Industrial Sector

Navigating GenAI in manufacturing: a CTO's guide

How long will it take for generative AI to become a game-changer in manufacturing? Discover real-life insights and strategies to learn how CTOs and C-Level executives can balance innovation with practicality, overcome cybersecurity concerns, and optimise efficiency as well as costs.

We're also planning an "Applied GenAI Masterclass for industrial CTOs". Don't miss to register here!

9 minutes to read
With insights from...

The pace of technological advances today is nothing short of breathtaking. Among these advancements, Generative AI (GenAI) stands out as a transformative force with the potential to redefine industries. For companies producing components, machines, or devices staying ahead of such innovations is not just a competitive edge – it's a necessity. This blog aims to provide a roadmap for CTOs and other C-Level executives to navigate the complexities of GenAI integration, addressing both internal and external challenges while offering practical, phased approaches. 

Managing a delicate balance and distinguishing genuine advancements from hype

As a CTO, you're at the forefront of a technological tug-of-war. Almost every day brings promising innovations that could revolutionise your industry, your organisation or your role – but there’s also the challenge of distinguishing genuine advancements from fleeting trends. Your mission is to allocate limited resources – both budget and talent – towards technologies that will deliver concrete benefits, not just hype.

The pressure is twofold: stay ahead of the curve while ensuring current systems run smoothly. This balancing act is further complicated by the need to align your technical vision with the overall business strategy. You must convince stakeholders of the value in new technologies, requiring not just technical expertise, but strategic acumen.

The stakes are high. Investing in the wrong technology could mean wasted money and resources or missing out on key advancements, finally leading to a significant competitive disadvantage. Your role demands a delicate balance: being innovative yet pragmatic, forward-thinking yet grounded in operational realities. In this landscape, making informed decisions about extremely fast emerging technologies like GenAI is as mission critical as difficult.

Overcoming cybersecurity concerns and breaking down resistance to change

In the industrial sector, where operational efficiency is key, adopting GenAI presents a unique set of challenges. The need for highly optimised systems and a high degree of automation often leaves little room for exploration, creating a tension between maintaining and improving current performance and embracing emerging technologies.

Not easing the challenge are corporate IT departments, which are essential for maintaining security and stability but can inadvertently become blockers of transformation. Their valid concerns about cybersecurity, the integrity of operational systems, and complex dependencies often lead to a risk-averse stance. This caution, though understandable, will slow down the exploration and adoption of GenAI technologies, becoming a risk in itself.

  • 'If we use internal IT, we can forget about making our AI project a success!'

    Vice President Product Group, Machinery Group in the DACH region

    Vice President Product Group,
    Machinery Group in the DACH region

Moreover, people working within highly operational setups are often resistant to change. Years of refining processes and achieving continuous improvements create a culture where predictability is prised above all else. Introducing GenAI into this environment can be met with scepticism or outright resistance and can pose its own risks. 

It’s crucial to recognise that new technologies like GenAI start with lower maturity levels. They require time, experimentation, and refinement to become truly efficient and effective within specific industrial contexts. This reality clashes with the expectation of immediate, measurable results that many decision-makers hold.

Zühlke Dominic Böni
' If you want to create the ‘next big thing’ with GenAI, you must be willing to go through a painful change and to invest on a long-term basis. '
Dominic Böni
Principal Consultant Zühlke Group

In the race for innovation, leaders often call for rapid adoption of technologies like GenAI, seeing them as a fast track to growth. However, this enthusiasm rarely takes into account the complex realities of implementation, such as vendor lock-in and the explosion of tools. This disconnect between calling for the ‘next big thing’ and underestimating the investment, time, and organisational change required creates a precarious situation. Unravelling this situation remains a critical challenge for industry leaders.

Real-life challenges and insights: rethinking the proof-of-concept approach

In the field of technological innovation, particularly with GenAI, many businesses eagerly jump on the bandwagon, starting with what they believe to be industry best practice: the proof-of-concept (POC). However, our observations in the field reveal a startling trend. Despite significant investments in POCs, a considerable number of these initiatives never transition into operational reality. This phenomenon begs the question: Why do so many promising POCs fail to gain traction?

Upon closer examination, we've identified three key factors contributing to this disconnect between POC success and real-world implementation:

Over-engineered technical focus:

Many POCs suffer from an overly technical approach, essentially transforming them into elaborate feasibility studies. While demonstrating technical capability is crucial, this narrow focus often overlooks critical real-world factors. User adoption, practical application in day-to-day operations, and the human behavioural aspects of technology integration are frequently sidelined. Consequently, these POCs, while technically impressive, fail to address the broader spectrum of challenges that arise in actual implementation.

The scale and quality paradox

Industrial companies, accustomed to operating mature, high-quality processes at scale, often apply a similar mindset to POCs. This approach leads to unnecessarily expensive and over-engineered proofs-of-concept. More problematically, when this mindset is applied for the full implementation, it results in a daunting financial prospect of moving from POC to real-world application, often deterring further progress.

Neglecting user adoption and process impact

Perhaps the most critical oversight in many POCs is the failure to adequately address user adoption and the technology's impact on overall processes. These factors represent the most significant risks and potential benefits of any new technology implementation. With some creativity in process redesign, the technical and quality risks could often be reduced drastically. However, by not sufficiently proving these aspects, POCs leave the most crucial questions unanswered, adding substantial uncertainty to any decision on moving forward.

The combination of these three factors – over-engineering, high costs, and unaddressed key risks – creates a perfect storm. The result is an expensive, high-risk investment proposition that's rarely attractive in a corporate environment where resources are finite and ROI is closely scrutinised.

A practical approach is key to successful GenAI integration

Our experience has shown that a practical, down-to-earth approach is crucial for successful integration, especially in manufacturing where the fundamental value is still the physical product and not data. But, there is a quick win after all. We have observed various companies where employees started to develop GenAI solutions on their own with significant business impact, proving that there are several low hanging fruits to be plucked, as a CTO of an industrial devices manufacturer proves, ‘Our field office developed a solution based on Custom GPTs which is already very successful’. However, these solutions often conflict with data privacy guidelines. To solve this, consider building a platform that allows quick configuration of single agent setups in a controlled environment, leveraging enterprise-grade LLMs from major cloud providers to ensure data security.

""

So, here's how companies can navigate this complex terrain effectively:

Avoid the POC trap

rather than investing in extensive, costly POCs, focus on clear, simple use cases that can provide immediate value. Start with small parts of larger processes where GenAI can make a tangible difference without requiring perfect accuracy or full automation.

“We have about ten thousand products. We need a guide for our field offices to support them in product selection”.
CTO, Component Manufacturer in the DACH Region

Understand and simplify the beginning

by thoroughly understanding the current issues in your processes. Look for opportunities where GenAI can support human decision-making rather than completely replacing it. For instance, using GenAI to guide newer sales and customer support employees to relevant documents or suggest options can be immensely helpful and easier to implement than full automation.

Start small, scale smart

the industrial sector is ripe with small, quick-to-implement use cases that can provide significant value when designed well. These lower-value but easier-to-implement applications serve as excellent starting points, allowing for quick wins and learning opportunities. However, since these use cases alone often don’t justify a full development project, it is important to set up a more scalable platform approach. This enables fast and cost-effective development of these use cases, which can then often be configured rather than requiring a full development effort. The more complex cases can follow later.

Prioritise operational integration

instead of getting bogged down in technological perfection, focus on getting GenAI into manufacturing operations as quickly as possible. Simplify everything - the solution just needs to provide value to the user and improve overall results. This approach addresses the critical risks of user adoption and system integration more effectively than non-operational POCs.

Manage complexities and risks fully

because integrating AI into operational systems carries significant risks, often slowing down adoption in corporate environments. To mitigate this, consider an isolated, separate setup not just for development and testing but also for operational use. Surprisingly, many use cases can function effectively without real-time access to operational data. When necessary, expose specific data and access through APIs, limiting risk while maintaining development speed. This approach allows for rapid development and deployment without compromising on security or compliance.

Steps to start: 

  1. Understand the process and where the actual challenges are.
  2. Pick a small part of the process which can be improved by guiding and supporting the human tasks.
  3. Make this small part operational instead of building a fully-fledged POC.
  4. Create a small, pragmatic platform that can provide good levels of data security and replicate the above process quickly.

Pro tip: Disconnect your platform from operational IT systems. This will reduce security complexity, increasing speed and reducing costs dramatically. Real-time connectors can be added later.

By adopting this practical approach, companies can navigate the complexities of GenAI integration more effectively. It allows for rapid learning, immediate value creation, and the flexibility to scale and refine solutions as they prove their worth in real-world applications.

' Don’t forget, once you have a valuable operational use case, no matter how simple and crude, you can always improve it later. But a shiny POC that never becomes operational is worth nothing. '
Dominic Böni
Principal Consultant Zühlke Group

No matter which hyperscaler you prefer, you get top-notch LLMs from them:

  • Azure – OpenAI: GPT4o, GPT4o mini
  • Azure – OpenSouce: Phi3 family, Llama 3 Family, others…
  • AWS – Antrhopic: Claude 3.5 Sonnet & Claude 3 Family
  • Google – Gemini 1.5 Pro, Gemini 1.5 Flash

GenAI is here to stay, what can companies do to leverage it effectively?

Generative AI, despite its current limitations, is already demonstrating significant value across various industries. It's crucial to recognise that what we're experiencing now is just the beginning - we're working with the least advanced version of GenAI we'll ever see. Yet, even at this stage, its impact is undeniable. 

We can expect improvements like more sophisticated models, seamless system integrations, the emergence of complex agent-based systems and GenAI-powered hardware that will dramatically reduce computational costs, increase speed, and enable local processing. Given this trajectory, it's clear that GenAI is far more than a passing trend. The pace of change is rapid, and its influence on businesses will only grow.  

Moving beyond PowerPoint presentations and theoretical discussions, it's time to trigger and drive real change within your organisation. The key lies in practical, hands-on experience - experimenting with use cases, learning from real-world applications, and gradually integrating GenAI into your operations.

Masterclass: Applied GenAI for industrial CTOs

We are planning an "Applied GenAI Masterclass for industrial CTOs", a masterclass designed specifically for CTOs and CEOs.

This workshop will help you evaluate and build a concrete use case, providing you with hands-on experience and practical insights.

If you want to receive updates about the upcoming masterclasses, register via the form below and we will reach out to you with all the details:

""