From POCs To Impact: A 3-Pillar GenAI Strategy For Growth

Forbes

03-07-2025

Preetpal Singh, Group Managing Director Global Head Product and Platform Engineering at Xebia.
Generative AI (GenAI) has the potential to fundamentally change the trajectory of nearly every industry. Experts estimate that more than 30% of all workers could eventually see at least 50% of their occupation's tasks impacted by the technology. Companies have been quick to act: recent research from Google Cloud shows that more than six in 10 large companies (more than 100 employees) are currently using GenAI, and 74% are already seeing a measurable return on investment.
These are impressive numbers, but they can obscure a key truth: GenAI is still in its infancy and many efforts remain narrowly focused on increasing the efficiency of existing processes. The reality is that most organizations are stuck in the experimentation phase. While GenAI proofs-of-concept (POCs) are everywhere, very few have been successfully scaled. Gartner predicts that one-third of all GenAI programs will be abandoned after the POC stage by the end of 2025, underscoring the difficulties facing companies as they try to operationalize functional progress.
The Scaling Challenge: From Pilot Projects To Enterprise Programs
While GenAI can be powerful, the journey to genuine impact isn't always easy. When businesses isolate a promising use case, they may have to deal with scaling safety or proving cost-effectiveness. That's because scaling GenAI isn't just about replicating use cases—it's about aligning technology, data, people and governance across the enterprise. As multiple departments, managers and processes get involved, complexity grows, leading to challenges.
Companies commonly encounter the following barriers when attempting to scale GenAI:
• Managing fragmented infrastructure that is siloed and difficult to integrate
• Preventing shadow IT (registration required) and decentralized experimentation with unvetted applications
• Improving enterprise-wide GenAI literacy and structured adoption plans
• Mitigating security and compliance risks across use cases
• Controlling cost overruns caused by unmonitored model usage
As industries move from GenAI experimentation to adoption, they need a careful implementation strategy—one that includes a foundational, scalable framework to organize the work ahead.
Building A Scalable Framework Around 3 Key Pillars
While small functional changes often target specific tasks, building and scaling enterprise-wide GenAI solutions requires a comprehensive, end-to-end approach. Given all the moving parts, having a framework in place helps ensure initiatives are thoughtfully crafted, efficiently deployed and continuously refined for maximum business value. Here is a breakdown of how that framework could look, along with industry examples:
When companies think about scaling GenAI, there is a natural inclination to start with the technology. This is a common mistake. Leadership teams should begin by aligning business goals and identifying high-impact use cases designed to drive value. By concentrating on outcomes first, teams can ensure initiatives are focused, scalable and tailored to the organization's needs. From there, it's important to fine-tune large-language models (registration required)—sometimes referred to as foundational models—with industry-specific data, or to experiment with multiple model architectures to find an optimal approach.
In 2024, JPMorgan rolled out a GenAI assistant to 60,000 employees to help with tasks like writing emails and reports. Rather than developing its own GenAI model trained on sensitive financial data, the bank designed its assistant as a portal to access external large-language models (LLMs). This is a strong example of starting with a business outcome—improved productivity—and selecting the right tech to meet it. If the bank had started with building its own model, it could have introduced risk that would hinder scaling.
In business, speed is critical—but even the fastest cars have brakes. Any GenAI operating system must include guardrails for data privacy, security, compliance and ethical use. This requires diverse development teams and rigorous bias detection and mitigation strategies. Many companies now embrace a "white box" approach where GenAI models provide insight into their decision-making processes, helping users understand the factors behind the outcomes.
Here are some examples of responsible AI governance:
• Establishing tool kits, like IBM's watsonx.governance, to help manage complex regulations with minimal manual oversight
• Creating ethical AI committees, like Microsoft's Aether, to certify the responsible development and deployment of GenAI technologies
To support scale, companies must invest in GenAI literacy and ensure systems can evolve with shifting business needs. This includes building a culture of experimentation and providing pathways for teams to grow their skills. Organizations can take action to drive readiness:
• Enabling employees to experiment with user-friendly, self-service tools
• Delivering continuing education and structured upskilling programs
• Deploying GenAI or agentic AI builders to support departmental onboarding at scale
In 2024, AstraZeneca announced a GenAI accreditation program—a self-guided training course to upskill employees on responsible GenAI use. The program aims to boost confidence and nurture the human skills needed to thrive in the AI era.
From Experiments To Enterprise Value
The next 12 to 24 months will define winners in the GenAI space. Now is the time to scale. Doing so requires the right framework, built around strategy, governance and people. Putting in this work supports adoption across the organization and ensures GenAI initiatives align with business objectives. With it, GenAI becomes a force multiplier for sustainable growth.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?