How are companies distinguishing between “moonshot” automation and validated digital tools that can be deployed today?

Cell therapy is no longer a scientific question but a scaling challenge. These therapies have demonstrated durable, potentially curative outcomes with a single dose, and a clear regulatory pathway now exists. The remaining hurdle is how to reliably manufacture and deliver them at scale.

The industry is now adjusting its view on automation. The debate is no longer “moonshot vs. validated tools”; instead, it is about timing, integration, and regulatory risk. Fully autonomous, end-to-end platforms remain compelling in theory, but in practice they introduce a different challenge since changing platforms during development or post-approval can trigger complex comparability exercises for highly sensitive, patient-specific products. That risk is often underestimated.

At CTMC, we see automation as the critical enabler of scale, but only when deployed pragmatically. The near-term focus is on modular, audit-ready systems that reduce variability, improve data integrity, and integrate cleanly into GMP workflows without disrupting regulatory continuity. In parallel, we selectively invest in transformative automation where it can fundamentally reduce labor and increase throughput in a regulatory low-risk manner.

The leaders in cell therapy will not be pursuing full autonomy but instead be focused on scaling intelligently without creating new regulatory hurdles.

The distinction between a 'moonshot' and a 'deployment ready' tool often comes down to whether it’s looking to augment an existing human process, or trying to replace humans altogether.  “Moonshot” automations are more typically the latter. These are high-risk projects, often with high barriers to adoption that must overcome the inherent conservatism of a regulated environment to succeed. One way I’ve found to distinguish between the two is by asking: 'Does this technology transform my business in five years, or does it make my scientists’ lives easier in five minutes?' Paradoxically, the tools that answer the second question are often more transformative. Focusing on “easy to use” digital tools that help to streamline the human workflow today, helps ensure adoption and build the foundations of data integrity that are required to actually reach those moonshot automation goals tomorrow. 

Agentic AI can deliver insights in seconds rather than days, and in an industry where that speed can accelerate lifesaving therapies to patients, the potential is difficult to ignore. Moonshot automation has its place as an aspirational north star, but without the right foundation it carries a higher risk of failure, delayed value, and lost opportunity. The pragmatic path forward lies in what is deployable and defensible today.

That distinction begins with architecture. Many organizations are still grappling with AI bolted onto fractured data ecosystems, producing fragile results rather than validated ones. Effective AI requires a clinical data intelligence platform where agents operate within a single, coherent clinical workflow. The most reliable agentic systems combine deterministic logic for rule-bound, auditable decisions with generative models for pattern recognition and insight synthesis. That hybrid architecture is what makes AI both intelligent and trustworthy in a regulated environment.

Equally important is governance embedded at the core, not tacked on after deployment. Role-based privileges, complete traceability, and audit-ready records are not compliance overhead. They are what make AI outputs defensible against GCP requirements and regulatory scrutiny. The tools driving real breakthroughs today are grounded in unified data, governed from the start, and built on the combination of deterministic and generative intelligence. That is the line between moonshot and deployable.

Companies are distinguishing between moonshot automation and audit-ready tools by shifting from an artisanal pilot phase to an industrial omics factory. While early AI proved it could find signals in data, leaders now prioritize the foreman’s mindset, moving beyond the question of whether the prototype works to asking whether they can scale the line.

Nowhere is this factory mindset more critical than in precision health projects that fundamentally depend on the explosion of multiomic data. The sheer complexity, size, and sensitivity of omics data make it uniquely challenging. When you layer multimodal data on top of that, the IT and operational hurdles rise exponentially.

This is exactly where the distinction between orchestration and execution becomes stark. Moonshots often result in scrap on the floor, meaning AI outputs that are not reproducible, compliant, or scalable across an enterprise. In contrast, validated tools provide a universal floor, functioning as an intelligent orchestration layer like DNAnexus that tames this data complexity and allows various therapeutic areas to sit upon a shared, governed foundation.

To support multibillion-dollar compounds in this environment, enterprise-grade governance is non-negotiable. Winners in this space are taking a regulatory-first mindset, working from the end goal backwards. They require validated tools to ensure the 1,000th iteration of a pipeline is as accurate, predictable, and safe as the first. Ultimately, ensuring that any and all AI abides by these strict, audit-ready standards must be top of mind for anyone looking to combine the power of AI with the immense potential of multiomics data.

Amid the current volume of AI buzz, many companies have felt pressure to demonstrate that they are investing in AI so they do not appear to be falling behind their peers. This phenomenon is described by Gartner as the “Peak of Inflated Expectations” within the Gartner Hype Cycle. Such phases are common with emerging technologies and are typically characterized by heightened enthusiasm and experimentation. However, according to Gartner, the industry is already beginning to move beyond this stage toward a more pragmatic phase focused on identifying and delivering real value. Regulatory agencies and domain experts are also playing an important role in guiding organizations toward proven investment decisions, particularly in foundational technologies that strengthen operational performance, improve observability and controls, and support long-term digital transformation.

The best way to distinguish between a validated automation solution and one that's still a moonshot is to seek tools that are designed to build trust and scale. Moonshot solutions are generally exploratory and still seeking use cases, but tools that are already validated will be able to demonstrate how users can build trust, showcasing results from real-world deployments.

One way to spot the difference is to ask for a live demo that addresses specific use cases in a production-like scenario. This may sound obvious, but every time you're investing in a digital tool, you should demand a live demonstration that isn't pre-recorded and shows in practice how manual workflows can be automated. For example, if you were considering a network management platform, you would want to see evidence that this platform's conversational AI worked as advertised and that it could really automate manual jobs like filing trouble tickets or flagging network anomalies that could be a security concern. Secondly, ask the person giving the demo how they would solve for your typical problems using the tool, and have them guide you through the process. This allows you to see how intuitive the solution is and also how responsive.

Finally, look for solutions that clearly state they offer a human-in-the-loop management mode versus solutions that sit unmonitored within your system. The tools that are built to be examined and build trust are often the solutions that have already been validated and have nothing to hide. Tools that don't offer an easy way for human intervention can cause unintended consequences, and in a healthcare environment, risk isn't an option.

Near-term systems — such as workflow orchestration agents (e.g., OpenClaw, NemoClaw), LLMs layered over governed data lakes (e.g., OpenAI-based enterprise deployments), simulation acceleration, data science copilots, and engineering automation — are designed to operate within existing compliance frameworks (GxP, audit trails, version control). They are human-in-the-loop, bounded in scope, reproducible, and capable of delivering measurable ROI within 6–24 months. These tools augment decision-making rather than replace it, integrate into established SOPs, and generate traceable outputs suitable for inspection and audit.

In contrast, “moonshot” automation — such as end-to-end AI drug discovery, autonomous labs, foundation biological models, in silico clinical trials, or generalized AI scientists — targets high autonomy and transformative scientific impact but operates under greater regulatory ambiguity and biological uncertainty. These efforts often require new validation paradigms and long feedback cycles for ground-truth verification, with ROI timelines extending 5–10+ years and inherently higher variance in outcomes. As a result, many organizations are deliberately separating portfolios: operational AI for immediate productivity and compliance gains, and frontier AI for long-term strategic differentiation.

At Minaris, we’re focused on driving near-term efficiencies through digital system automation and workflow orchestration (e.g., deploying best in class native SaaS products with open APIs that drive end to end enterprise Integrations, linking data lakes, leveraging MS Power Applications, MS Fabric, bringing AI in-house, and building regular use case LLMs for operational efficiencies), while also deploying targeted test cases to explore novel concepts that could support future moonshot opportunities. The key is building capabilities today (in partnership with our technology vendors) that serve both short-term performance and long-term strategic objectives keeping quality and regulatory compliance at the center of our transformation(s).

Which way to the Moon? Agile innovation requires clear separation of operating modes. Moonshot technologies must iterate quickly outside the validated environment to mature and stabilize. When capabilities transition into GxP relevant use, changes must be managed through established change control, impact assessment, and re validation processes.

Some approaches are easier to validate than others. Often, automation suppliers aren’t used to the requirements of regulated environments. This causes friction at the interface between innovation and change control. Organizations can address this by assembling cross-functional teams that combine automation and compliance expertise.

In a moonshot world, the traditional waterfall process model doesn’t always work: You can’t put together a user requirements specification (URS) or functional specification if you are still exploring what your new technology is even capable of. Companies are recognizing that these approaches often lack the maturity, transparency, and compliance frameworks required for GxP use today.

In the GxP space, digital tools have always been designed with validation, traceability, and data integrity at their core. These solutions prioritize compliance with regulations such as 21 CFR Part 11 and Annex 11, ensuring that electronic records are secure, attributable, and reproducible. Rather than replacing human oversight, they augment it, by capturing structured data directly from instruments, enforcing standardized workflows, and maintaining detailed audit trails that stand up to inspection.

Focusing on deterministic behavior, clear logic, and configurable rules that can be documented and tested makes tools far easier to validate and qualify.

So, we need to consider how we can make solutions easier to validate. By adopting a pattern-based approach, we can reuse the same artefacts, as they always do the same thing. This leads to easier validation and reduces the impact of changes. We can then form one common validation blueprint to reapply, reducing the time to validate.

Ultimately, organizations are adopting a dual-track strategy: investing in innovation for the long term, while prioritizing proven, compliant technologies that can be implemented today to drive measurable operational and regulatory benefits.

Moonshot automation is often framed as an aspiration for end-to-end autonomy, but that perspective overlooks where the true breakthrough lies. Moonshot isn’t simply automating decisions or existing processes. It’s recognizing that a problem exists in the first place, especially when that problem could not be identified with traditional methods — and surfacing a solution that was previously out of reach by synthesizing vast amounts of data with real context.

This distinction is critical in high-risk, precision driven industries like biopharma and pharmaceutical logistics, where companies are increasingly reframing “moonshot” automation to focus less on making decisions faster, and more on navigating decisions dependent on too much data and contextual nuance to be reduced to a single linear workflow.

In these cases, AI paired with technologies like knowledge graphs act as decision amplifiers, surfacing relevant, connected context that strengthens and supports human judgment rather than removing people from the loop. AI doesn’t need to be fully autonomous to be moonshot level. Enabling better decisions in highly complex scenarios is impact enough.

By contrast, audit ready automation targets well understood, repeatable tasks to enhance productivity where it counts. What unites both concepts is the need for strong data hygiene and governance to trust data output.

“Moonshots” are difficult in biomanufacturing. The technologies that have broadly changed this field in the past — like the shift from paper documentation to electronic records, or stainless-steel equipment to single-use systems — have done so slowly and incrementally. One reason for this is that biomanufacturing processes are complex, and no two are alike, which has resulted in a wide variety of vendors supplying a wide variety of drug manufacturers, each with different standards and goals, such that no single organization has the power to drive innovation. Moreover, it is a conservative industry, and many vendors develop products only when there is a clear and growing market. It is difficult to drive innovation while seeking the assurance of an established market.

At Ensorcell, we anticipate that robotics and automation will be increasingly employed to improve the productivity of biomanufacturing facilities. A major barrier to this future is that bioprocess equipment is still designed mainly for human operators. Focusing on single-use systems, which have already proven to be efficient and flexible, we are designing compact, battery-powered equipment engineered to be user-friendly for human operators, but also to allow automated portability and integration with the autonomous factories of the future.

With increasing data volumes, advances in hardware for data processing and equipment control, as well as software for extracting, transforming, and inferring information, automation technology is rapidly advancing.

Digital transformation has been driven by efficiency through automation tools replacing standardized manual work. Examples include automating data input with online equipment interfaces, detecting anomalies with statistical analysis, and integrating system data. Optimizing business processes with laboratory information management systems, automating visual inspection with vision AI, and reducing repetitive tasks with robotic process automation tools have also been key.

As digitalization expands the scope of automation, human operators still make important decisions and orchestrate multiple workflows. However, recent advances in AI and robotics are enabling systems to handle these functions as well.

The difference between “moonshot” automation and conventional digital tools lies in the degree of human intervention and the purpose of the system. In contrast to the automation of individual tasks with human operators intervening to make key decisions, moonshot automation aims for autonomous, end-to-end capabilities. The system designs and performs the actions necessary to achieve its objectives, allowing human operators to intervene only in exceptional situations and evaluate the performance.

Regulated industries, pharmaceuticals especially, are navigating two fundamentally different classes of technology. On one side sit audit ready, validated systems designed for defensibility and repeatability. On the other, fast moving AI and automation tools optimized for speed and exploration. The real question is no longer whether to adopt these technologies, but which problems are they fit to solve. The answer has less to do with technical capability than with trust and liability.

When humans remain in the decision loop and the consequences of error are limited, AI can accelerate experimentation and insight generation. But where decisions carry personal legal responsibility, a Qualified Person under EU and UK medicines law, for example, the bar for validation rises dramatically. These systems operate in a different risk category entirely. Treating them as interchangeable is not merely a compliance risk; it’s a strategic misjudgment.

The industry is also moving on from multi year proof of concept cycles. Shorter, more targeted pilots have become the norm, and rightly so. In a landscape evolving this quickly, hesitation carries its own opportunity cost. At the same time, early hopes that AI could simply work around poor legacy data have largely faded. That workhorse foundation—validated ELNs, LIMS, and regulated data platforms—still matters. AI derives its value from those systems; it does not replace them. The moonshot sits above the mandate, not instead of it.

The remaining risk is drift. Tools introduced quickly in R&D can inch closer to regulated decision making without a corresponding reassessment of validation and accountability. Governance must evolve at the same pace as adoption.

Before any tool goes live, one question cuts through the complexity: who is accountable if this is wrong — and what happens to them?