This article is an updated version of “How Pharma Can Win With RPA” originally published October 2023.
When we first wrote about RPA and its potential for pharmaceutical revenue management in 2023, bots that could log into state Medicaid portals and retrieve invoices without a human doing it manually felt like a real step forward.
That post aged quickly.
The underlying technology has moved fast enough that what felt cutting-edge two years ago is now a ceiling. The question for pharma revenue teams in 2026 is not whether to automate Medicaid invoice retrieval. It is whether the automation you are running is built for what comes next.
RPA Was the Right Answer. It Is No Longer.
RPA works well for processes that never change. Medicaid invoice retrieval is not one of them.
State portals update layouts. CMS modifies MDRP file format requirements. New programs come online. Email-based invoices sit alongside portal downloads. Two-factor authentication is now standard. Every one of those changes requires a human to intervene, update the bot, test it, and redeploy it. For manufacturers managing invoices across dozens of states and hundreds of programs, that maintenance burden compounds every quarter.
“RPA tools reliably automate defined sequences of actions within specific applications, but as stand-alone approaches, they struggle when processes involve exceptions, require decision-making, or need coordination across multiple systems.” — Automation Anywhere, 2026
That is not a flaw in any particular product. It is a structural limitation of rule-based automation in a dynamic environment. The industry is moving because the environment keeps moving.
Where the Industry Is Heading
Model N’s 2026 State of Revenue Report, based on responses from over 400 life sciences executives, found that 97% of companies now use AI in some capacity in their revenue management operations. Yet 75% still rely on two or more disconnected solutions, and manual processes persist across the board.
“The revenue leaders who embrace data integration, automation, and AI will be best positioned for success in 2026 and beyond.” — 2026 State of Revenue Report, Model N
Automation is widespread. Integration is not. That gap is where agentic systems do their best work.
Gartner predicts 40% of enterprise applications will include task-specific AI agents by end of 2026, up from less than 5% in 2025. The adoption curve is steep and the direction is clear.
What Agents Do That Bots Cannot
A bot executes a fixed script and breaks when reality diverges from it. An agent observes, decides, acts, and adjusts. The difference is not cosmetic.
In practice, for Medicaid invoice retrieval: when a portal changes its layout mid-cycle, an agent adapts and continues. When invoices arrive by email instead of portal, an agent handles both in the same workflow. When authentication requirements change, an agent adjusts. When something fails, it surfaces the issue in real time rather than failing silently.
The result is a system that monitors, retrieves, processes, and reports, with humans overseeing exceptions rather than executing every step.
A Word on the Hype
Gartner has also flagged that more than 40% of agentic AI projects could be canceled by end of 2027 due to escalating costs and unclear business value. They coined the term “agent washing” for vendors relabeling RPA and chatbots as agents without meaningful capability changes.
The test for Medicaid invoice retrieval is operational, not marketing: can the system handle portal changes without human intervention? Can it process email invoices in the same workflow as portal downloads? Does it provide real-time visibility into what ran, what failed, and why? Does it require a services team to stay functional?
Those questions cut through the noise quickly.
The Window Is Narrowing
If your team is still running manual or RPA-based Medicaid invoice retrieval, you are not behind the curve yet. But the Medicaid portal landscape keeps expanding, CMS keeps changing file format requirements, and quarterly cycles do not forgive systems that need two days to recover from a portal change.
The manufacturers building an agentic retrieval foundation now are not just solving a workflow problem. They are building the data layer that makes every downstream process, validation, reconciliation, dispute recovery, faster and more defensible.
The rest are still maintaining bots.
Sources:
Model N, 2026 State of Revenue Report, February 2026
Gartner, “Gartner Predicts 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026,” August 2025
Gartner, “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027,” June 2025
Automation Anywhere, “Agentic AI Platforms: 2026 Buyer’s Guide,” March 2026
