How to Move Fast With AI Without Breaking Things

Many IT leaders are feeling pressure from both leadership and the board to not only deploy AI but to do so at speed, or risk becoming obsolete. Recent research from McKinsey & Company has shown the outsized potential of AI, noting for example that applications of generative AI hold the potential to add as much as $4.4 trillion in economic value to the global economy.

But what is the point of deploying at speed if the result is a trail of broken and poorly implemented AI use cases? Or worse, AI that puts the company at risk for privacy or reputational issues? When it comes to this revolutionary emerging technology, moving faster is not a direct path to better outcomes.

In the race toward AI adoption, some organizations are forging ahead so quickly that they are failing to fully think through the regulatory, legal or financial impacts. Moving at a rapid pace, while seemingly desirable, presents several challenges.

But that doesn’t mean there is not a path to deploying AI at speed. With the right technology and business strategy, organizations can accelerate their AI usage effectively.

Spotting the risks

Moving quickly has its benefits, but it also opens the door to very real risks. Constantly accelerating the adoption and use of AI without accounting for the current data governance and compliance landscape could mean falling into non-compliance, resulting in excess costs, fines and effort to come into alignment with regulations.

From a security perspective, moving quickly presents challenges too. To be effective, AI needs to draw on huge volumes of data. So, when moving at speed, the data being utilized needs to remain secure throughout the process.

“AI has the potential to be a catalyst for faster evolution across industries, but most companies struggle in moving from concept to production,” says Michael Curry, president of the data modernization business unit at Rocket Software, a global technology company that helps its customers successfully modernize critical IT infrastructure.

It’s not just a matter of security and governance. If AI is built from faulty data, or even used in an unconstrained way, it doesn’t matter how fast you’re going; the resulting output is all but certain to provide little actual value and potentially even outright incorrect information.

Companies that have been successful in deploying AI have been able to overcome these challenges by prioritizing data governance and safety. To do that, AI has to be placed into the broader modernization strategy of enterprises, assessing the requirements and impacts across people, process and technology.

Thinking through the lens of people and processes

AI technology has been leading the news cycles. Each new release of a new AI model or GPU intensifies the buzz around AI, so naturally most companies start their approach to the problem through a technology lens. Most companies have established AI teams who research the latest technologies and approaches, and build out proofs of concept for how the technology can be applied. However, often these projects are not grounded in broader context of people and processes, and hence they never achieve escape velocity against the pull of compliance and business concerns.

AI needs to be considered and used within the context of the people and processes where it will be applied. This idea of contextual AI aligns its use to the business drivers you are looking to achieve from it. AI is best suited for automating processes and making people more productive, so it makes sense that people and processes should be central to the thinking of how AI will be applied. Understanding how AI will impact and enhance people and how it will automate processes are key to unlocking the value of AI within your business. In addition, safe use of AI requires a new set of processes to be wrapped around the technology to ensure privacy and safety in the output of AI.

Prioritizing data governance and safety

With the rapid rate of acceleration surrounding AI, data governance needs to be a top priority in enabling adoption at speed and scale. More businesses are realizing this reality, opting for solutions that deliver robust and agile data governance capabilities, especially in highly regulated industries. Effective data governance involves obtaining a complete understanding of the data landscape within the enterprise, how it is all related and how it must be protected and managed. Automation has become a requirement to achieve effective governance against an exponentially growing amount of data.

Data governance is foundational to AI because it pinpoints the downstream privacy risks, and also ensures that the data you are using for decision-making is completely trustworthy. Research from IDC paints an even clearer picture of just how important effective data governance is to accelerating AI safely across an organization.

In fact, the research indicates that those with high levels of data intelligence saw a 40% boost in financial performance, a 20% increase in operations success and a massive 200% surge in the rate of reported improvements in data management metrics.

In addition to data governance, model safety has emerged as a requirement for AI. This is something that is unique to AI since there is often a lack of certainty around what the AI model will produce. Despite the intentions of the teams that build AI projects, the results can reflect biases contained in the data used to train them, or even produce outright factual inaccuracies. This can create safety or reputational issues when deploying AI. Many technology approaches have been invented to address these issues, but those are often not accommodated in the proof of concept phase, and must be hardened into place before deploying more widely. Again, there is a degree of automation that can be applied to simplify and codify these approaches, but often this isn’t well-accommodated in vendor technologies.

“Our customers are able to adopt generative AI more quickly because Rocket^® Mobius^® automates many of the privacy and safety measures required to give companies confidence in deploying it,” says Curry. “Without that level of automation, most companies get stuck trying to define and implement their own measures, which raises the bar on costs, risks and skill requirements.”

Whether it’s new technologies altogether or the deployment and rapid acceleration of AI, automation can help organizations move toward continuous monitoring of their data—something that would otherwise be difficult to achieve for a data governance team already strapped for time.

Leave the rip-and-replace approach behind

“Achieving AI at scale is a complex endeavor that must be integrated within the broader context of IT modernization to ensure long-term success and utility,” states Steven Dickens, Chief Technology Advisor at the Futurum Group. “Rocket Software’s focus is on fostering a sustainable approach that leverages existing infrastructure while embracing new technologies to expedite AI deployment.”

For AI to function correctly, and at speed, organizations need a solid foundation of data infrastructure to support its use. One of the key values of the existing infrastructure is that it is usually already well-protected through existing security and access control policies. The idea of a rip-and-replace approach—pulling the data out of the systems where it is managed today into the cloud to enable it for AI—sounds like a quick solution. Still, this approach is costly, fraught with disruption and new risks.

“Rip-and-replace is not guaranteed to be successful. Instead, organizations should orient themselves toward a more nuanced approach—modernizing in place with hybrid cloud,” says Curry. “You can continue to leverage the security and reliability of existing mission-critical infrastructure while also introducing flexibility that enables the use of new tools and technologies to help accelerate AI adoption.”

This approach is echoed by Dickens, who notes just how important it has become to move away from a rip-and-replace mindset, instead favoring an approach of modernizing in place. He explains, “At the Futurum Group, we believe that the key to successful AI deployment lies not just in speed, but in strategic modernization that aligns with core business objectives. Aligning with this approach allows Rocket Software to harness the true potential of AI without the disruption of traditional rip-and-replace methods.”

Get started with what you know

Many software companies are already starting to release new features to enable you to take advantage of AI within the context of the use of their software. Often these new features are the fastest path to adopting AI, without requiring a large investment in new AI skills. Although this offers a faster path to AI, it is still important that you confirm that you aren’t unintentionally opening the door to new risks. Some vendors might want to use your data to train their models, as an example, which could create privacy or intellectual property risks. As long as you apply the right rigor to assessing your vendor, however, these mechanisms can provide a scoped approach to AI adoption that already fits into your existing processes and is already somewhat familiar to your people.

This is also true for the systems you currently use to manage your data and content. Many of these are being extended to enable new AI capabilities to be applied directly, without requiring new infrastructure and skills. Rocket Mobius is a good example of this, enabling existing customers to apply generative AI techniques like question answering and summarization directly against the content managed within the platform, without requiring any new skills, and while adhering to the existing policies and controls protecting that content today. These types of approaches can be huge time-savers, enabling you to get value from AI much faster by making it an incremental step from where you are today. They can also reduce risk by leveraging the existing controls you already have in place within and around your software.

Taken together, these approaches can help you to move fast with AI without breaking things, and give your executive teams more confidence in adopting these market-changing technologies in a safe and effective way.