Discover proven ways to navigate common risks, optimize performance, and build capabilities as your organization moves beyond initial GenAI experimentation.

Despite its potential to boost productivity, many organizations are still hesitant to scale generative AI (GenAI) beyond proof-of-concepts. Commonly cited concerns include data privacy, output accuracy, unclear return on investment, and potential legal and regulatory implications.

Robust AI governance, policies and standards, clear usage guidelines, and deliberate rollouts are building confidence and clarity. However, demonstrating clear ROI with AI to justify project costs, especially at the C-suite and board levels, remains a significant hurdle.

This challenge stems partly from the difficulty in quantifying the productivity gains in knowledge work that generative AI optimizes. For example, translating improvements — such as reducing resolution time from 10 hours to 1 hour for an HR chatbot — into business value is not always straightforward.

Without clear metrics to map tangible outcomes, it becomes challenging to convince management boards to invest further. Given these hurdles, what can IT leaders do?

Five steps to better GenAI ROI
With a wide range of powerful GenAI models now available, organizations have unprecedented opportunities to leverage these tools for productivity and creativity.

These models can be applied across domains, including text generation, coding, trend analysis, translation, video analysis, and content creation. Selecting the right model for each business need is essential.

1. As a first strategy, make sure to choose the right GenAI platform by:

  • Prioritizing flexible and task-appropriate model choices rather than defaulting to the most sophisticated option.
  • Embracing simplified workflows and robust safety guardrails for AI application development
  • Ensuring the development process is as efficient and secure as possible

One useful reference is the deployment of an AI and machine learning platform in the public sector, which streamline document processing, improve insight extraction, and save substantial staff hours across multiple organizations.

2. The second step is to improve AI outcomes through prompting and context. Performance is not determined by model selection alone. Combining fit-for-purpose models with best practice prompting techniques — such as prompt engineering — can significantly enhance accuracy and cost-effectiveness:

  • Multi-shot prompting, where users provide several examples of desired outcomes, helps calibrate model responses for greater consistency and relevance.
  • Retrieval-augmented generation (RAG) is another effective tactic. As generative models are limited by their original training data, RAG supplements a model’s responses with up-to-date or context-specific information, enhancing accuracy and relevance while reducing the need for post hoc human intervention. For example, onboarding chatbots built with retrieval augmentation are able to handle complex customer queries rapidly and accurately.

3. The third strategy is determining the right size, speed and efficiency of a GenAI model to use. While advanced models can address most tasks, using them for simple operations often increases cost and latency unnecessarily. The recommended approach is to select the smallest, fastest, and most efficient model that reliably fulfills each requirement.

4. Next, the importance of human oversight and responsible AI cannot be overstated: GenAI should not be left to operate in isolation. Essential practices include:

  • Maintaining human oversight, curation, and active feedback loops
  • Adhering to established responsible AI guidelines to safeguard outcomes and maintain business trust
  • Recognizing that, as of now, end-to-end business process automation with generative AI remains out of reach; effective human-AI collaboration is critical for robust and trustworthy systems

5. Finally, building organizational capability is crucial. Realizing GenAI’s full potential depends on workforce development. Regular, tailored skills training ensures teams can deploy, supervise, and continuously improve AI systems effectively.

Significant investment in national and enterprise AI talent and infrastructure is increasing globally, supporting accelerated adoption through access to advanced tools, computing resources, and expert consultative support.

By following these approaches: thoughtful model selection, prompt engineering, retrieval techniques, human oversight, and talent development — organizations can capitalize on GenAI benefits safely, securely, and at scale.