“If I have one more person talk to me about agentic AI and then realize they’re still using paper batch records, I’m going to scream.” That frustration captures the central tension in CDMO digital transformation today says Matt Lowe, Chief Strategy Officer at MasterControl.
Matt has spent nearly two decades at MasterControl, rising to CSO with responsibility spanning competitive intelligence, strategic partnerships, pricing, and product direction, making him one of the industry’s sharpest observers of where life sciences manufacturing technology stands and the distance it still needs to travel.
In the latest PharmaSource podcast episode, Matt maps the digital manufacturing journey contract development and manufacturing organizations (CDMOs) must take, from eliminating paper on the shop floor to achieving the industry’s most coveted goal: golden batch and real-time release.
The Digital Maturity Gap Is Wider Than CDMOs Admit
Survey data and conference room reality tell very different stories. In MasterControl’s own primary research, fewer than 1% of respondents said they were still running undigitized operations. But when Matt asks delegates at industry conferences how many still have paper on the shop floor, every hand goes up.
“Many CDMOs have digitized parts of their operations. ERP (enterprise resource planning) for instance. That ship sailed decades ago. But walk out onto the shop floor and logbooks are everywhere. Paper batch records, travellers, routers. Anything in a paper record, AI can’t get to.”
The problem is compounded by the CDMO model itself. Running high-variability, low-volume programs across multiple client products on shared equipment makes it structurally harder to adopt enterprise manufacturing solutions designed for captive production environments. “When you’re building five products that you sell yourself, it’s one thing. When you’re building five products times 100, for 100 different customers, it’s a whole different level of complexity.”
Download PharmaSource and MasterControl’s CDMO AI and Digital Maturity report
Three Stages: The CDMO Road to Autonomous Manufacturing
Matt maps CDMO AI adoption across three distinct stages, and most organizations are still at the start.
Stage 1: Get digital. Digitizing batch records and shop floor logs isn’t an innovation initiative. It’s table stakes. “It’s got to be digital. If we’re not there yet, let’s get there first.” Without this foundation, every conversation about AI is premature.
Stage 2: Operator co-pilot. Once data is accessible, AI can support operators in real time, monitoring the manufacturing process, identifying patterns that have historically led to scrap events or deviations, and flagging them before they happen. “We as humans can’t remember the last 100 batches we produced. The machine is very good at that.”
Stage 3: Lab in the loop. The longer-term vision connects quality control (QC), laboratory information management systems (LIMS), manufacturing execution, quality management systems (QMS), and asset management into a unified data environment where AI operates across the full process.
“The more data accessible across all the variables in the manufacturing process, the better AI will be able to operate.”
Full autonomy remains constrained by regulatory frameworks built around personal liability.
“How do you hold a machine liable? You can’t put it in jail. That’s at the heart of most of our regulatory frameworks, and regulation has a lot of catching up to do.”
Golden Batch and Real-Time Release: The Industry’s North Stars
When MasterControl runs its annual customer hackathons, where clients co-create and vote on product priorities in real time, two themes consistently dominate in the CDMO manufacturing segment: golden batch and real-time release.
“One kind of feeds the other. If you can arrive at the golden batch, you should be able to release it really fast. This is where you go beyond an operator co-pilot to asking: at what point do agents start making decisions and enacting them in the manufacturing process to achieve optimal yield and throughput?”
The regulatory constraints are real, but Matt sees the data infrastructure work happening now as what makes that future possible, and the CDMOs building it today as the ones best positioned to move when the framework catches up.
Making the Business Case: Start Low-Risk, High-Impact
For CDMO leadership teams building an internal case for AI investment, Matt’s advice is straightforward: don’t start with the most complex workflows.
“Use a risk-based approach. Identify low-risk but high-impact workflows where you can introduce automation and learn. If it goes wrong, it’s not going to be the end of the company. You get ROI data. Then you leverage that in higher-risk use cases with even greater impact.”
On implementation timelines, historically one of the biggest barriers to ROI justification, Matt points to MasterControl’s own client data. “We average 117 days to go live with a manufacturing execution system. Absolutely possible within three to six months.”
On Vibe Coding and the Build vs. Buy Question
As AI coding tools accelerate and SaaS valuations wobble, the question of whether CDMOs could build their own quality and manufacturing systems is becoming harder to dismiss. Matt takes it seriously, and dismisses it just as seriously.
“You can vibe code very simple systems. What people fail to realize about enterprise systems, especially regulated systems of record, is the complexity of the workflows, the interactions between them, the security requirements, the domain and regulatory expertise required to keep those things in compliance.”
His bottom line: “Do you want to be a software developer, or do you want to be a CDMO?”
Learn more about MasterControl on their website
