SAP Southeast Asia News Center

Headwinds Against Scaling of AI Projects

Digital data of particle wave and network connection on black background.

Artificial intelligence has penetrated many business models today. One of the most sought-after objectives was to automate manual processes into a digital mode to improve overall business performance. However, AI capabilities can achieve so much more. While many businesses have attempted to leverage AI to solve challenges beyond automation, there remain challenges that require a better understanding of AI workings to meet the needs of enterprise-wide scaling and integration.

Operationalizing AI – Why It Remains Challenging?

According to PWC research, AI will be a defining technology of this decade. Yet for many organisations, the benefits of AI remain elusive. Only 26 per cent of the 405 global executives surveyed by Harvard Business Review Analytic Services in December 2020 and who work at organisations that have an active interest in AI, including those that have initiated AI projects, have met most of their AI operationalisation goals within the stipulated timelines. Only five per cent of them completed the projects in full.

One of the most pertinent reasons why few companies have deployed AI with scalability in mind from the onset but ended up rolling it out as an experimental initiative is due to the lack of visibility on return on investment. Other challenges include accuracy in reporting, lack of accessible quality data and investment continuity in a phased AI project.

Another crucial factor lies in the project objective of AI implementation: is it solving a qualitative issue or improving the business bottom line? At Singapore Tech Week 2022, my colleague Manik Saha, Managing Director of SAP Labs Singapore, shared these sentiments with other AI practitioners during a panel discussion. The panellists unanimously agreed on the need to have a comprehensive understanding of the mission and purpose of AI and as such, have embedded intelligence into business models with the end goal in mind.

To realise the full potential of AI in an organisation, a “full adoption mindset” needs to be cultivated.  In the HBR article, it highlighted that 80 per cent of those who succeeded in fully rolling out their AI applications found it worthwhile at the end of the day.

Overcoming the Challenges 

Here are four ways enterprises can consider getting AI projects over the hump to benefit operations:

  1. Define a clear and standardised AI strategy to guide the process: A clear AI strategy is needed to guide the process in understanding the value that can potentially benefit the organisation and outline the steps needed to achieve the goal. A typical AI strategy should clearly state the problem statement, the impact of the solution and the reporting required by the stakeholders.
  2. Retire legacy systems and start improving interoperability internally: To redesign an existing business model, retiring legacy systems and improving interoperability across the technology infrastructure must remain a key priority in the endeavour. Only by overtaking the current IT applications with new systems, can businesses develop the right capabilities in using data, models, and software to address the problems detailed in the strategy. Project leaders will also be required to fundamentally prepare sufficient and quality data and start labelling them into useful assets to power the AI model.
  3. Data quality, quantity, and labelling – Another critical element involves hiring and developing specialists in data engineering, data science, data and AI ethics, model security, and machine learning engineering. They should work as a team to optimise the complex navigation around the interplay of software and hardware. Before acquiring the right capabilities to operationalise AI, it is also important to set up internal processes for end-to-end and top-down management of the project.
  4. Ensure scalability in the project scope – Finally, as with all projects, the owners need to ensure scalability in the project scope from the onset and work towards achieving the complete purpose. Not realising scalability after the first phase will lead to poor reporting, resulting in difficulty in securing future investments.

The Road Ahead

Operationalising a promise into a reality is what sums up a typical AI deployment. In the current decade, many businesses believe AI is key to their organization’s future, with  seventy-six per cent of the global executives surveyed by HBR stating that successfully deploying AI is critical to achieving their organisation’s strategic goals. This suggests that some organisational disconnect exists when it comes to conceptualising AI’s role, especially given how effective operationalising AI is when achieved and how important it is for companies going forward.

Today, we are seeing various industries deploying AI to improve their value chain to stay resilient in the fast-changing macroeconomic uncertainties. At SAP Labs Singapore, we are bringing in engineering talents to realise our vision of becoming the AI hub of the region, with an emphasis on embedded AI that runs directly on devices rather than in a data centre. AI has the potential to scale for huge business impact, and I am optimistic about its endless possibilities to drive innovations to better the way we work and live in the future.

This article was originally published on TransformLife.sg on December 6, 2022.


Nivedita Salam Mohapatra is Chief Operating Officer for SAP Labs Singapore.

Exit mobile version