After our recent webinar on practical AI for government contractors, we received dozens of questions from CGOs and executives. The questions revealed a common thread. Leaders want clarity on what the government is actually doing, which tools are acceptable for contractors, and how to stay compliant while improving internal speed and accuracy.

This is the first in a three-part series. The questions below came directly from executives who attended our Deltek webinar on practical AI for government contractors.

Proposal Evaluation Status

Question: Where does the current market stand on using AI for solicitation submission evaluation. Are we actually seeing more use of AI to review, score, and award solicitations?

Agencies are beginning to use AI to assist with the administrative and analytical work around proposal evaluations, but people are still making the final award decisions. Under FAR 15.305, evaluators are required to assess proposals against the stated factors and document their rationale. AI can organize, summarize, and check data, but procurement staff remain responsible for the judgment and selection process.

Where Does AI Adoption for Proposal Evaluation Stand Today?

The federal acquisition community is in the assistive phase of AI adoption. Agencies are using tools that reduce repetitive tasks, not those that automate decision-making. Here’s a rundown of where the market stands today:

• Governance and policy. OMB M-24-10 requires oversight and risk management for any AI in use. NIST’s AI Risk Management Framework and Generative AI Profile set the standard for responsible deployment.
• Measured adoption. GAO reports that government AI use cases nearly doubled between 2023 and 2024, focused mainly on administrative and analytical functions
• DHS pilots. The DHS Procurement Innovation Lab is testing AI tools to help evaluators locate and interpret CPARS and past performance data more efficiently. Phase II expanded to 10 agencies to explore retrieval and triage tools for evaluators.
• GSA’s Procurement Co-Pilot. GSA uses internal copilots for market research, “prices paid” analysis, and clause checks. These assistive tools improve speed and consistency.
• DoD and DHA pilots. DoD components are experimenting with AI-enabled acquisition support, including Army SBIR topics focused on source-selection assistance and CDAO-led pilots on AI governance. The DHA’s “Ask Sage” operates in a secure enclave under a published Privacy Impact Assessment, showing production-level use under strict policy controls.

Agencies are moving quickly to test AI as a time-saver and analytical assistant but remain cautious about delegating judgment or making source selection decisions. For industry, proposals must still persuade humans and be structured clearly for machines, structured clearly with Sections L/M, and evidence-based to help evaluators (and their tools) find and score your strengths efficiently.

What is Permitted?

Question: What AI software IS permitted for government contractors to use and is in compliance with DoD GovCon rules?

Commercial GenAI tools are permitted when used appropriately. The key is aligning the tool with your data’s sensitivity, the network environment, and contract clauses.

We recommend breaking this into three questions:

1. Where are you working?
   • On a government network or government-furnished system, use only agency-approved tools.
   • On your company network: move to step 2.
2. What data will touch the tool?
   • Public or marketing data and general BD research: Use commercial enterprise tools like ChatGPT Enterprise, Claude Enterprise, or Gemini Enterprise.
   • Company confidential (not CUI/CDI): Use enterprise editions only, using SSO, logging, DLP, and “no training on your data.”
   • Federal sensitive data (CUI/CDI, export-controlled, PII): Use government-authorized clouds that meet contract security clauses (e.g., DFARS 252.204-7012, NIST SP 800-171/CMMC). Typically, Azure Government/Azure OpenAI, Microsoft 365 Copilot in GCC or GCC High. For DoD CUI, host in environments aligned to DISA IL4/IL5.
   • Classified: Do not use public GenAI. Follow your classified system rules.
3. Do your contract or agency policies restrict GenAI?
   • Some solicitations and ATOs add specific prohibitions or data-handling rules. If a clause conflicts with your tooling, the clause wins.

Practical Guardrails

• Use enterprise versions, not free consumer tools.
• Enable SSO, audit logs, retention controls, DLP, and “no training on your data.”
• Tenant boundaries matter. Use GCC/GCC High (or agency-approved equivalents) when handling sensitive or DoD data. Validate where prompts and responses are processed and stored.
• Always maintain human review for anything affecting compliance, pricing, or award.
• Document your AI use in a short SOP covering approved tools, data types, reviewer roles, and retention.

Which Tool Should I Use?

Question: Do you have any insights on the maturity of AI solutions for supporting proposal development and management. Any known leaders. What sizes do they support. Are they costly or affordable?

For using AI in government contracting, you could either go with an AI software that is tailored to the GovCon growth lifecycle or integrate one of the leading foundation models within your process.

GovCon Specific AI Tools Overview

Over the last 2-3 years, AI tools for federal growth have evolved rapidly. There are more than 100+ AI software tools specifically tailored to the GovCon growth lifecycle. Most of these platforms support BD research, streamline capture functions, and support the full proposal lifecycle, from outline development to content drafting and review, with flexible pricing by seat or API usage. The key is to always validate vendor claims before integrating into your workflow.

While many of these tools have different benefits, the best tool may depend on your organization. This could be based on your expected use case, anticipated ROI, data sensitivity, and team size. We hesitate to make any blanket recommendations since there are many tools that could provide value based on your company’s objectives.

Foundation Model Overview

Below are the more commonly used GenAI foundation models used by federal contractors and the government community. These are broad summaries, and we recommend companies evaluate each to determine which is most suitable for their needs.

• Microsoft 365 Copilot (GCC/GCC High). Ideal for teams working in Word, PowerPoint, Outlook, SharePoint, or Teams. Drafts and edits inside Word and PowerPoint while pulling facts from SharePoint/Teams, which speeds section writing and review without leaving your tenant.
• Azure OpenAI / ChatGPT in Azure Gov. Use when you need Microsoft cloud boundaries and custom apps/agents. Great for retrieval over your library and heavy analysis. Pairs well with Copilot.
• ChatGPT Enterprise (OpenAI). Use to synthesize market research, summarize RFPs, and structure proposal sections from scattered notes or legacy content. It supports collaborative use across teams for win themes, storyboards, and draft volumes. Keep to non-sensitive data unless your policy allows more.
• Google Gemini Enterprise / Workspace. Best suited for teams standardized on Google Workspace. It can summarize large volumes of source material from Drive, analyze RFP attachments, and help draft outlines directly in Docs and Sheets.
• Anthropic Claude Enterprise. Excellent with long, precise instructions and big document sets. Good for long RFPs+ attachments, policy summaries, and orals scripts.
• xAI Grok Business/Enterprise. Useful for fast ideation and long-context brainstorming or competitor snapshots in low-sensitivity scenarios. Has growing enterprise features. No federal authorization today.

Tool Pricing

When it comes to cost and scalability, most of today’s AI and proposal platforms can support companies of any size.

• Small businesses: $30 to $100 per seat/month for GenAI tools like Copilot, ChatGPT Enterprise, or Gemini.
• Mid-sized firms: Layer in automation and retrieval tools alongside pricing systems like ProPricer. Annual costs typically run in the mid–five figure range, but is often offset by time savings in proposal generation and compliance checks.
• Large companies: Enterprise-level environments combining multiple AI tools under a single governed architecture. These deployments can reach six figures annually but are justified by their ability to manage hundreds of pursuits, maintain security and auditability, and integrate with CRM or CLM systems.

Across all tiers, the best way to control cost and maximize ROI is to pilot one or two tools on a live opportunity before scaling across your organization.

 Trusting AI

Question: How can you ensure that if you trust an AI tool—that it does what you’ve asked it to and the information is valid?

Treat GenAI models as if they were the smartest junior analysts who always had an answer. You establish your goals and objectives, verify the work, but still question what you might get. It helps to start by telling the tool exactly what “good” looks like.

How to Build Confidence

• For detailed and complex tasks like a compliance matrix, define the scope and output format upfront. For example: “Capture only ‘shall’ and ‘must’ statements. Return a table with Req ID, text, source page, and factor. Quote the exact sentence used.” Clear direction reduces ambiguity and speeds review.
• After you verify the result, we recommend running a second check with a different model or a QA tool, then spot-check the sources yourself. Keep reference RFP excerpts to test prompts and catch drift.

Keep Humans in the Loop

Always have a person in the loop to validate accuracy, whether it’s the user who tried the prompt or a second resource to review and document excerptions.

• For data, sources, metrics, or figures, request and require sources every time. This includes asking the model to quote the clause, policy cite, or file and page it used.
• Treat anything without a source as unverified until a person checks it against the original document.

Pick the Right Model for the Job

• Claude Enterprise: Handles long, precise instructions and large document sets.
• Gemini for Google: Strong for environments and long-context research.
• ChatGPT Enterprise: Best for deep research and agent workflows.

 Up Next

In Part 2, AI in Action: How Growth Teams Are Using It to Win Smarter, we’ll cover how to use AI to speed opportunity identification, where to find proven prompt templates for capture work, how to gather customer intelligence, and how to integrate AI into gate reviews without losing decision quality.

The tools are here. The compliance frameworks are clear. The question is whether you’re set up to use them effectively or still treating AI as a science project.

If you want to move from experimenting with AI to operationalizing it across your growth lifecycle, Red Team can help. Our advisors can assess where AI fits into your BD, capture, proposal, and pricing processes, develop a compliant governance model, and build a practical roadmap your team can execute.