Rethinking past performance in the age of GenAI

Today’s pace of emerging technology development in the “Age of GenAI” is unprecedented.  Its disruptive impacts span nearly every facet of our collective workforce in ways traditional AI does not. This widespread impact across both technical and non-technical disciplines, in government and industry alike, requires that we continue to rethink how we manage this disruption.

The state of the art is advancing so far ahead of adoption in government and much of industry that if all of today’s engineers simply stopped working, it would take many years to fully realize the transformation possible with what’s available today. This rapid progress means the pace of digital and business process transformation must also accelerate if we are to achieve optimal mission outcomes that fully leverage the technology’s capabilities.

Such acceleration is crucial to maintaining the United States' leadership in researching, developing, and evaluating the robustness, resilience, and safety of artificial intelligence systems across domains.

Traditional acquisition creates obstacles

One significant hurdle in this accelerated transformation is the lengthy process of government technology acquisition and associated evaluation criteria. Traditionally, for procurements outside of research and development or other specialized transaction authorities, agencies rely heavily on a vendor's related past performance, looking for evidence of similar work done for other agencies as a measure of reliability and capability.

However, as the pace of technological advancement quickens, there simply won't be a broad history of past performance to draw from. This reliance on historical data becomes increasingly problematic while accelerating adoption, creating a bottleneck in the adoption of cutting-edge solutions and imposing an undesirable lack of competitive diversity. Agencies must find new ways to evaluate potential partners and their ability to deliver innovative technologies effectively if we’re to move at speed.

Hot: Current performance. Not: Past performance

To address this challenge for technologies such as GenAI, agencies need to shift their focus from solely evaluating past performance to assessing a contractor's current and active use of the technology. Take an approach that considers “Client Zero”—the provider itself—as an authoritative source. By examining how vendors are leveraging these advanced capabilities within their own operations, agencies can gain a more accurate and up-to-date understanding of their expertise and readiness, and the veracity with which they implement their own offerings and capabilities. This approach not only provides a clearer picture of a contractor's ability to innovate and adapt, but also ensures that the solutions being implemented are grounded in real-world application and experience. This shift in evaluation criteria is essential for keeping pace with the rapid advancements in generative AI and other emerging technologies.

We’ve seen this challenge, and the solution, play out firsthand over the past two years as we took our AI @ CGI Federal story to market and engaged our clients well before the government adoption curve began to rise. Our Federal Emerging Technologies Practice leaned in aggressively very early as generative AI technologies became commercially available. We recognized the imperative to embrace AI to drive our own digital transformation and improve internal business processes.

As the pace of AI technology advancement continued to accelerate in the early phases, we detected an ever-widening workforce skills gap that needed to be mitigated to adapt and thrive. AI training was widely employed, terabytes of it. However, until we actually deployed a generative AI solution into the hands of our own workforce, there was limited closing of that skills gap achievable. Moreover, just like our federal agency customers and partners, our workforce had access to a growing portfolio of generative AI solutions outside the secure boundary of our network, tools that in most cases were neither safe nor legal to use with the data inside our network. Until we delivered safe, secure, responsible, and monitored generative AI capabilities to the workforce, we were doing nothing beyond policy to truly mitigate the risks.

We faced the dual challenge of meeting our own operational efficiency goals while simultaneously learning and understanding AI capabilities to stay ahead in the U.S. Federal market. This required us to effectively progress through the AI maturity cycle—crawling, walking, running, and eventually flying—while mitigating risks and ensuring security compliance in line with stringent federal information security regulations.  The result was an entire corps of newly developed elite GenAI subject matter experts in CGI Federal, an enterprise-scale and agentic generative AI platform for use across all sectors of our business operations—and a workforce of nearly 8,000 partners—and a robust business transformation in flight.  

That was a year ago at this point, and it fundamentally shaped our conversations with existing and potential clients assessing their own journeys in this space. Our secure, compliant, and remarkably capable internal use at scale predated federal adoption, while our ability to introspectively demonstrate our own business process reengineering and return on investment was the north star proving the technology’s value.  In applications from software development to talent management, from business development and proposal content generation to help desk support, we’ve consistently reduced weeks to hours, and hours to minutes.

We haven’t stopped innovating, and our own transformation continues at speed. As we continue to accelerate achieving mission outcomes for government in the Age of GenAI, there will be no better measure to assess a partner to build with than a deep and introspective look at how those technologies are used safely, securely, and profitably within. I personally invite you to connect and embark on that conversation with us to see the impact that’s possible.