1. Is the Einstein 1 Platform's AI truly 'trusted'?
Salesforce is branding its relaunched platform around trusted AI. The Einstein 1 Platform is designed as a secure, governed environment for AI. But what does "trust" mean in practice? For developers, the hidden question is about data privacy, model transparency,
and handling hallucinations within the new Einstein Copilot Studio. How are customer data and prompts being used, and what controls do we have to prevent sensitive data from ending up in a training model? The trust layer is only as strong as its most granular permissions.
2. How 'real-time' are Data Cloud-triggered flows?
A major announcement is the ability to trigger Salesforce Flows directly from Data Cloud insights. This promises to close the gap between insight and action. The technical question is about latency. While marketing materials promise real-time automation, developers know there's always a delay. Is it milliseconds, seconds, or longer for complex, multi-system data changes? Understanding the true speed of this new capability is critical for designing responsive customer experiences.
3. Will Copilot Studio empower developers or abstract them away?
Einstein Copilot Studio gives admins and specialists tools like Prompt Builder and Skills Builder to create generative AI experiences. The question for pro-code developers is where they fit in. Will the studio be an extensible environment where developers can build complex, custom AI skills, or is it primarily a low-code tool that simplifies AI at the cost of deep customization? The answer will determine if this is a new playground for developers or a bypass around them.
4. What's the real story with 'Headless 360'?
Salesforce is pushing a "headless" architecture, allowing its platform's business logic to be called by external systems via APIs. This is a huge architectural shift. The hidden question is how this new model coexists with the traditional, UI-centric platform. Will headless capabilities get first-class treatment and investment, or will they lag behind the core CRM user interface? For architects, this determines whether a composable enterprise vision is truly viable on Salesforce.
5. Is Zero-Copy a complete data-sharing solution?
Data Cloud's Zero-Copy integration with platforms like Snowflake is a headline feature, promising to eliminate data duplication. The technical question is about performance and bidirectional data flow. While it's great for bringing external data into Salesforce for analysis, how efficiently can insights generated within Salesforce be pushed back to the source without creating new copies or complex ETL jobs? The 'last mile' of data activation is what truly matters.
6. What are the practical limits of 'Agentforce'?
The vision for an "Agentic Enterprise" involves AI agents autonomously performing complex tasks. This relies on the new Agentforce platform. The question is about the scope and reliability of these agents. Can they handle multi-step, long-running tasks that cross different systems? And what happens when an agent fails halfway through a process? The underlying governance and error-handling frameworks are more important than the flashy demos.
7. How do we govern costs in a consumption-based AI world?
With powerful generative AI features comes a new consumption-based pricing model, likely involving credits. The pressing question for IT leaders and architects is about cost governance. How do you monitor, control, and forecast AI spending when every user-generated prompt or agent action could be consuming credits? Without robust observability tools, the cost of innovation could spiral out of control.
8. What does the new Life Sciences Cloud mean for platform customization?
Salesforce announced a new Industry Cloud for Life Sciences. For developers working in that sector, the question is how prescriptive this new cloud will be. Will it provide a flexible, extensible foundation, or a rigid, locked-down solution that's difficult to customize? Industry clouds can sometimes accelerate development, but they can also constrain it if not designed with extensibility in mind.
9. Are free Data Cloud licenses a gateway or a gimmick?
Salesforce is offering free starter licenses for Data Cloud and Tableau to many customers. This lowers the barrier to entry, but the technical question is about the limits. The free tier includes up to 10,000 profiles, but many enterprises have millions. Is this enough to prove genuine value and build a business case, or will technical teams hit a hard ceiling just as they start making progress, forcing a costly upgrade?
10. Does Slack's deeper integration create new development silos?
Slack is becoming the conversational interface for the Agentic Enterprise, with record channels and native bots. The question for developers is whether this creates a parallel development track. Building rich experiences in Slack requires its own skillset (Block Kit, workflows) that is distinct from Salesforce's core platform tools like Lightning Web Components. Will this integration truly unify workflows or just create another silo to manage?
