget that effectively managing and applying data can also be a source of differentiation.
Summary
Dreamforce 2024 provided many insights. The main focus was Agentforce, which if its potential is realised, could significantly help Salesforce customers. Salesforce’s offer of free AI training and certification will also help drive this forward.
Enhancements across the platform, particularly for SLDS 2, Flow, Slack, Data Cloud and Prompt Builder further empower Salesforce teams to tailor solutions. Moreover, Salesforce Foundations is an important announcement and where appropriate, may help businesses work towards a truer end-to-end journey for their customers. This is being driven forward by Salesforce’s ongoing commitment to build solutions onto core, which will likely yield further innovations in time.
The AI landscape remains varied, with challenges and opportunities. As businesses increasingly look to rollout or hone their AI offerings, it will be important to engage a wide array of stakeholders. Not only will this allay concerns and increase adoption, but will be essential in ensuring suitable guardrails are built in the world of autonomous agents. Throughout, maintaining a viewpoint of cautious optimism is prudent.
It will be interesting to see how things progress by Dreamforce 2025!
Until next time, thanks for reading! If you found this useful, don’t forget to like & follow SFDC Penguin on LinkedIn or Twitter/X.
Bonus Penguin Fact 🐧
An “absolute unit”, “chonky” and “linebacker”. These are some of the affectionate nicknames given to a nine-month old king penguin chick at Melbourne’s Sea Life Aquarium. Weighing in at an impressive 23kg, Pesto eats around 25 to 30 fish a day; a pretty healthy appetite!
Pesto’s size is in part due to his downy brown feathers but also due to his genetics. King penguin chicks tend to lose their downy feathers when they experience their first molting at around 10-12 months old. At which point, Pesto will start growing his sleeker, waterproof feathers and soon after, begin his swimming lessons!