Why are practical AI projects failing or taking longer than expected ?

It is an important to understand why AI projects across the world are slow to take off or do not yield the required results. The study one on thousands of exeutives by the authors o a HBR published study has attempted to find what are the reasons for that. It is found from the study that the problem for failures or very late realisation of success of AI projects is mainly due to human shortcomings and mis-understanding.

1. a very constrained and restricted approach instead of a cross organisational approach

2. from leadership-driven decision making instead of a data-driven decision making
3. from a risk-averse setup instead of an experimental and agile setup

I was going through the article, Building the AI powered organisation, by Fountaine, McCarthy and Saleh in HBR, Aug '19, click here for the paper,

Even though executives think AI projects are like plug-n-pay setups, actually it is not so. Only about 8% of all AI projects have given any semblance of meaningful returns yet.

There are many steps that could be initiated that could help address this issue quite amicably and effectively.

• First and foremost, get the workers to realise that AI is not meant to replace them but augment their existing skills and help them become better decision makes and efficient workers.
• The second approach would be to focus on initiatives that bring results within a matter of months than years.
• It does not benefit the industry to concentrate more in the central hub installation an less in the spokes. The applications need to be spread across installations as need arises. The decentralised arrangement was found to be successful in AI applications than the centralised hub and spoke model. Decentralising implementation at the spokes can be a better way to ensure success of AI initiatives.
• As we have seen and studied from digital companies across the past seven decades that have grown and matured, AI organisations also need to develop a unique style of development, planning and implementation, untested as of the present, to ensure AI projects do not get dragged and start giving their intended benefits at an early date.

New future AI applications will need difficult and new workflows, processes and culture at the workplace which we are yet to understand and define in clear terms. Experience and feedback from executives can contribute a lot to understand and navigate this area better.

George..