Inside TDX 2025: Agentforce Evolution and the Community Adoption Challenge

TDX 2025 recently concluded at Moscone Center in San Francisco, where Salesforce unveiled significant enhancements to its Agentforce AI platform amid cool, cloudy late winter days. While the technical advancements were substantial, the event revealed a widening and increasingly troubling disconnect between Salesforce’s ambitious AI strategy and its traditional developer and admin community’s readiness to embrace this vision.

Executive Vision vs. Community Reality: The Adoption Gap

TDX attendance appeared noticeably lighter than previous years, with a subtly different atmosphere permeating the event. Patrick Stokes, Salesforce EVP for Product and Industries Marketing, kicked off the keynote by introducing the concept of “Agentblazers,” describing them as “pioneers who are building a world where humans and digital agents collaborate.” He emphasized AI’s role in addressing workforce challenges caused by slowing global population growth—a message that seemed disconnected from the immediate concerns of developers and administrators in attendance.

I caught myself wondering whether a diminishing workforce is a CEO or policymaker concern, not something the developers and admins in the room could relate to.

The event’s reduced emphasis on community traditions—no high-profile “Golden Hoodie” presentation and fewer iconic Salesforce characters and decorative elements—contributed to what several attendees described as a “less magical” experience. The recent dismantling of Salesforce’s Well Architected program and team appeared to dramatically compound community concerns, with some attendees questioning the company’s commitment to foundational platform education amid its AI push.

Informal feedback gathered throughout the event revealed the sentiment gap:

• “Agentforce is impressive, but I’m not sure if it directly impacts my day-to-day work yet.”
• “It feels more tech-driven than customer-driven, unlike earlier Salesforce innovations.”
• “Agentforce hooks up to your existing Flows and actions, but they aren’t teaching us enough about the basics!”
• “I wish there were better forums for community members to present their own solutions.”

Agentforce 2dx: Technical Advancements in Detail

Despite these tensions, Salesforce delivered substantial technical advancements through Agentforce 2dx:

Developer Tools and Infrastructure

Adam Evans, Salesforce EVP & GM for Agentforce, unveiled the new Developer Edition with Data Cloud integration, describing it as a “free, dedicated environment featuring 10GB of Data Cloud storage and up to 150 large language model generations per hour,” designed specifically to reduce barriers to AI experimentation. This announcement generated the most enthusiastic response during the keynote, and addresses concerns we had about developer access back at Dreamforce.

Alice Steinglass, Salesforce EVP & GM for Platform, showed significant improvements to developer productivity tools:

• Enhanced Agentforce CLI capabilities enabling developers to “directly script complex agent-driven operations into their DevOps workflows”
• Advanced VS Code support with real-time Lightning Web Component updates (hot module reload) and improved debugging tools
• Scratch Org snapshots delivering 600%-2400% performance improvements

These improvements address longstanding developer pain points, potentially offering more immediate value than some of the more conceptual AI capabilities.

Integration and Connectivity Ecosystem

Salesforce emphasized its “deeply unified platform” vision through enhanced integrations across its product portfolio:

• MuleSoft Integration: Providing the foundation for agents to autonomously engage in workflows and backend system interactions. As demonstrated during the keynote, this integration “dramatically expands agent capabilities by bridging front-end communication with backend processes.”

• Slack Integration: Facilitating real-time collaborative interactions between human users and digital agents, though executives acknowledged “sticky authentication and other issues remaining to be resolved to have seamless Slack-Salesforce interactions.”
• Tableau Integration: Empowering businesses with advanced analytics embedded directly within Agentforce, transforming decision-making through real-time, actionable insights.
• Data Cloud Integration: Harmonizing structured and unstructured data to “enrich Agentforce’s capabilities, providing agents with comprehensive data context to enhance decision-making accuracy.”

Marketplace and Partner Ecosystem

Christophe Coenraets, Salesforce SVP for Developer & Admin relations, introduced the AgentExchange Marketplace, describing it as “a structured, metadata-driven marketplace enabling developers to quickly integrate pre-built agent actions, such as those from DocuSign, into their workflows.”

The exhibition floor featured a robust presence from Salesforce DevOps vendors, including ACCELQ, Copado, Elements.cloud, Gearset, Hubbl Technologies, Opsera, Pharos AI, Prodly, Provar, Salto, and Sweep.io. Some vendors took alternative approaches, with Metazoa and SRE.ai sponsoring offsite events rather than traditional booths.

Monitoring and Governance

Steinglass revealed improvements to observability dashboards powered by Salesforce Event Monitoring, emphasizing the enhanced ability to “track agent performance and interactions more effectively.” The “Agentforce Interaction Explorer” and comprehensive Testing Center were highlighted for providing deeper visibility and control over agent behaviors.

Elements.cloud: Process Configuration Mining

The most popular demonstration came from Elements.cloud, which unveiled its new “Process Configuration Mining” feature. This system scans organizational metadata to automatically construct process maps that reveal unintended configuration consequences.

Ian Gotts, CEO of Elements.cloud, explained: “What the team has achieved is remarkable, but only possible because of the 8 years of development on the Elements.cloud platform to be able to sync and analyze the customer org’s metadata, dependencies and usage. It provides the insights that every customer has been asking for since we launched Elements.cloud.”

Gotts also shared a case study where their automated solution improved upon analysis produced by a large consultancy that had required “thousands of hours of billable work,” demonstrating AI’s potential to commoditize and democratize complex organizational analysis.

Robotforce: AI Meets Physical Systems

Charlie Isaacs, Salesforce’s CTO for Customer Connection, provided insights into Salesforce’s integration of Agentforce with physical robotics systems. He detailed a pioneering implementation where “most robots run on Python, but ours is running on Ubuntu with C++ pulling from the Agentforce API through MuleSoft,” enabling “agents communicating with agents”—a capability he claimed was previously unprecedented.

“We’re moving towards having supervisory agents that oversee operations, creating new knowledge from interactions and adding it back into the network. That’s exactly where we’re going,” Isaacs explained, echoing CEO Marc Benioff’s recent references to “Robotforce” as the company’s vision for sophisticated agent orchestration.

Isaacs demonstrated these capabilities through a scenario involving a malfunctioning windmill at the event booth: “The human creates the work order manually, but through proximity sensors and RFID, Agentforce instantly communicates a fix directly to the IoT device. We’re seeing the beginnings of exponential learning here.”

He acknowledged implementation challenges, sharing how robots had to be rapidly adapted for demonstrations: “We consolidated functionalities from multiple robots on the spot,” highlighting the critical role of Salesforce’s architecture in enabling instant scalability.

Admin-Focused Innovations: Agentforce for Setup

Salesforce showcased “Agentforce for Setup,” a conversational AI designed to enhance administrator productivity within the Setup environment. Key capabilities included:

• Natural language assistance for common tasks (“How do I freeze a user?” or “How do I deactivate a user?”)
• Automated creation of custom report types
• Clear explanations of connected apps and user permissions

Future enhancements will include formula generation from natural language prompts, automated object and field creation, and intelligent navigation help. These practical tools received positive feedback for their immediate utility in solving daily administrator challenges.

Developer Experience: The Technical Reality Check

A conversation with Dan Fernandez and Khushwant Singh from Salesforce revealed both progress and persistent challenges in the developer experience. According to Salesforce, approximately 20-25% of new Salesforce development is already AI-assisted using Agentforce for Developers, demonstrating real-world adoption with tangible benefits. However, community feedback gathered from LinkedIn paints a more critical picture of fundamental platform challenges that many developers believe should be addressed before further AI advancements.

Nicolas Vuillamy, CTO at Cloudity France and creator of MegaLinter.io and sfdx-hardis, garnered dozens of reactions when he outlined four critical DevOps priorities that should precede Agentforce development:

1. Deployment fundamentals: “Maybe let’s start with allowing to deploy correctly the big daily pain points that can be sharing rules, duplication rules, roles, profiles…” Vuillamy emphasized the need for “meaningful deployment errors in all cases” rather than generic “unknown error” or “incorrect cross reference id” messages that lack specific information.
2. Performance limitations: Vuillamy highlighted “the dramatically awful performances of Apex tests during deployment, forcing us to invent awful workarounds to not run all of them, injecting tech debt in our release management pipelines.” He suggested parallel test execution as a potential solution.
3. Developer standards: He called for Salesforce to “resuscitate well architected team and extend it to Apex guidelines more precise than ‘don’t code in triggers’” and recommended implementing “default class naming norms and folder hierarchy.”
4. AI prioritization: Only after addressing these foundational issues should Salesforce “talk about Agentforce,” according to Vuillamy.

Additional Concerns

Added developer concerns raised in the discussion thread included:

• Christopher Hickman, Enterprise Solution Architect at DoorDash, asked, “When is Revenue Cloud going to be deployable?” and questioned “why doesn’t Salesforce make it easier to treat configuration data as metadata (without relying on third-party solutions)?”
• Matt Pieper, Director of Engineering at LeafLink, emphasized the need for “TESTS FOR FLOW,” stating, “I don’t understand how we’re pushing agents being built without any testing framework behind the automation that drive them.”
• Teja Amerineni, co-founder of Salesforce partner CloudFulcrum, called for addressing “basic issues on the platform level” including easier profile management in source format, better support for unlocked packages, and improved Omni Studio build processes.
• Alan McQueen, Director of Business Systems at Fike Corporation, requested enabling “source tracking across all sandboxes” and removing timeouts “for large orgs with many metadata elements.”

Chris Pearson, Salesforce Engineering Director at Jostens, raised concerns about Agentforce’s practical implementation, noting “It takes too long to build an agent from scratch” and questioning “Why can’t Agents have built-in extensions to connect to other commercial apps and perform actions vs. having to get Mulesoft and/or Data Cloud involved?”

Salesforce acknowledged these issues during TDX sessions, emphasizing their ongoing focus on Apex recompilation efficiency, proactive AI-driven fixes, and more transparent observability. However, the prevalence and intensity of these DX concerns raises significant questions about whether Salesforce’s community engagement and platform fundamentals are receiving sufficient attention alongside its ambitious AI strategy.

Evaluating TDX Through the VE Economics Lens

Salesforce’s Agentforce advancements at TDX 2025 strongly reflect the principles of Virtual Employee (VE) Economics—an analytical framework examining how AI-driven autonomous agents impact organizational structures, costs, and labor markets. Analyzing the event through this framework reveals both Salesforce’s strategic vision and potential community adoption challenges.

The Agentforce 2dx platform enhancements directly embody the three core VE Economics principles:

1. Law of Infinite Scale: The new Developer Edition with Data Cloud integration enables organizations to deploy AI agents at minimal marginal costs. As Evans emphasized in his keynote, this democratized access aims to remove “barriers to AI experimentation,” potentially allowing businesses to scale cognitive capabilities without proportionate resource increases. This marks a significant shift from traditional labor-intensive CRM customization to scalable AI-driven solutions.
2. Law of Cognitive Commoditization: The AgentExchange Marketplace represents a clear implementation of cognitive commoditization, converting specialized knowledge work into reusable, low-cost assets. Pre-built agent actions from partners like DocuSign and Copado reduce previously high-cost cognitive processes to accessible platform capabilities, fundamentally altering traditional professional service models. The Process Configuration Mining case study cited by Elements.cloud is additional evidence of Cognitive Commoditization.
3. Law of Exponential Learning: The enhanced MuleSoft and Slack integrations demonstrated by Coenraets showcase Salesforce’s commitment to network-driven knowledge propagation. As agent interactions generate insights across connected systems, these improvements “dramatically accelerate performance enhancements,” as stated in Isaacs’ robotics demonstration. This approach contrasts sharply with traditional organizational learning models, where knowledge transfer faces significant friction.

Too Much Too Soon?

However, evidence from community feedback suggests a significant gap between this VE Economics-driven vision and Trailblazer community priorities. While the framework elegantly explains Salesforce’s strategic roadmap, the traditional customer base appears more concerned with fundamental platform reliability and developer experience issues than autonomous agent capabilities.

This tension reveals broader implications for enterprise AI adoption: advancing theoretical capabilities may outpace organizational readiness and practical implementation capacity. As VE Economics predicts, organizations capable of bridging this gap stand to capture disproportionate competitive advantages as AI integration intensifies across industries. This may foreshadow potential conflicts between CEOs and executives hot to implement AI, while their workforce may be resistant to change.

Competitive Landscape Considerations

Salesforce Agentforce leads the industry in embedding AI capabilities within a tightly organized business database and security system. However, the company’s comprehensive, enterprise-grade approach faces potential competitive challenges from nimbler, purpose-built AI solutions. It’s a natural impediment because Salesforce needs to satisfy existing customers and minimize disruption can stifle innovation.

This makes Salesforce vulnerable to competition from overlaid AI solutions. As noted by community member Andrew Russo, Salesforce Architect at BACA Systems, alternatives such as Intercom’s AI agent Fin lets users bypass the complexity and security rigor of Salesforce for some applications, such as a company FAQ chatbot.

This tension—between comprehensive enterprise capability and streamlined, purpose-built AI tools—represents a strategic challenge as Salesforce advances its AI vision.

Bridging the Adoption Gap: Critical Challenges

For Agentforce to achieve broad adoption, Salesforce must address several key challenges found during TDX 2025:

• Observability Maturity: While Salesforce has taken substantial steps forward, comprehensive agent observability and debugging tools are still in early stages compared to mature DevOps ecosystems.
• Core Platform Education: The heavy focus on AI appears to have reduced attention to community engagement and operational needs, with some attendees noting insufficient education on platform fundamentals.
• Community Education Gap: A sizable portion of Trailblazers appear unsure about how to effectively leverage Agentforce in their daily work, requiring more accessible educational resources.

Looking Forward

TDX 2025 reinforced Salesforce’s technical leadership in enterprise AI while highlighting the increasingly troubling challenges of aligning innovation with community readiness. The company’s success will depend on effectively bridging this gap—demonstrating how Agentforce can enhance rather than replace the human-centered values that have traditionally defined the Salesforce ecosystem.

Patrick Stokes, addressing some of these concerns during his keynote, emphasized that Agentforce aims to augment human capabilities rather than replace them. However, translating this message into practical, immediate benefits for traditional Trailblazers remains a critical challenge that appears to be worsening rather than improving.

As Salesforce executives remain open to criticism, with some community advocates feeling their voices are heard, the company has an opportunity to address these concerns directly. Increasing the rhythm of community education and more clearly communicating the practical, incremental benefits of Agentforce for traditional use cases may help bridge the current adoption gap.

TDX 2025 demonstrated both Salesforce’s ambitious technical vision and the increasingly serious real-world challenges of bringing its community along on this transformative journey. The months ahead will reveal whether the company can successfully navigate this transition, repairing a strained community foundation while advancing its AI-driven future.

“This isn’t just about technology choices,” concluded one long-time Salesforce consultant. “It’s about whether Salesforce remembers that its greatest strength has always been its community. The AI future they envision won’t materialize if they leave that community behind.”

• TrailblazerDX 2024 for Salesforce Devops
• The Transformative Power of AI-Driven Salesforce Administration
• Dreamforce 2024 DevOps Roundup
• Salesforce Releases Agentforce: Autonomous AI Moves Beyond Chatbots
• Agentforce Pushes Salesforce Forward at Dreamforce 2024