Participants: Eric Shupps, Paul McCollum, Ian Gotts, Jade Leong, Nadina Lisbon
Main Themes:
AI Adoption & Hosting: The discussion focused on the challenges and implications of AI adoption, specifically concerning public cloud versus on-premise hosting. The consensus leans towards public cloud due to the significant computational power required for large language models. Concerns were raised regarding data security and compliance for specific industries like finance and government, potentially driving the need for hybrid or "cloud-gapped" solutions.
AI Capabilities & Competitive Advantage: The group discussed the current state of AI capabilities, highlighting the dominance of text-to-text models like those offered by Salesforce. While acknowledging the potential of other AI types like text-to-video or computer vision, the consensus suggested that competitive advantage currently lies in the effective utilization and fine-tuning of existing models rather than developing proprietary ones.
Salesforce's "Agent Force" and its Impact: The conversation delved into Salesforce's newly announced "Agent Force" and its reliance on a consumption-based pricing model. This sparked debate on cost predictability, potential challenges for implementation teams, and the need for sophisticated architecture to manage agent interactions and prevent runaway costs. Concerns were raised about potential customer pushback and the risk of Salesforce repeating the mistakes of Microsoft's consumption-based pricing model.
Governance & Technical Debt: A key concern highlighted throughout the discussion was the lack of governance and technical debt prevalent in many Salesforce implementations. Participants agreed that AI, particularly Agent Force and Data Cloud, would exacerbate these issues, potentially leading to project failures, data breaches, and unforeseen costs. The group emphasized the need for a fundamental shift in mindset, moving away from the "anyone can do this" approach to one that prioritizes rigorous data governance, implementation practices, and specialized AI architectural expertise.
Key Ideas/Facts:
AI-specific workload compute clusters: A prediction that cloud providers will offer specialized AI-focused infrastructure to cater to the increasing demand for computational power.
Cost unpredictability of consumption-based pricing: A significant concern raised across various cloud providers, leading to project delays and customer dissatisfaction.
Salesforce Agent Force pricing: Initial pricing for Agent Force is $2 per conversation, raising concerns about cost control and potential for abuse.
Data Cloud implementation challenges: Successful Data Cloud implementation requires a rigorous approach and highlights existing data governance issues within organizations.
The need for AI architecture expertise: The increasing complexity of AI implementation demands a new breed of architects specializing in AI governance, cost management, and agent interaction architecture.
Notable Quotes:
"The size of your models is limited by your horizontal computational power." - Eric
"I think the pure computational power to host and expand these models is going to preclude 95% of organizations from being able to host it themselves." - Eric
"The differentiator isn't the model. The differentiator is your ability to use it." - Ian
"I would hope that Salesforce will study Microsoft's mistakes and not go too far down that route." - Eric
"Consumption requires success." - Ian
"Data cloud and AI will quickly amplify and expose problems with bad data and poor data governance." - Paul
"The ease at which it and speed at which it can retrieve information exposes how bad your practices were in storing that information in the first place." - Eric
"Infinite computation gives you the ability to immediately shoot yourself in the foot with a cannon as before when you were just doing it with a BB gun." - Paul
"The people who got us to where we are now are not the people who are going to get us to where we need to get to." - Anonymous Salesforce executive quoted by Ian
Action Items:
Further research into cost management strategies and best practices for consumption-based AI services.
Develop frameworks and methodologies for AI governance, implementation, and agent interaction architecture.
Investigate emerging technologies and ISV solutions aimed at addressing AI cost and risk management.
Engage with Salesforce and other cloud providers to better understand their roadmap for addressing AI governance and cost predictability challenges.
Conclusion:
The discussion highlighted the excitement and potential of AI, particularly within the Salesforce ecosystem, while simultaneously exposing critical concerns regarding cost, governance, and implementation complexity. Addressing these challenges through proactive architectural planning, robust governance frameworks, and specialized expertise will be crucial for successful and sustainable AI adoption.
