Webinar Recap: The Future of Winning, AI's Next Frontier in GovCon Business Development

Procurement Sciences recently joined GovConWire and Executive Mosaic for a forward-looking webinar on where AI in government contracting has been, where it is now, and what teams need to prepare for over the next two to five years.

Bruce Feldman, AI Platform Strategist at Procurement Sciences, led the session. Bruce brings thirty years of GovCon BD experience across capture management, proposal management, and business development leadership, and he approaches AI from the perspective of a practitioner who uses it every day rather than an engineer who builds it. That framing shaped the entire conversation: less about the technology in the abstract, more about what it actually means for the people doing the work.

Three Years In: What the Transformation Actually Looks Like

Generative AI has only been widely available for about three years. By almost any measure, that adoption curve has been extraordinary. More than eighty percent of GovCon companies are now using generative AI in some capacity, a penetration rate that exceeds anything seen with previous technologies, including mobile phones and the internet. And that number likely understates the reality, since it does not capture employees using AI outside official company channels.

The efficiency gains are real and measurable in the places where they show up most clearly. Writing is the most obvious one: AI produces grammatically correct, coherent content quickly, and it eliminates the blank page problem that slows every proposal process. Research is the other major area. Capture managers carry an enormous cognitive load, spending hours scouring websites, pulling competitive intelligence, synthesizing data from dozens of sources. AI automates most of that research, delivers the gist, and cites its sources, freeing capture teams to focus on judgment rather than data collection.

Beyond individual task efficiency, AI has started addressing structural problems inside BD organizations: data silos, collaboration barriers, bottlenecks that slow work from one phase to the next. And there is a human dimension that does not show up in ROI calculations but matters to anyone who has spent nights in a proposal room: AI gives people time back. Burnout is a real cost in this industry, and tools that reduce it have genuine organizational value.

What Teams Got Right in Early Adoption

Bruce spent time on the lessons learned from three years of watching companies bring AI into their BD processes, because the adoption experience has not been uniform.

The companies that have gotten the most out of AI share a few patterns. They started small, meaning a focused group of ten to twenty people with room to experiment, rather than a broad rollout before the platform was understood. They used a crawl, walk, run methodology: start with current work, use the AI intensively on real tasks, build proficiency over days and weeks rather than months, then expand scope and scale. They identified internal champions early and gave those people the latitude to become genuine experts and evangelists, not relying on the vendor to fill that role.

They also took their skeptics seriously. Rather than dismissing resistance, the most successful adopters actively engaged their harshest critics, gave them coaching and support, and made space for their concerns. A skeptic who becomes an advocate is more valuable than an enthusiast who was never tested.

One adoption failure mode Bruce flagged: leadership not framing a clear vision before deployment. When a new AI platform arrives without context, the workforce fills the gap with anxiety. The message that needs to come from leadership, clearly and early, is that generative AI is an enabler for people, not a replacement for them. The strategic work still requires human beings. That is not spin; it is accurate. And teams that hear it clearly tend to adopt more effectively.

Measuring ROI: The Real and the Hard

Return on investment from AI is real but genuinely difficult to measure cleanly, and Bruce was candid about why.

Win rate is the obvious metric, but win rate is affected by too many variables to isolate AI's contribution. Efficiency improvements are easier to see in specific tasks: a compliance matrix that used to take days now takes hours or minutes; a proposal volume that took a week to recover from red team can now be revised in hours of labor rather than calendar time. Those are meaningful and observable.

Some of the clearest ROI shows up in what does not happen: fewer all-hands proposal crunches that pull subject matter experts away from customer engagement for weeks; fewer consultants needed for surge capacity; less time spent on mechanical research that a capture manager had to do manually.

Bruce's practical recommendation: start measuring specific task-level timelines now, before AI is fully embedded. Without a baseline, you cannot demonstrate improvement. The AI itself can help capture those metrics over time.

Human in the Loop: Still Non-Negotiable

Bruce returned to this point several times throughout the session, and it is worth stating plainly: human oversight of AI-generated content is not optional in GovCon.

Proposals are contractually binding documents. An AI-generated assertion that turns out to be inaccurate, a hallucinated legal citation in a protest, a commitment your company cannot keep: these are not editing problems. They are grounds for award rescission, sanctions, or worse. Bruce cited the GAO and Court of Federal Claims' documented unhappiness with AI-hallucinated legal references in protests, and noted that companies have already been sanctioned for submitting them.

The right frame for human engagement is not human in the loop as a checkbox, but human in the lead, meaning the human drives the process and the AI carries the work underneath it. The AI researches, drafts, reviews, and flags. The human decides what to use, what to change, and what the strategy actually is. That division of responsibility is where the gains come from without the risk.

What Is Coming in the Next Two to Five Years

This is where Bruce spent the latter half of the session, and it is where the conversation gets genuinely interesting for teams thinking about how to position themselves.

Agents and agentic systems. Agents are already here in limited form: tools that take action on your behalf without requiring you to explicitly invoke them, like a research tool that automatically searches multiple sources when you ask a question. What is coming is more sophistication and more automation, including agentic systems where multiple AI models work in concert, each specialized for a different task, coordinated by an orchestrating model. The practical implication is cost optimization: you use a more expensive, higher-capability model for complex reasoning tasks and a less expensive model for high-volume, repetitive ones.

Knowledge graphs. Gartner projected that seventy percent of leading generative AI implementations would be incorporating knowledge graphs by now, and the trend is accelerating. The distinction matters: a standard database holds facts about entities; a knowledge graph holds the relationships between them. For capture, that means connecting a contracting officer to their job history, connecting that history to the RFPs their previous office released, and connecting those RFPs to the patterns that emerge across similar acquisitions. That relational layer dramatically enriches what the AI can surface and significantly reduces hallucination.

Predictive analytics. AI-assisted prediction is moving beyond P-win calculators. The next generation of tools will help teams forecast likely evaluation criteria before an RFP drops, estimate price-to-win ranges earlier in pursuit, model what labor categories and solution components a requirement might demand, and identify which opportunities in a broad pipeline are genuinely worth pursuing versus which should be no-bid. These are capabilities that exist in early form now and will become much more sophisticated.

Performance monitoring and dashboards. Right now, most AI platforms offer limited visibility into how the system is actually being used and where it is performing well or poorly. That is changing. Expect platforms to surface metrics on task completion times, token costs, hallucination rates, and system bottlenecks, with dashboards that let leadership make informed decisions about training, platform configuration, and data management.

Multimodal environments. Text is still the dominant mode for BD and capture work, but platforms are moving toward high-quality integration of text, image, voice, and structured data. The practical implications for GovCon will become clearer as these capabilities mature.

The AI Super Cycle and Why Cost Matters

Bruce raised the macroeconomic context because it has direct implications for anyone managing an AI platform budget.

The infrastructure buildout happening right now, data centers, chip capacity, electrical grid investment, connectivity infrastructure, follows a recognized pattern from previous technology transitions. It took roughly twenty years to build out the cell tower infrastructure that made mobile phones ubiquitous. The AI equivalent is happening much faster, but the underlying dynamic is the same: enormous capital investment today in anticipation of future demand, and someone has to pay for it.

Token costs, meaning the cost of actual AI usage, are already surprising teams that did not model them carefully. The example Bruce cited: a major company burning through its entire annual token budget in the first month of the year. As platforms become more capable and more embedded in workflows, usage will increase. Teams that are not tracking token consumption and building it into their cost models will face budget surprises.

The implication is not to avoid AI investment. It is to go in with eyes open about what the ongoing cost structure looks like and to build governance around it.

How to Choose a Platform That Will Last

The final section of Bruce's presentation addressed platform selection, framed around a question that matters more than most teams ask upfront: will this platform keep pace with what is coming?

The factors he highlighted: Does the platform have a roadmap that credibly addresses agents, agentic architectures, knowledge graphs, and predictive analytics? Can it integrate with your existing data sources and legacy systems, or will you be constantly porting data? Is it genuinely BD domain-aware, meaning built from the ground up for government contracting, rather than a general-purpose tool with a GovCon layer on top? Is the vendor financially stable and established, with a customer base that can speak to performance? And critically: can it accommodate regulatory and policy changes, like a shift from cost-reimbursement to fixed-price contracting, without requiring you to rebuild your workflows from scratch?

The platform you choose today needs to still be serving you in three to five years, through a period when the technology is changing faster than it ever has before.

The Call to Action

Bruce closed with a few direct recommendations for teams at any stage of AI adoption.

Start preparing for outcome-based pricing, both as a government expectation and as an internal management discipline. AI can help you understand what things actually cost to produce inside your organization, which is the foundation for identifying and eliminating inefficiencies.

Build user trust deliberately. Adoption stalls when people do not trust the output. Training, clear governance, and a platform that cites its sources and invites verification are all part of building that trust over time.

Future-proof your architecture. The regulatory environment is shifting alongside the technology. Teams that are wholly dependent on a single model or a single vendor approach are exposed when the environment changes. Flexibility is a feature.

Keep experimenting. The teams pulling ahead right now are not the ones who implemented AI once and called it done. They have people whose job it is to watch the frontier, test new capabilities, and bring what works back into the organization. That kind of continuous learning posture is how you stay ahead of a market that is not slowing down.

Watch the webinar to learn more.