Field Notes: Freeing Process Data For Automation At Kivi Bio AI Summit

By Todd Kapp, Kivi Bio

Across the life sciences industry, organizations are rapidly adopting automation, cloud infrastructure, and AI. Despite these investments, most labs and manufacturing environments remain constrained by a persistent and systemic barrier: data readiness.

An estimated 70%-80% of process data is still trapped in instruments, PDFs, spreadsheets, and legacy systems, forcing scientists into manual error‑prone workflows that introduce 3%-9% of errors under normal conditions and up to 18%-40% under stress.

In April, Kivi Bio hosted a conference at the Kenosha Innovation Neighborhood. The event brought together thought leaders from Kivi, Gilson, Wunderlich‑Malec, and Foundate Systems to examine the shared challenge and outline a path toward automated and AI‑enabled scientific operations.

Here are some of the key discussion highlights from the event.

Structural Disconnect Between Data Generation And Data Usability

We identified the core industry problem as a structural disconnect between data generation and data usability and reframed digital transformation around a simple premise: before we can even start talking about AI, we must address the data bottleneck.

Well-governed cloud infrastructure is the missing layer required to unify fragmented systems, eliminate manual reconciliation, and enable real-time decision-making.

We also highlighted some of the challenges when company leadership lacks knowledge of cloud and automation along with resistance to change by IT departments and others. For AI to be meaningfully deployed, leaders will have to buy in. Once they do, there's a practical, obvious maturity curve that happens in four phases.

Phase 1: Connect, consolidate, and clean the data.

Phase 2: Digitize workflows with a hybrid local/cloud architecture moving the cloud-ready data streams. Some data will remain on local drives at this phase.

Phase 3: Combine analytics with AI and digital twins to begin scaling.

Phase 4: Start predictive and semi-autonomous operations.

A robust cross-functional digital governance infrastructure reveals itself with fewer deviations, better yields, and faster tech transfers.

Islands Of Automation

We highlighted how even highly-instrumented environments — equipped with sensors, process analytical technology (PAT), supervisory control and data acquisition software (SCADA), manufacturing execution systems (MES), and enterprise resource planning (ERP) — still suffer from islands of automation and value leakage. Case studies demonstrated that bridging the integration layer is essential for accelerating batch release, predicting deviations, and reducing operational variance.

In one example, a company running validated processes at manufacturing sites in several countries had no automated data collection infrastructure. To solve it, they used servers and message queueing telemetry transport (MQTT), a lightweight protocol, to push data into a unified namespace (UNS) platform, which organized and contextualized the data.

When they turned it on, it gave leaders immediate visibility of all assets at their various plants from any device with a browser.

The case study underscores that automation alone is insufficient without end‑to‑end data connectivity.

Data Integrity And Decision Support

We addressed the problem from the bench-level perspective, focusing on operator fatigue, workforce shortages, and the persistent risk of manual errors. We showed how connected tools and digital traceability strengthen data integrity, improve reproducibility, and support regulatory compliance.

Sourcing automated tools is largely a committee-based task, and each stakeholder has their own list of priorities, which makes settling on a where to put investment first a daunting task. However, offloading even the smallest incremental steps to automation can reduce operator 

fatigue and deliver strong ROI. For companies just dipping their toes in, add automation starting with the most mundane and error-prone tasks.

Critically, key decisions must always happen with direct human oversight of automated tools. It's important to view AI as decision support. Even in fully automated labs, humans remain ever-present.

Capturing The Data You Need Without Errors

Data has two distinct types: process-related and analytical. Before we can build automated, end-to-end workflows and enable higher functionalities like visualization and modeling, it's important to understand how to manage both types.

Process data comes from manufacturing equipment, including bioreactors, chromatography skids, and fermenters. Rugged programmable logic controllers (PLCs) and OPC servers produce continuous, reliable process data.

Analytical data, on the other hand, is not standard like process data. It relies on manual transcription and is prone to error or omission. Scientists may remove data from failed experiments, which robs AI of the chance to digest what doesn't work. Beyond that, data streams for the same critical process parameters often use conflicting formats, which makes it difficult for multisite teams to know precisely what their counterparts see.

For example, scientists in different countries may be working on the same molecule but using different equipment to measure the same critical process parameters, like pH, lactate, and dissolved oxygen. To make the most of their data, companies must capture and structure data at the moment of creation, eliminating copy‑paste workflows and ensuring that the full data set is available for analytics and AI.

Data Readiness Determines AI Readiness And Automation

Together, these perspectives converged on a unified conclusion: data readiness determines AI readiness and automation. The session provided a practical framework for modernizing scientific operations through cloud‑connected systems, automated data capture, and integrated workflows — laying the foundation for faster, more reliable, and more intelligent decision‑making across the life sciences ecosystem.

About The Author:

Todd Kapp is the founder and chief executive officer of Kivi Bio, a life sciences consulting company in Kenosha, Wisconsin. He received an undergraduate degree in chemical engineering from the University of Connecticut and an MBA from Loyola University. Connect with him on LinkedIn.