Stumbling blocks to avoid on the AI journey

Harkrishan Singh, Director, Application Development, at In2IT Technologies

As with any new technology, there are a number of potential stumbling blocks to avoid. Making the most of AI means minimising these pain points to maximise returns and benefits.

Data is at the heart of AI

AI would not be possible without data, and data is therefore typically the first stumbling block in a successful implementation. If AI tools do not have clean, consistent and complete data from which to draw insight, the insights they provide will be flawed or incomplete. There is also the potential for inherent biases in the data to assert themselves and grow more pronounced with time. Learning algorithms will reinforce these flaws and issues, resulting in AI tools that fail to perform as expected.

For completeness of insight, data needs to be collected from multiple sources and held in multiple formats. In addition, it should be accurately and clearly labelled for ease of access. Once data is collected, its security should also be a top priority, particularly with regard to numerous data privacy and protection legislations.

Potential pitfalls to avoid

AI offers many positive elements and benefits for business, but while embracing AI, enterprises must also be cognisant of the risks involved. For example, the long commute from storage can negatively affect the performance of AI systems, a challenge that will increase along with data volumes. As data grows from terabyte to petabyte and beyond, the time it takes to transport this data closer to compute resources and perform data processing and analytics takes longer, impeding the agility of the organisation. This needs to be factored in to planning to mitigate the impact.

There may also be challenges associated with ingesting, sorting, linking and properly using data. The volume of unstructured data from sources such as the web, social media, mobile devices, sensors and the Internet of Things (IoT) has increased dramatically. This in turn increases the potential for inadvertently using or revealing sensitive information hidden among anonymised data. Businesses could find themselves at risk of non-compliance if they are not aware of this issue and taking steps to mitigate it.

Another emerging issue is the potential for fraudsters to exploit seemingly non-sensitive marketing, health and financial data that companies collect to fuel AI systems. If security precautions are insufficient, customer data could be compromised, and individuals could fall victim to this fraud. In addition, technology and process issues across the entire operating landscape can negatively impact the performance of AI systems.

The key to success lies in the vision

AI is an expensive technology to implement, because it is new, inherently complex and because it requires specialist skills that are in short supply. When a business takes the decision to implement AI, all necessary steps should be taken to maximise the chance of success. One of the most crucial steps is having executive vision and sponsorship.

Once this is secured, business and IT need to work together to brainstorm potential AI project use cases, estimate potential value, assess feasibility, rank ideas and select a viable project to kick-off the program with a quick win pilot project. When these steps have been addressed, it is essential to provide access to tools and data to create and then deploy the models. A cross-functional project team is required to implement an AI solution

While there are a number of potential stumbling blocks surrounding AI, it is a powerful enabling technology that is creating entire new industries. Combined with key technologies such as IoT, big data analytics and/or Blockchain, and with economic growth, innovation and investment, AI will be the main factor in job creation. It is also a significant driver of innovation and competitive advantage, and businesses need to keep it on their radar to remain relevant in the coming years.