By Karun Sarabhai, Senior Director for Agile Program Management, Publicis Sapient
As Generative AI (GenAI) reshapes how organizations engage customers, create content and unlock new avenues for innovation, the question is no longer “if” but “how”, and more importantly, “how fast and how well?”
Traditional program management thrives on predictability—defined milestones, fixed scope, and linear progress. GenAI flips this paradigm. It replaces deterministic logic with probabilistic creativity, where outcomes emerge through experimentation. Program managers must, therefore, shift from rigid roadmaps to dynamic hypothesis backlogs, treating each sprint as a learning loop rather than a delivery checkpoint.
Building the GenAI Operating Model: From Strategy to Execution
In today’s AI-era, cross-functional collaboration is essential. GenAI teams blend data scientists, prompt engineers, legal advisors, and product strategists. Roles blur: product managers must grasp prompt tuning, while data scientists must consider UX. This convergence demands leadership that cultivates psychological safety, encourages experimentation, and aligns diverse disciplines around shared outcomes.
Scaling GenAI isn’t just about the tech; it’s about getting every layer right. Here are the essentials leaders must consider:
Navigating the legal maze: GenAI systems can hallucinate, reflect bias, or generate harmful content. This calls for embedding a Responsible AI framework early, one that is complete with laws and regulations, human-in-the-loop reviews, and traceability. As AI regulations evolve, continuous legal engagement is critical.
Ethical implications: Organizations must ensure that their GenAI systems are designed and deployed in a manner that is ethical, transparent, and aligned with societal values. This involves implementing ethical guidelines, conducting regular audits, and engaging with diverse stakeholders to address potential biases and ensure fairness. Leaders must also prepare their teams to operate confidently in an environment defined by ambiguity.
From Control to Curated Chaos
Culturally, GenAI introduces curated chaos. Traditional leadership relies on control; GenAI demands comfort with ambiguity. This can introduce discomfort. The onus is on project teams to put a comprehensive training plan in place to ensure that we address stakeholders’ fear of losing control, disbelief around potential and efficiency, initial frustration especially if the users are not well trained on prompt engineering and other such issues. However, change management isn’t a support function – it’s a core capability in GenAI delivery. Teams must invest in training to overcome fear, disbelief, and frustration, especially around prompt engineering. A robust management strategy is key in supporting this transition.
Unleashing the Power of Data
Data is the foundation of GenAI success. But more than just volume, it’s about intentionality. GenAI systems need purposeful curation of data – prioritize high-quality, domain-specific datasets to reduce hallucinations and improve model alignment. There’s a strong requirement to implement data governance through consent tracking and audit trails to ensure regulatory compliance. Also important is to embed continuous feedback loops and democratized access to accelerate learning and innovation. Finally, data must connect to outcomes – whether it is improving response quality or accelerating content generation, every data decision must be tied back to measurable business outcomes.
Reinventing Testing & Models
Testing GenAI systems requires a new approach. Instead of binary pass or fail tests, assess outputs against acceptability thresholds such as intent, tone, and factual accuracy. Prompt testing simulates real-world user intent, while automated pipelines flag anomalies. QA teams must include linguists, ethicists, and domain experts to evaluate appropriateness, inclusivity, and safety.
Before selecting a model, teams must evaluate suitability, ethical implications, and integration readiness. Maturity isn’t static since new models and regulations emerge constantly. Assessments must move away from demos and simple use cases. They need to be contextualized to the domain, tested in real world scenarios and align to the risk profile and business ambition. To do it, teams must:
Establishing a safe environment for teams to learn
The iterative nature of GenAI development means that teams must be prepared for continuous learning and adaptation. This involves not only technical adjustments but also shifts in organizational culture and mindset. Encouraging a culture of experimentation and learning is crucial. Teams should be empowered to take calculated risks, learn from failures, and iterate on their approaches. This iterative process helps in refining models, improving accuracy, and enhancing overall performance.
As organizations look to scale their GenAI initiatives, they must address challenges related to infrastructure, data management and resource allocation. Investing in scalable infrastructure, leveraging cloud-based solutions, and adopting best practices for data management are essential for supporting the growth and expansion of GenAI programs. Scaling solutions must also factor in talent – upskilling existing talent or hiring specialists which are going to be crucial to supporting large scale GenAI solutions.
The old playbooks aren’t necessarily applicable when managing GenAI programs. Emerging technology demands new playbooks that are written in real time. These programs challenge our assumptions about delivery, governance and leadership. Ultimately, it requires a shift from certainty to curiosity, from control to orchestration, and from static plans to adaptive systems. As leaders, our role is to foster safe experimentation, ethical grounding and creative empowerment.
The future of GenAI is still unfolding. Those who lead with clarity, courage and collaboration will not only deliver successful programs but also define the blueprint for enterprise innovation in the AI era.