One Day at a Time: FDA’s New AI-Informed Inspection Pilot and What It Means for Industry

We have written on this blog about FDA’s modernization agenda from several angles lately—from the psychedelics Executive Order to the peptide compounding landscape (part 1, part 2)—but here is a storyline that cuts to the operational heart of regulated industry (or at least appears to): FDA has quietly been knocking on facility doors for single-day inspections, and this past Wednesday the Agency told the world about it.

On May 6, 2026, (now former) FDA Commissioner Marty Makary announced at the Food and Drug Law Institute’s (FDLI) annual conference that the Agency has launched a pilot program of one-day “inspectional assessments” for facilities that artificial intelligence (AI) has identified as lower-risk—both for domestic and overseas sites alike.  The announcement landed with a certain matter-of-factness that belied what is actually a meaningful shift in how FDA approaches inspection resource allocation.

Let’s unpack what we know, what we don’t, and what industry should be doing right now.

The Basics: What FDA Announced

The press release framed this as a “broader initiative to make its inspectional resources more targeted and efficient.”  The pilot is being conducted across multiple FDA inspectorates, including human and animal foods, biologics, medical products, and clinical research programs.  Facilities are selected using risk-based criteria: product type, prior inspection outcomes, and operational characteristics.

As of late April 2026, FDA had already completed approximately 46 of these one-day assessments.  So, what were the early returns?  Most one-day inspections resulted in “No Action Indicated” (NAI) outcomes—which is either reassuring about the quality of AI-based risk stratification, or a sign that, well, low-risk facilities really are low-risk.  Some assessments did extend beyond one day when investigators identified significant observations, which FDA described as built-in flexibility—not a flaw.  FDA investigators are not required to disclose on arrival that they’re planning for a one-day inspection, meaning firms may not know they’re receiving a one-day inspection until it’s done.

(Former) Commissioner Makary put it plainly at FDLI: “[I]f there are any safety issues on that one-day safety screening inspection, our inspectors reserve the right to extend it,” but for AI-identified low-risk facilities, “we are going to be doing more one-day inspections in the U.S. and abroad.”

The pilot will continue through the end of  fiscal year (FY) 2026, with FDA developing evaluation metrics focused on inspection duration, escalation rates, and the utility of findings in guiding risk-based decision-making.

One critical clarification from the outset: these assessments are not a replacement for standard FDA inspections, or for FDA’s existing alternative tools.  They are an additional, complementary tool.  However, unlike observations identified by FDA through the use of alternative tools, observations identified during a one-day inspection will be documented on an FDA Form 483.  One-day inspections are for facilities with a “strong compliance history” and are not intended for “complex or high-risk” products.  Higher-risk and more complex facilities and facilities with significant prior inspectional observations will continue to receive the full FDA inspectional treatment.

The AI Piece—and the Big Unanswered Question

Here is where things get interesting—and, candidly, where transparency remains thin.  FDA is using AI to identify which facilities are suitable candidates for one-day screenings.  (Former) Commissioner Makary confirmed as much at FDLI.  What he did not say—and what FDA has not published nor clarified—is which specific AI tools the Agency is using, what data inputs feed the risk-scoring model, or how the threshold between “low-risk” and “not low-risk” is determined.

This is important.  Facilities cannot optimize for a scoring system they cannot see.  And from a legal and due process standpoint, there is a reasonable question about transparency in how AI-driven risk stratification operates when it has direct regulatory consequences (i.e., whether your site receives a one-day screening vs. a multi-day full inspection).

On the same day as the announcement, FDA also released an update to its internal AI tool, Elsa—though the relationship between Elsa and the inspection risk-scoring function is not clearly articulated in public materials.

FDA Associate Commissioner for Inspections and Investigations Elizabeth Miller indicated that the agency is analyzing “trends in outcomes, risk signals, and investigator feedback” to refine its approach, and that data from the assessments—including compliance patterns, facility-specific risk scores, and discrepancies between registered and actual operations—will feed back into FDA’s broader risk models.  In other words, the AI gets smarter with every visit.  That is a meaningful feedback loop that industry should take seriously.

What Are Inspectors Actually Doing in a Single Day?

FDA has not released detailed inspection protocols for these assessments, so some inference and speculation are required.  Based on the program’s stated purpose and what has been reported, a few things are likely front-of-mind for investigators:

The registered vs. actual operations check. FDA explicitly flagged “discrepancies between registered and actual operations” as a risk signal feeding into site selection.  This is a fast, targeted confirmation: does the facility do what it says it does per its registration and listing?  If the answer is no—even on a one-day visit—that is the kind of observation that could convert a screening assessment into something longer.

Follow-up on prior 483 observations. If a facility had findings in its last inspection, investigators have reason to verify whether corrective actions were actually implemented.  That is a focused task that does not require starting from scratch on a full inspection.  This would presumably only be applicable if the prior inspectional observations are limited in both number and complexity, as only facilities with a “strong compliance history” are being targeted for one-day inspections.

A facility walkthrough and housekeeping. A visual assessment of cleanliness, equipment condition, and general facility status can identify obvious problems quickly.

Limited record review. Select batch records during a manufacturing inspection or Individual Case Safety Reports during a postmarketing surveillance inspection could be reviewed as representative examples of operations as a whole.

For context on what standard drug inspections entail: FDA’s internal inspector guidance provides that full drug manufacturing inspections must evaluate at least four of the manufacturer’s six systems—quality, facilities and equipment, materials, production, packaging and labeling, and laboratory controls.  Abbreviated site assessments must cover at least two.  One-day assessments, in the context of this pilot, are not formally classified against that framework, but investigators technically retain authority to examine anything relevant.

The Foreign Inspection Context

This pilot does not exist in a vacuum.  FDA has struggled for years with the logistical and resource challenges associated with foreign manufacturing facility inspections.  See our coverage of FDA’s anticipated expansion to the far east here.  The Agency has historically given foreign firms more advance notice of inspections than domestic manufacturers—a disparity that has drawn scrutiny from Congress and industry alike.  FDA announced plans to expand unannounced foreign inspections last year, but DOGE-related reductions in force complicated the logistics: inspector positions were not cut, but many support staff were—including the staff responsible for arranging foreign travel.

One-day assessments offer a partial workaround, assuming foreign facilities are eligible.  Planning to cover several one-day inspections  rather than a single multi-day full inspection would presumably minimize the chances that the trip will result in a “washout” in which the investigator is unable to complete an assigned inspection.  (Former) Commissioner Makary said it plainly: shorter screenings allow FDA to inspect more facilities overall, reserving deeper resources for sites that present greater safety or compliance risks.  “On a population level,” Makary said, this “increases the net level of safety.”  It is a resource optimization argument, and it is not unreasonable—at least theoretically.

What Should Industry Be Doing Right Now?

Nothing in this pilot changes the evidentiary standards FDA applies to compliance determinations.  But it does change the odds that any given facility will receive an inspection visit—and it raises the operational stakes for facilities that might have been lulled into comfort by long gaps between inspections.

A few practical takeaways:

Evaluate your facility’s prior inspection history. That record is one of the explicit criteria driving site selection.  If your facility has open 483 observations or prior Warning Letter history, it almost certainly affects your risk score—and your likelihood of receiving a visit.

Confirm that your registered operations mirror actual day-to-day activities.  FDA flagged operational discrepancies as a specific risk signal.  This is a quick compliance check that facilities should be doing regardless of any inspection program, but it bears renewed emphasis now.

Clearly document and evidence your corrective actions. If your facility had 483 observations in a prior inspection, thorough Corrective and Preventive Action (CAPA) documentation is the first line of defense when investigators arrive for a follow-up.

Treat a one-day screening as a potential on-ramp to a full inspection. A screening that surfaces significant observations will likely be extended.  That is a feature, not a bug, of this program’s design.  The safest assumption is that anything a standard inspection could catch is technically in play—the one-day format simply limits how far investigators can go before they decide to extend.

A Note on Modernization—and the Open Questions Ahead

(Former) Commissioner Makary consistently framed these initiatives as part of a broader “modernization agenda” for FDA.  One-day inspections, AI-informed risk stratification, and expanded remote assessments—these represent a genuine evolution in how the Agency intends to deploy its regulatory resources.  In principle, a risk-tiered inspection approach that concentrates resources on higher-risk facilities while providing lighter-touch feedback to compliant ones has intuitive appeal.

But intuitive appeal is not the same as validated, in-force, or long-term policy.  The pilot runs through the end of FY2026, and FDA is still developing its evaluation metrics.  The key questions that will determine whether this approach earns permanence are: Do one-day assessments reliably identify compliance gaps, or do they miss things that longer inspections would catch?  Is the AI risk model accurate enough to be the gatekeeper for which facilities receive lighter scrutiny?  And will FDA publish enough about the model’s inputs and logic to allow meaningful industry and public comment?

We will continue to monitor the development of this pilot and report back as evaluation metrics are published and the program evolves beyond FY2026.  In the meantime, if you have questions about your facility’s inspection readiness or compliance status, please do not hesitate to reach out.