Quick answer: AI in pharma marketing has made promotional content faster to create, but it hasn't added capacity to the MLR review process. In The AI Efficiency Trap, a webinar from Vodori and Compliance Forward, Jason Cober of Compliance Forward and Annalise Ludtke of Vodori explained why faster content doesn't automatically mean faster approvals. The teams getting real value from AI measure the full content lifecycle, give AI tools the regulatory context they need, use AI to cut preventable rework, and start small before scaling.
AI has made content faster and cheaper to produce. Another draft, another variation, another localization. What it hasn't changed is the number of people available to review all of it.
That gap was the focus of The AI Efficiency Trap, a live webinar hosted by Vodori and Compliance Forward on September 14. Jason Cober is Senior Consultant, Regulatory Affairs, Ad Promo & Digital at Compliance Forward. He spent about 17 years at FDA, most of them in the Office of Prescription Drug Promotion (OPDP), where he led the office's AI development efforts from 2019 to 2025. He joined Annalise Ludtke, Director of Marketing at Vodori, to talk through what AI is actually doing to the MLR review process: where it helps, where it quietly moves work downstream, and what to have in place before scaling it.
Here are the key takeaways. To hear the full conversation, watch the session on demand.
Content teams are using AI to keep up with rising volume expectations, and Vodori sees it in its own data. In end-user surveying for its 2026 benchmarks report, about 30% of respondents said they're seeing an influx of AI-generated content. Review capacity, meanwhile, isn't growing to match.
Volume is only part of the problem. AI-generated content isn't always higher quality, so reviewers are handling more submissions and, sometimes, weaker ones. Jason's advice is for MLR teams to understand how that content enters the funnel in the first place. Much of today's variable and modular content is built from patient and HCP personas. Knowing what those profiles contain and how the AI applies them helps reviewers spot where it's likely to go wrong.
Annalise recommended tracking process metrics like first-pass acceptance and circulation counts. They're some of the clearest signals of submission quality, and they show where AI-generated content is costing time. If any of these terms are unfamiliar, our MLR terminology guide explains how each metric is defined.
Most AI business cases for content stop at creation. But every piece still has to go through the MLR review process, then approval and distribution. If faster drafting leads to more circulations and longer reviews, the savings never reach the market. AI content can also sound polished while hiding hallucinations, semantic drift or plain inaccuracies that only surface once reviewers get into the details.
The answer is to measure total lifecycle time, not just creation time, and to agree as a team on which KPIs matter most. For some organizations that's first-pass approval. For others, it's total review time.
Jason added that time savings may not show up where you expect. In his experience, AI often trims 10 to 15 minutes from a task, and people reinvest that time in higher-quality work rather than moving on to the next thing. KPIs should account for quality gains, not only time saved.
Reviewers now have to judge whether AI output can be trusted. Jason shared a simple two-question test for any AI tool. First, ask a fact-based question, such as listing the warnings and precautions in a product's label. If the tool can't answer, it likely doesn't have access to the right information. Second, ask an applied question that requires using that fact in a realistic scenario. If it fails there, the tool knows the information but can't apply it, which is a common cause of hallucinations.
The fix is context. General-purpose AI models are generalists, and much of what makes a reviewer effective is unwritten knowledge of how regulations, guidance, and enforcement letters have been applied. If it isn't written down, AI can't use it. Jason's starting point: list the five things you'd most want an AI pre-check to look for in a promotional piece, then test, refine and build from there.
Guardrails matter on the review side too. Annalise shared that on Vodori's recent system owner panel, customers noted that reviewers often approach AI-generated content through a different lens, and can be quick to reject it. To give content creators and reviewers a shared standard, some teams are putting guardrails in writing:
A few use cases stood out as practical starting points:
The message that came up again and again: start small, measure and improve over time. Jason framed AI as an operational transformation rather than a technology deployment. When email arrived, it replaced processes people already understood, like interoffice mail and fax. AI doesn't have that kind of analog, so its best uses aren't obvious. It takes patience and a disciplined approach to find where it delivers.
In practice, that means defining guardrails and business rules before full-scale rollout, asking reviewers for feedback along the way, and tying every pilot back to the metrics your team cares about. These steps build on the cornerstones of an effective MLR review process. It also means being realistic about scope. Jason recommended building guardrails before applying AI pre-checks to modular content, and noted that some presentations, like those consistent with the FDA-approved label but not drawn directly from it, are harder for AI to evaluate than others.
AI is increasing the volume of content entering the MLR review process, especially variable and personalized content.. Review capacity at most organizations is staying flat, so more submissions, and sometimes lower-quality ones, are reaching the same number of reviewers.
Not automatically. AI can speed up content creation, but if that content leads to more circulations or longer reviews, total time to market doesn't improve. Teams see the biggest gains when they measure the full content lifecycle and use AI to reduce preventable rework before content is submitted. Vodori's Content Preflight catches typos, formatting and branding issues before content reaches reviewers, so their time goes to the questions that need human judgment.
Test it two ways. Ask a fact-based question, such as the warnings and precautions in a product's label, to check whether the tool has access to the right information. Then ask an applied question that requires using that fact in a realistic scenario. A tool that passes the first test but fails the second has the knowledge but can't apply it, which is a common source of hallucinations.
An MLR pre-check runs before content is formally submitted for medical, legal and regulatory review. It flags mechanical issues like typos, formatting problems, broken links and incorrect brand elements so they're fixed before a reviewer sees them. Vodori's Content Preflight is built directly into the platform.
No. Evidence does not suggest that AI is at a level where it can reduce review headcount. Today, AI is most useful for surfacing issues, drafting options and organizing context. Approval decisions stay with the review team.
The AI Efficiency Trap covered more than a single post can hold, including audience Q&A on modular content and AI pre-checks. Watch the full session on demand to hear the complete conversation between Jason Cober and Annalise Ludtke.