Flirt with failure; or drive strategic value with AI

ARTIFICIAL INTELLIGENCE (AI) is currently at a crossroads between hype and actual value-adding implementations. According to Gartner studies, while 92% of companies are increasing their AI investments, 74% of initiatives fail before they are ready for production. This discrepancy entails a number of fundamental problems, ranging from personal sensitivities, to operational feasibility, and the strategic viability of the projects. Critical errors made along the way are almost always due to poor methodology right at the start of development. 

The problem: most organisations treat AI implementation as a technological experiment and not as a strategic value driver. Unstructured approaches lead to the typical pitfalls: solutions looking for a problem, use cases without business objectives; and pilot projects becoming stranded in the “proof-of-concept desert” due to insufficient consideration of measurability and subsequent scalability. However, corporate practice shows things can be done differently.

The reason for this is down to the ability of many companies in identifying use cases that make a clear business impact, instead of utilising purely technology-driven solutions focused approach. Starting here enables them to eliminate poor investments at an early stage. It also incorporates a structured evaluation system that maintains, from the outset, the scalability of pilot projects in a way that moves them towards true industrial value creation long term. 

That in mind, what follows is a step-by-step approach that all manufacturers – especially those within the automotive and aerospace – can take towards building a successful AI business case, starting with the creative ideation and multi-stage quality gates, that is followed by the implementation and review of the proof of concept (PoC). Each phase serves as a strategic filter that checks for relevance, feasibility and scalability. Following a structured methodology overcomes typical AI traps. These include: avoiding data dilemmas by means of early-stage fitness checks, closing acceptance gaps with user focused co-creation and overcoming barriers to scaling with the aid of infrastructural roadmaps. 

Step 1: Use case ideation – Many organisations have a difficult time with knowing where to start with AI. To change this, manufacturers would do well to carry out a systematic pain-point analysis. This includes various explorative methods to identify appropriate business-relevant AI use cases for their organisation.  These use cases should offer clear economic value and strategic business relevance not just for the ‘now’, but for the future.

Step 2: Use case evaluation and prioritisation – This step usually involves considering multidimensional evaluation criteria (e.g. strategic fit, ROI, feasibility) to help with identifying the use cases that offer a high potential. At this stage, it’s also worth creating a value–effort matrix to ensure maximum transparency that supports ongoing decision-making around AI.

Step 3: Use case refinement – The next step is to develop a detailed catalogue of the various requirements that will make suggested use cases operable and feasible. This also helps with use case refinement as it identifies gaps between existing and required resources. 

Step 4: Proof of concept (PoC) development – Agile implementation is the next step. This enables organisations to transform their concepts into validated prototypes under real-life conditions with a focus on their performance and user acceptance.

Step 5: Use case review and scaling – The last step involves a final evaluation to review the use case and how to scale it – if that is appropriate for the organisation (e.g. manufacturer). All this is made possible by developing and tracking various kinds of quantitative KPIs and qualitative factors. This will then result in well-informed decision-making for AI projects. Organisations can then decide on scaling, optimisation or discontinuing a particular project.

Systematic frameworks for AI drive success

The five-step process from use case ideation to PoC review forms the systematic framework for ensuring successful identification, evaluation and implementation of AI use cases.

Additionally, following this kind of framework enables organisations to thoroughly and consistently evaluate AI applications and their quality at all stages. Following a structured approach also helps to create clearly defined handover points within organisations and enables companies to manage AI initiatives more transparently, effectively, efficiently, and to generate quantifiable business value. It means companies won’t flirt with failure as they deploy AI. Instead, the real success, comes from delivering projects that drive organisational value, and which are sustainable long-term – while also offering greater organisational agility. 

AI culture, AI translator, AI transformation

Further, for many organisations adopting AI means that they need to develop a sense of open-mindedness about AI, as it will likely require a fundamental transformation in terms of culture. However, if AI succeeds, this will be worth it. Additionally, a clearly communicated AI strategy has the potential to become an organisational compass. It can set and synchronise resource allocation, talent acquisition and innovation culture.

What is more, as AI strategies are deployed, it’s worth establishing ‘AI translator roles’ across the organisation. ‘Translators’ bridge the gap between data scientists and departments, while targeted upskill programs enable employees to actively shape the digital transition. In terms of process development, a balance between standardisation and flexibility needs to be achieved too. While the end-to-end process serves as a guiding principle for use case development, modular entry points and opt-in or opt-out scenarios facilitate tailored adaptations for the use of AI across organisations, which will be different for everyone.

For example, an engineering company with a developed data infrastructure will likely start immediately with advanced analytics use cases, while a logistics provider would probably initially invest in data lakes and literacy programs. This adaptability transforms the process from a rigid framework into a living organism that adapts to the maturity level of the company.

If you consider this further from a purely ‘technological perspective’, the coherent integration and efficient adaptation of various existing solutions becomes a critical enabler too. For instance, the seamless connection to ERP/MES systems via OPC UA interfaces; the use of established DevOps pipelines for MLOps; the embedded architecture of AI models in edge device clusters, and so on. These are all more than just technical details. They are vital requirements to be considered that deliver fast, predictable and viable success for companies.

Conclusion

Driving AI value across organisations requires a strong combination of culture, process and technology, and aligning it into a coherent value added chain. At this stage many organisations do not have the proven capabilities inhouse that enable them to drive AI value. This is where proven holistic digitalisation partners can support: from initial strategy workshops, through to developing industrialised AI operation models. They often have the expertise and experience in developing effective integrated methodologies that combine sector-specific domain knowledge with highly developed technological expertise – and always with the aim of establishing AI not as an isolated technology but as a driver of innovation within the system.

In a world where 74% of AI pilots never make production, a holistic approach towards AI deployments is becoming a critical competitive advantage for manufacturing and engineering organisations; and it can truly shift the needle for these companies, including within automotive and aerospace. 

With that in mind, the companies that succeed in understanding the five phases: not just as a linear process – but as an iterative learning cycle and that activate cultural, process-related and technological levers in a synchronised manner – they will be in a strong position to not only be AI users, but position their organisation as AI innovators that drive value and results.

Bodo Philipp is CEO at MHP Consulting UK

www.mhp.com