This month, I had the opportunity to join thousands of developers, administrators, partners, and ecosystem enthusiasts at the Moscone West convention center in the heart of San Francisco for the Salesforce TDX conference. Over the span of two days, we immersed ourselves in a whirlwind of networking, events, and hands-on sessions to learn all about what’s new and what’s coming in the Salesforce world. It will probably not come as a surprise to anyone in the industry right now, but generative AI took center stage and found its way into nearly all of the sessions.

Salesforce has brought together all of their generative AI offerings into the Einstein 1 platform and there were many opportunities for TDX attendees to get hands-on with some of the new features (and even set a new Guinness World Record to commemorate the occasion).

Each of the sessions were led by knowledgeable members of the Salesforce architecture team. I was impressed with how simple all of the new features were to configure from an administrative perspective. Many of Salesforce’s generative AI products are generally available now or will be in the very near future, although I do think they will take another few cycles of refinement and updates to features and integration capabilities to get close to their true potential.

The value proposition for generative AI for business is definitely compelling, but as I reflect on the keynotes and the sessions, I see three main hurdles that will affect companies’ ability to really get excited about it (some of which Salesforce is addressing better than others):

1. Security – The primary hurdle with companies allowing employees to use models like ChatGPT is concern related to safekeeping of sensitive company and customer data. Salesforce recognizes the importance of this concern and has addressed this with the Einstein Trust Layer, which TDX attendees were also given a deeper dive into. The ability for Salesforce to automatically mask sensitive data and their zero-retention agreement with OpenAI (and others in the near future) are hugely important building blocks for any AI related features.
   
   
2. Accuracy – Large Language Models (LLMs) such as ChatGPT are incredibly useful, but, as anyone that has used them much knows, they do not always provide accurate information. Using prompt engineering and LLM training/grounding, companies will need to learn how to refine the inputs into these powerful tools (and the tools themselves) to get outputs that are as accurate and relevant to their business processes as possible. Salesforce has introduced tools to allow admins the ability to refine the inputs so that all of their users get a consistent experience from the AI features, regardless of the individual users abilities when it comes to using LLMs. I think these tools need more features and capabilities to really get there, but what I saw during TDX was a great start.
   
   
3. Cost – These products certainly have the potential to allow Salesforce customers to reduce staffing and/or increase the efficiency of staff, but the licensing, usage based costs (tokens), and implementation costs will be significant and will need to be committed up front. These costs will also likely exclude some small and midsize businesses (SMBs) out of the gate, regardless of their level of enthusiasm. I think Salesforce is still figuring out the best way to package and market these features and I hope that they are able to make at least some of them accessible to all users, similarly to what we have seen from other tech leaders like Google and Microsoft. I also spoke to several ISVs (AppExchange partners) at TDX who are already leveraging generative AI, so they will likely fill some of those gaps for SMBs. This all makes for a complex evaluation process, and this is where strong internal Salesforce teams and Salesforce partners (such as EpiGrowth) will be critical to developing and planning implementations with realistic expectations and good outcomes.

These are just a few of my AI-related takeaways from TDX 2024. Feel free to reach out directly if you want to continue the conversation or if you’re curious about how Salesforce’s new generative AI features might fit with your organization’s needs. Figuring out where and how much to invest in these exciting new technologies can be challenging and we’re always happy to share our knowledge and honest opinions on anything Salesforce related.