How government and industry can leverage AI to transform the federal IT marketplace

AI offers the potential to dramatically transform government, both in terms of technical capabilities and operational impacts. But it is also resulting in significant marketplace disequilibrium. In particular, that disequilibrium calls into question many of our steady-state assumptions at the heart of the government’s acquisition strategies and industry’s processes for designing winning solutions to address the government’s requirements.

As the government takes its first tentative steps on the AI regulatory front, there is a near-term need to protect the government’s core networks and applications from vulnerabilities, especially zero-day bugs, such as those uncovered by Anthropic’s Claude Mythos. Technology operations and maintenance (O&M) processes and procedures that had been a matter of routine operational hygiene, such as patch management, have taken on a new level of urgency. They must be handled at scales and speeds that the government and its contractors have never had to deal with before. The potential impacts are orders of magnitude greater than those from recent generations of cyber threats.

The turbulence and uncertainty in the IT marketplace, brought on suddenly by the AI labs, call into question the superiority of technical solutions that have proven effective in recent years. Many of the “black belt” processes and government procedures that help ensure strong, consistent performance on today’s contracts may make it difficult to innovate at speed by applying the latest AI models to reengineer our approaches to challenges such as network modernization. 

Many of the platform, infrastructure, and software systems that have proven their worth for enterprise resource planning, customer relationship management, financial operations, and human resource management are now at risk of becoming obsolete due to the power of AI models, which enable rapid, low-cost development and customization of tailored replacements.  

Shifting landscape for bidders

The challenges posed by the marketplace AI revolution will clearly impact government contractors, requiring bidders to reinvent well-established strategies, processes, and tools to win contracts. A major part of the challenge in assessing the government’s requirements — the primary job of industry’s business development specialists — is developing a clear understanding of the government’s practical and perceived constraints.

On top of the challenges for industry in requirements assessment, there’s a whole new set of challenges in solution development.  

AI has expanded the range and quality of technical and management solutions available to optimize mission support and modernization solutions. As the decision space has expanded, the analysis of alternatives has grown more complex. How much optimization is too much? How do we retool operational processes and procedures to enable greater AI decision support and more automation? Which systems and tools are going to continue to deliver high-quality results, and which ones will fall victim to marketplace consolidation and obsolescence brought on by AI? Which architectures will work best to maximize the benefits of AI solutions?

Then we come to the executives and spreadsheet-jockeys responsible for developing pricing strategies, building price-to-win models, and putting together the price proposal. It’s hard enough to price bids in normal times. Industry is just starting to figure out how to get control over the costs for AI tokens. How will they be able to project these costs over the life of a five-year or 10-year government contract when the AI marketplace is so immature and unstable?

Perhaps the toughest challenge of all: How to estimate the Level of Effort (LoE) required for next-generation solutions that leverage AI? Greater automation and improved decision support through the application of AI should reduce the need for the number of government and industry personnel involved in more routine tasks. What basis will industry provide for claiming adequate labor hours in their staffing plans when there’s so little real data to underpin fair and reasonable Level-of-Effort estimates?

And what basis will government evaluators use to distinguish between proposals that make legitimate claims to improvements in efficiency by leveraging new AI technologies and those that hide behind AI to low-ball their bids?

Adapting to a new procurement era

The AI shock is so recent, technical developments are so fast-paced, and the commercial IT markets are so unstable, that it’s hard to develop strategies that won’t become outdated in a matter of weeks when the AI labs announce their next breakthroughs and the impacts ricochet through corporate America and our government and military IT hierarchies. The federal IT marketplace has always demanded decision-making under uncertainty, but we’ve never experienced a time with such extreme levels of unpredictability.

First and foremost, this is a time to explore new approaches that can help federal agencies, military services, and departments benefit from the tremendous potential of recent advances in AI while protecting against the growing list of major risks. Let’s start with the basics.

On the government side, market research is a built-in requirement for every acquisition, but we need to do more than issue a request for information shortly before developing an acquisition strategy and formulating a request for proposal.  We need to make a stronger commitment to maintaining open communication between government and industry so that both sides get smarter about the growing number of technical options and are better equipped to adapt to a radically changed decision space.  

The government needs to hear firsthand from the OEMs and cloud service providers (CSPs) who are working at a furious pace to develop new AI offerings and integrate AI features into established products. Don’t overlook the system integrators, consulting firms, and Value-Added Resellers (VARs) who are designing and refining tools to support AI governance, network and operational performance impact assessments, and FinOps cost management optimization.  

A model for building autonomous networks

A huge challenge facing the government in this time of marketplace turbulence is the need to build frameworks to guide AI adoption. The TM Forum, for example, created a roadmap for “… building an industry standard architecture that enables self-healing, self-optimizing network domains, readying the network for autonomous ‘Zero-X’ (zero wait, zero touch, zero trouble) operations,” which they have labeled Autonomous Network Management (ANM). 

Their Autonomous Networks Project created a six-step taxonomy that CSPs can use to measure their progress in moving from fully manual to fully autonomous operations. The steps include Manual Operations and Maintenance (Level 0), Assisted Operations and Maintenance (Level 1), Partial Autonomous Network (Level 2), Conditional Autonomous Network (Level 3), Highly Autonomous Network (Level 4), and Fully Autonomous Network (Level 5). Major players in industry are applying this framework for infrastructure management, and it’s worthy of the government’s attention.

The ANM architecture provides a useful model for refining and standardizing our approaches to network modernization. This would help the government accelerate the development of its acquisition strategies, reduce the time required to develop RFPs, improve the quality of proposals, and establish more consistent standards for post-award contract management.  

That will be important in light of the fact that the government continues to work under serious constraints these days, with fewer personnel available to perform critical technical, operational and acquisition functions. Much of industry is under similar constraints, with companies cutting back on payroll and investing in AI products and application development.  

We can’t leave this to the CDAOs and CTOs. The turbulence in the IT marketplace, brought on by the rapid acceleration and commercialization of capabilities from AI labs, has the potential to transform every aspect of government and military operations, and every segment of industry that supports them, for better or worse.

We need to be all in on the government, military, and industry sides to figure out where we can and should draw the lines between autonomous and human-in-the-loop operations, between hype-cycle claims and real capabilities, and between leveraging new AI tools to build streamlined, optimized program management organizations and the risky hollowing out of mission-critical support and delivery teams.

These recommendations will require greater participation from both sides of the federal acquisition market research process — and a significant investment of time and resources by both the government and industry. That includes not only preparing and evaluating proposals that leverage AI technologies, but also the additional post-award effort required on both sides to achieve results.

Even though AI promises to save time by automating coding, streamlining routine processes, and speeding up data analysis to identify action alternatives, for important tasks like root cause analysis, it will take both industry and government dedicated time, intense focus, and loads of effort, sweating the details to get targeted results. 

It will also be important to leverage the enterprise-level agencies with the heft of DISA and GSA, who have the engineers and acquisition specialists to do the heavy lifting needed to minimize the load on end-user agencies, services, and departments.  

Despite the powerful potential of AI technologies, we can’t outsource our brains or our decisions, and we can’t dodge our responsibility to navigate today’s turbulent markets to deliver meaningful results for the American people.

Warren Suss is president of Suss Consulting, Inc., which currently has over 70 federal civilian, defense, and industry IT experts specializing in federal information technology, market analysis, and proposal development, serving systems integrators, network service providers, equipment manufacturers, software developers, and cloud service providers.