How to maximize GenAI Return on Investment (ROI)

Unlocking the full potential of Generative AI or GenAI starts with understanding its ROI. For businesses navigating GenAI, knowing how to maximize returns the right way is key to transforming AI from a cost into a powerful asset.

I’m Maud Stigter, an AI expert, and together with my colleague Willem Zeiler, we’ve developed practical strategies to help businesses unlock cost-effective GenAI adoption.

By focusing on key factors like gradual implementation, making smart “buy vs. build” decisions, and aligning AI investments with core business priorities, we will show you how to turn GenAI into a powerful and value-driven part of your operations.

Full utilization of features for GenAI ROI

To make AI cost-effective, it’s essential to fully leverage the available GenAI features, especially those in platforms like ServiceNow. While deploying all features immediately may not be feasible, having them on the roadmap and incrementally adopting them is a smart approach. 

Start with foundational technologies like performance analytics and process mining to track and optimize your processes. From there, move on to predictive intelligence, which can predict service offerings or assist with tasks like field assignments, helping agents resolve issues faster by identifying similar incidents from the past. Each step improves efficiency, laying the foundation for future GenAI adoption. 

GenAI's role in automation and predictive intelligence 

As you move forward, integrating GenAI capabilities will further increase cost-effectiveness.

Automation plays a crucial role here.

For example, by utilizing GenAI to automatically summarize incidents or cases, or predict solutions, the time spent on repetitive tasks decreases, allowing employees to focus on more value-driven work. Over time, these efficiencies lead to significant cost savings. 

Buy existing AI solutions, customize, or build your own?

To maximize GenAI ROI and potential, organizations need to carefully assess their options. Should you buy existing AI solutions, customize them, or build your own? Here’s how to make that decision: 

• Buying pre-built solutions: Many organizations start with pre-built GenAI capabilities like ServiceNow’s NowAssist. These solutions offer a quick, cost-effective entry into AI because ServiceNow maintains the whole stack. This makes it a commodity solution for many businesses. 

• Customizing AI for tailored solutions: If your organization needs more tailored GenAI solutions, customization is a viable option. While more resource-intensive than pre-built options, working with partners allows you to tweak and adapt existing AI models to better suit your unique business needs. 

• Building your own AI models: For organizations with significant intellectual property or specialized needs, building your own AI or large language model (LLM) from scratch may be worthwhile. However, this is the most expensive option, requiring a high level of expertise, resources, and ongoing maintenance. It’s recommended only when full control over data and processes is important. For example, in areas dealing with proprietary information or critical business functions. 

Balancing GenAI costs and long-term benefits

Determining ROI isn’t just about evaluating the costs—it’s about what you’re putting into the system and the long-term benefits you’ll gain.  

For example, in cases where intellectual property is at the core of your business, developing your own GenAI models might offer a higher ROI as it ensures full control over your data and processes. However, for more standardized tasks like incident summarization, pre-built solutions are more cost-effective and deliver immediate value. 

Choosing between in-house AI and off-the-shelf solutions

If your organization relies heavily on proprietary data or intellectual property, building your own AI model may provide higher value in the long term.

Having an in-house model ensures that sensitive information remains secure and under your control. However, this route is costly and resource-intensive, requiring specialized teams to develop and maintain the model. 

On the other hand, for routine tasks like incident summarization, where AI solutions are already mature and readily available, purchasing off-the-shelf GenAI tools is a far more efficient and cost-effective approach.