This post was originally published on this site.
Salesforce, with its existing Einstein predictive AI services and the emerging Agentforce digital worker platform, is among the world’s largest producers of enterprise AI services. Over the last decade, the company has worked to integrate a thoughtful and innovative AI safety and ethics program. Now, with talk of digital workers performing role-specific functions like a customer service representative (CSR) or sales development representative (SDR), there are worries about the impact of systems like Agentforce on jobs, especially those assistant or highly systematic roles. As Agentforce moves into headcount budgets, is Salesforce responsible for the loss of entry-level jobs that may result?
Salesforce has taken a proactive approach to these challenges under the leadership of Paula Goldman, its Chief Ethical and Humane Use Officer. During a recent conversation with Goldman, I gained insights into how her office operationalizes AI ethics and fosters practical alignment for Salesforce enterprise customers and society at large.
Table of contents
Building the Foundation of Ethical AI at Salesforce
“When my role was created over six years ago, there wasn’t a real precedent for it, especially in enterprise software,” Goldman remarked during our discussion. “Salesforce’s leadership recognized early on that as technology grows more powerful, it’s critical to think ahead about its consequences.”
Goldman’s team is charged with an ambitious mission: to develop policies, frameworks, and systems that not only ensure the ethical use of AI but also anticipate its societal impacts. This vision was set into motion by Salesforce’s early decision to embed trust as a core corporate value. Goldman credits this foresight with enabling her office to address the ethical questions that arise as AI capabilities evolve.
Operationalizing Ethics in AI
A standout feature of Goldman’s work is her team’s integration into Salesforce’s product development process. She explained, “We’re deeply involved in product roadmaps, ensuring that ethical principles are built into the systems from the start.”
One of her office’s key initiatives is adversarial testing. “We take products that have a higher potential impact on people’s lives and try to break them before they ship,” she said. “This helps us identify and fix issues, ensuring the systems are reliable and cannot be exploited in harmful ways.”
Her team also oversees the development of “standard patterns for human-AI interaction,” addressing critical questions like: “How do you ensure users know they’re interacting with AI? How do you display citations transparently? How do you build trust without overwhelming users?”
Customer-Centric AI Alignment
Goldman emphasized that customer demand drives much of her office’s work. “Our customers want to know their data is used responsibly, that they can rely on AI outputs, and that systems are monitored for toxicity,” she noted. “In the enterprise space, these concerns are far from theoretical; they’re practical demands.”
Salesforce’s Trust Layer—a set of security, privacy, and ethical guardrails—is a direct response to these demands. “We’ve worked on features like toxicity detection and runtime guardrails to help customers monitor and adjust their AI systems as needed,” Goldman said.
Navigating the Workforce Impacts of AI
As AI reshapes the workplace, Goldman’s team is addressing the human side of these changes. “A big part of our focus is on trust and human-AI interaction,” she explained. “It’s essential that users have the right amount of trust—not too little and not too much—so they can interpret AI results effectively.”
To this end, Salesforce engages employees in “trust testing,” a participatory process where diverse teams attempt to break products to surface potential biases or usability issues. “We partner with our business resource groups to ensure a variety of perspectives are represented,” Goldman said. “This feedback helps us create systems that work for everyone.”
Goldman also highlighted Salesforce’s broader workforce initiatives, such as AI Learning Days and AI accelerators for nonprofits. “These programs aim to upskill employees and prepare them for an AI-driven world,” she said.
Challenges and Opportunities in AI Alignment
One challenge Goldman identified is the relatively nascent partner ecosystem for observability in AI. “Observability is critical for ensuring systems remain trustworthy,” she said. “There’s an opportunity for vendors to step in and build better logging and monitoring tools, which would benefit everyone.”
Goldman also addressed the complexities of human-AI teaming. “We’re exploring how service agents hand off tasks to AI, and vice versa,” she said. “As these interactions become more seamless, they’ll redefine roles and workflows, creating both efficiencies and new challenges.”
Salesforce’s Role in Shaping Industry Standards
In addition to its internal initiatives, Salesforce is active in shaping public policy and industry standards. Goldman serves on the National AI Advisory Committee and collaborates with organizations like the NIST AI Safety Institute. “It’s important to focus not just on AI models but on the broader systems and their societal impacts,” she said.
Goldman’s work reflects a commitment to making AI alignment practical and actionable. “The ethical landscape is constantly evolving,” she said. “Our job is to stay ahead of these changes and ensure our customers can trust the technology they’re using.”
Public Fear and the AI “Manhattan Project”
While Salesforce’s ability to operationalize ethics in its operations, its recent moves into the digital workforce raise greater questions. The rollback of AI governance efforts by the Trump administration, combined with the breakneck pace of innovation, has heightened public fear around the unchecked power of artificial intelligence. Figures like Marc Benioff, Sam Altman, and Dario Amodei may be seen as modern-day Oppenheimers—brilliant minds at the forefront of technological revolutions. Yet, unlike the Manhattan Project, where decisions about nuclear weapons ultimately rested with President Truman, AI operates in a decentralized, fragmented ecosystem. With no singular authority mediating between developers and corporate leaders deploying AI, questions about responsibility and accountability loom large.
For Salesforce, whose Agentforce platform aims to revolutionize enterprise roles with digital workers, this raises critical ethical questions. If AI-driven systems displace entry-level jobs—such as customer service representatives or sales development representatives—should Salesforce bear responsibility for those societal impacts?
During our conversation, Goodman acknowledged the need for transparency and accountability, stating that Salesforce’s customers demand systems they can trust. However, trust in a tool does not erase its broader consequences. This recalls the ethical dilemmas faced by Oppenheimer after Hiroshima: is the creator of a powerful technology responsible for its ultimate use, or do the consequences rest with those who wield it?
Salesforce has taken proactive steps to address such concerns, embedding ethical guardrails like the Trust Layer and adversarial testing to mitigate harm before products ship. But these measures focus on the technology itself—ensuring that it works responsibly and transparently—not necessarily on its macroeconomic or societal ripple effects. Much like the generals and politicians who decided how to use nuclear weapons, the corporate CEOs purchasing and deploying Agentforce bear significant responsibility for the consequences. Yet without a broader regulatory framework, the question of ultimate accountability remains unresolved, creating fertile ground for public fear and mistrust.
The AI revolution, much like the Manhattan Project, forces society to grapple with uncomfortable truths. AI is a powerful force for change, capable of immense good but also potential harm. As digital workers like Agentforce increasingly impact headcount budgets, there is an urgent need to establish who is accountable for these transformations. Are companies like Salesforce responsible for the job loss that may result? Or does the responsibility lie with the leaders who deploy these systems and fail to address their human cost? These questions, much like those surrounding Oppenheimer’s legacy, resist easy answers—but they must be asked if we are to navigate this new technological frontier with clarity and purpose.
The Weight of Innovation
Salesforce’s approach to AI ethics, led by Paula Goldman, exemplifies a proactive effort to align innovation with accountability. From its Trust Layer and adversarial testing to transparency in human-AI interaction, the company has taken significant steps to ensure its technologies are safe, trustworthy, and beneficial to its customers. However, as platforms like Agentforce blur the lines between automation and the human workforce, the ethical questions grow more complex. Is Salesforce merely providing tools for its customers to use responsibly, or does it bear a greater obligation to anticipate and mitigate societal impacts, such as job loss? These dilemmas evoke the weighty moral questions of the Manhattan Project: how much responsibility lies with those who create powerful tools, and how much rests with those who decide how to use them?
Public fears about AI mirror the existential anxieties of past technological revolutions, driven by a lack of clear governance and an accelerating pace of change. Without a central authority like the one that governed nuclear technology, the responsibility for AI’s societal consequences is fragmented across companies, governments, and end users. In this environment, Salesforce’s commitment to ethical guardrails and customer trust is essential, but it cannot fully address the larger questions of economic and societal disruption caused by AI. As Goldman pointed out, trust is central—but so is the need for ongoing dialogue about AI’s broader impacts.
Ultimately, the AI revolution is a collective challenge that requires shared accountability. Companies like Salesforce can lead by embedding ethics into their systems, but governments, industry leaders, and civil society must collaborate to develop frameworks that balance innovation with responsibility. Without such efforts, the potential for backlash, mistrust, and inequality will only grow. As with the scientists and leaders of the Manhattan Project, today’s AI pioneers must grapple with the dual-use nature of their creations—acknowledging both their transformative potential and their unintended consequences. It is only through this clarity and collaboration that we can guide AI toward a future that benefits everyone.