AI Creates New Jobs That Demand Human Judgment

In the past, technical roles like software engineering often followed a predictable pattern: build a product in isolation, hand it off to customers, and move on to the next project. This model emphasized coding proficiency and technical deliverables as the primary measures of success, with implementation teams focused on deploying solutions rather than ensuring they drove tangible outcomes. According to an analysis by Indeed and the Financial Times, job postings for forward deployed engineers (FDEs) saw an 800% spike between January and September 2025, signaling a shift as companies adopt AI agents rapidly but lack the expertise to deploy them effectively. This surge mirrors the rise of data scientists a decade ago, when Harvard Business Review deemed that role the sexiest job of the 21st century, highlighting how transformative technology creates entirely new positions rather than eliminating them.

FDEs exist to close the gap between AI capabilities and real-world business value, operating as a hybrid role that combines coding, consulting, and on-site collaboration with customers. Unlike traditional engineers who work remotely or in silos, FDEs pack a laptop and go directly to client sites, aligning their efforts with specific business outcomes and staying accountable until those goals are met. Sarah Khalid, an FDE Director at Salesforce, explains that implementation teams build solutions, but FDEs make sure those solutions drive value, emphasizing a shift from predefined checklists to outcome-oriented work. This approach reflects a broader change in how AI is integrated, moving beyond mere software implementation to deeply understanding and solving customer problems.

Ology for developing FDE roles involves internal restructuring and targeted training programs, as seen at Salesforce where the company tripled its FDE team in just six months by drawing from three internal teams: engineering, professional services, and customer success. To bridge skill gaps, Salesforce built a six-week onboarding program in September 2025, with two weeks of focused technical training and four weeks of on-the-job experience, prioritizing curiosity over formal credentials. Ruth Hickin, VP of Agentic Workforce Strategy and Innovation at Salesforce, notes that the most important criterion is whether someone can solve a customer's problem, with around 40 to 50% of FDE hires coming from internal movements. This internal focus allows for a cumulative growth path, as exemplified by Khalid's transition from developer to architect to FDE, which combined depth and perspective rather than following a straight line.

From this approach show that FDEs require a unique blend of technical and non-technical skills, with coding described as table stakes and human judgment emerging as the key differentiator. Khalid reports that working with AI platforms like Agentforce demands new abilities such as prompt engineering, instruction-writing, and comfort with nondeterministic outputs, which are less black-and-white than traditional coding. The pace of innovation also accelerates, shifting from three releases a year to weekly or biweekly updates, making adaptability a baseline requirement rather than a bonus. Strong FDEs excel in qualities like pattern recognition, delivering hard truths to stakeholders without alienation, and fostering trust, as demonstrated in cases where technical s were overshadowed by the need to rebuild customer confidence.

In practice, FDEs spend less time on administrative tasks due to AI automation, which has reduced activities like meeting summaries and status updates from up to 40% of their workload, freeing them to focus on hacking, building, and solving customer issues. Khalid provides concrete examples, such as helping a tech customer achieve a 70% deflection rate for customer service, which led to expanded use cases, and assisting an insurance company with accuracy targets that required experimenting with different LLM models. These cases highlight that success often hinges on non-technical skills like communication and trust-building, with Khalid reflecting that there's not much coding involved—it's more about judgment and skills that AI cannot replace. This evolution suggests that human jobs become more human as AI handles routine tasks, allowing professionals to engage in higher-value, creative problem-solving.

Of the FDE role extend beyond immediate job creation, serving as a leading indicator of where technical work is heading: closer to the customer, more dependent on judgment, and more integrated with business outcomes. Hickin draws a parallel to data science, which started as a niche specialty in 2012 but evolved into a baseline expectation across job functions, suggesting FDEs may follow a similar arc toward broader literacy. Khalid proposes reinterpreting the FDE acronym as future defining engineers, emphasizing that these professionals are not just defining AI applications but also shaping the future of the role itself. This perspective underscores how AI-driven transformations repeat historical patterns of job evolution, but at an accelerated pace, reinforcing that technology augments human capabilities rather than replacing them entirely.

Limitations of the FDE model include its reliance on internal talent development and of scaling such personalized, on-site roles across diverse industries and company sizes. The analysis by Indeed and the Financial Times focuses on job posting spikes without detailing long-term sustainability or potential saturation in the market, and Salesforce's experience may not generalize to all organizations, especially smaller firms with fewer resources for dedicated FDE teams. Additionally, the rapid pace of AI innovation requires continuous learning and adaptation, which could strain professionals unprepared for constant skill updates, and the emphasis on non-technical skills like judgment and trust may be harder to quantify and train compared to technical competencies. These constraints highlight the need for ongoing evaluation as the FDE role matures and spreads beyond early adopters.