Contractors, take note. The government has been quietly, but very clearly, testing tools that could transform how past performance is evaluated in federal contracting. If you think CPARS narratives are just paperwork to be completed, that could change. Here’s a look into the recent Department of Homeland Security (DHS) AI-based past performance pilot program, and what you can do now to prepare for the future of past performance evaluations.

What is CREPES?

The DHS Procurement Innovation Lab (PIL), in partnership with firms like CORMAC, advanced a platform called CREPES (CORMAC’s Envisioning and Prediction Enhancing System). The goal was for the system to operate as a production software service to support evaluators across agencies. It used machine learning and natural language processing to compare solicitation requirements with past performance records in CPARS, pull out relevant narratives, compute relevance scores, and help contracting officers find the best matching projects.

As of mid-2024, public reporting confirmed that DHS’s CREPES pilot had ended. Without finalized updates to the Federal Acquisition Regulation (FAR), the adoption of AI-assisted past performance evaluations is likely on pause, but the CREPES pilot demonstrated the concept’s viability and makes it plausible that similar tools will be used in the future once policy and data access align.

Why CREPES Matters

Tools like CREPES can process CPARS narrative data and ratings far faster than humans. They can compare past projects by technical scope, performance, and dollar value, and reduce subjectivity. Contractors with inconsistent or vague narrative data will be at a disadvantage. With AI-assisted relevance scoring, evaluation rationales may become more audit friendly, but only if records are structured and narratives are clear. As policy catches up, agencies will move toward formats that reward data that is consistent, quantified, and easy to evaluate.

What Contractors Need to Do

Act now. Although the DHS pilot ended, the possibility of using AI in past performance evaluations is on the horizon. Here are four ways that you can prepare now:

• Start with a focused audit of CPARS and past performance records. Read your narratives. Are they full of vague adjectives like “performed well,” or do they include specific outcomes like “delivered three months early” or “met 98 percent of milestones?” Standardize terminology so AI parsing does not miss key indicators. Ensure records are complete and clearly linked to solicitations. Document scope, period of performance, contract size, and the performance markers that show delivery. If there were modifications or unusual events, document and explain them so automated systems and human evaluators do not misinterpret the context.
• Collect and highlight measurable data on every delivery. Track cost savings, schedule changes, quality metrics, and customer satisfaction, and embed those numbers in your performance reports. Keep internal evidence that supports your record, even if it never makes it into CPARS. This allows you to add context later if an AI tool surfaces only a portion of your narrative as relevant to a solicitation.
• Stay alert to policy updates. Monitor FAR Part 42 and related acquisition regulations. When new rules are published, review them immediately to understand the implications for AI assisted evaluations. Where it makes sense, engage through industry comments to shape how narrative data is treated.
• Train proposal and capture teams on the new expectations. Assume that evaluations can be scanned by AI for relevance and signal strength, not just tone.

The Bottom Line 

AI is not yet grading your past performance on a wide scale, but the technology is proven, and policy will eventually catch up. For Chief Growth Officers and capture executives, CREPES is a peek into where federal evaluations are headed. Past performance has always been a record of delivery, and now that record has the potential of being clarified, quantified, and compared at machine speed.

Treating CPARS like routine compliance can potentially leave you behind. If you treat it as a strategic asset that is curated, validated, and communicated with the same rigor as capture intelligence, you will be ready when AI-assisted evaluation becomes standard practice.

Want to know what you can do now? Check out Neal Levene’s companion article, Five Steps to Stronger CPARS Ratings Today, for practical actions contractors can take to strengthen CPARS and prepare for the AI-enabled future.