Your Proposal Has Two Audiences Now đ¤Ż
GSAâs new AI rules and machine-first scoring just changed what âcompliantâ means.
Bidding in the Age of the Machine Evaluator
The agencyâs AI reads your proposal before a qualified human ever sees it. Build for both readers, or get cut before the people who could award you ever look.
Weâll Cover
Bottom Line Up Front â the one shift that changes how you write every section
đ Core Intel Report â why your proposal now has two readers, and how to win both [members only]
đ The Edge â the AI disclosure posture thatâs about to separate awardable firms from disqualified ones [members only]
Opportunity Alerts â where the AI money is actually moving this quarter
Signal vs Noise â the AI bidding advice to ignore before it costs you an award
đ The Play of the Week â five moves to make your next bid machine-readable and verified [members only]
This Weekâs Win â a real operator result, broken down
Bottom Line Up Front
Federal agencies have started using AI tools to screen proposals before a human evaluator engages. That means your bid now has two readers: a machine that filters, and a person who decides.
Most contractors are still writing only for the person. Theyâre getting cut by the machine first.
The move this week: stop using AI to write your proposal faster, and start using it to make your proposal survive the screen. Different goal, completely different output.
Week in Numbers
$13.4B â DODâs requested FY2026 budget for AI and autonomy, its first dedicated AI line. So-what: this is a standing category now, not a pilot. Position for multi-year work, not one-off task orders.
$800M â the Pentagonâs agentic-AI awards, split $200M each across four AI firms. So-what: the primes donât have cleared AI talent in-house. Thatâs a subcontracting lane, not just a headline.
$32B â DODâs committed AI, cloud, cyber, and data contract ceiling in the first half of FY2026. So-what: ceiling is committed; the work flows as task orders. If youâre not on the vehicle, youâre not in the room.
23% â the statutory floor for small-business prime dollars. So-what: the money is mandated. The new gate isnât access â itâs whether your bid clears the machine screen.
Figures: DOD FY2026 budget request and USASpending.gov FY2026 data. Budget lines are requests, not obligations.
Core Intel Report
Hereâs what changed, and almost nobody is pricing it in.
Agencies are no longer reading proposals cold. Source selection teams are running AI tools to deconstruct submissions, flag unaddressed requirements, and surface vague language before a human reads a word. Industry reporting this spring confirmed what proposal shops already suspected: if an AI tool catches a gap, the governmentâs tool catches it too.
So your proposal has two readers now. The machine reads first and filters. The human reads second and decides. Write only for the human and you lose before the decision ever happens.
This is the part most contractors get backwards. Theyâre buying AI to write proposals faster. Speed isnât the edge. The machine reader is a structure-and-compliance reader â it rewards proposals that map cleanly, name requirements explicitly, and leave no unaddressed gaps. A beautifully written narrative that doesnât map to the solicitation reads as non-responsive to a machine, no matter how good it sounds to a person.
Thereâs a second front opening at the same time. GSA released a draft clause this spring â GSAR 552.239-7001, âBasic Safeguarding of Artificial Intelligence Systemsâ â that would force any contractor delivering or using AI to disclose the tools, their provenance, and grant the government expansive data and use rights. Itâs headed toward a Schedule refresh. In 11+ years operating under a GSA prime, Iâve watched draft clauses like this go from âproposedâ to âflow-down requirement you signed last yearâ faster than most firms update their compliance binder.
And the risk side is no longer theoretical. In the GAOâs Raven case, a protester let AI generate content unverified â and it cited cases that donât exist. Thatâs the trap. The same tool that helps you bid faster will get your bid thrown out if you submit what it generates without checking it. One hallucinated past-performance reference, one invented certification, one fabricated metric, and youâre not just losing the award. Youâre a responsibility and misrepresentation problem.
What This Signals Next (analysis)
Expect machine-first evaluation to spread from large IT buys to civilian services procurements over the next two to four quarters. HHS, VA, and SSA are sitting on operational backlogs and growing AI budgets; theyâre the likely fast-followers.
Expect the GSA disclosure regime to harden. The draft is prescriptive on data rights and provenance. Firms that build their AI-use documentation now will answer a future representation in an afternoon. Firms that donât will scramble mid-solicitation.
And expect a quiet bifurcation. The contractors who treat AI as a compliance-and-structure engine will see win rates climb. The ones treating it as a ghostwriter will start losing on responsiveness and, eventually, on integrity findings.
This connects straight to the Velocity Framework. Proposal discipline is the muscle you build at the local and state level so that when youâre bidding federal â where the machine screen is sharpest â the habit is already automatic.
The Edge
Certification / Vehicle / Compliance Spotlight: AI-Use Disclosure & Provenance
This isnât a textbook clause. Itâs a posture, and right now itâs an edge.
Why it matters this quarter. GSAâs draft AI clause would require contractors to disclose every AI system used in performance, document its provenance, and accept broad government data rights. Itâs pointed at Schedule holders first, but the disclosure logic spreads. The firms that get there early look governable. The ones that wait look like risk.
Who should move on it. Anyone delivering AI, automation, or analytics â and anyone using AI to build deliverables for the government. Thatâs most of you now, whether youâve admitted it in a proposal or not. Who can wait: a pure staffing or construction firm with zero AI in the delivery chain. For now.
Where it pays off. GSA MAS holders first. Then DOD IT and any NAICS under 541511, 541512, 541519, and 518210 where AI is in the solution. If your past performance touches data or software, this is your lane.
The one mistake that kills it. Using AI in proposal prep or delivery and saying nothing â while FAR 3.104 still governs source-selection and proprietary information, and OCI rules still apply to tools trained on prior agency work. Silent AI use isnât a shortcut. Itâs an undisclosed liability sitting in your file.
The one move this week. Write a one-page internal AI-use statement: what tools you use, where in the workflow, what gets human verification before it ships. Youâre not publishing it. Youâre making sure that when an agency asks â and they will â you answer in minutes, not weeks.
Competitive Advantage Monitor. AI provenance documentation is becoming more valuable than contractors realize. Right now it reads as compliance overhead. Within two or three quarters, as OMBâs transparency preferences and the GSA clause mature, it becomes an evaluation discriminator â agencies will favor vendors who can show governance over vendors who merely claim capability. The cost of waiting is that youâll be assembling this under solicitation deadline pressure while a competitor drops a clean package into their proposal and scores the trust points you couldnât.
