AI is predicted to add up to $15.7 trillion by 2030, but three main aspects prevent successful adoption within companies.

Delivering a keynote speech at the recent Big Data LDN conference, IBM’s general manager of Data and Watson AI, Rob Thomas, discussed three main factors that often befall plans to implement AI.

Describing AI’s potential as “the greatest opportunity we’ll ever see in our lifetimes”, Thomas suggested the following:

1. A lack of data literacy

The potential that AI has is by no means lost on business leaders. According to a survey by MIT, 85% see AI as an opportunity to gain value from company data.

However, decision-makers often struggle to get the best out of AI due to uncertainty surrounding how it would fit into business practices.

According to that same MIT study, 81% of business leaders are unable to grasp the data needed for AI development.

“There is no AI without an IA, information architecture,” Thomas added.

Digital transformation requires robust IT architecture planning

Emerging technology and radically new models for IT delivery are making architectural planning more important than ever before. Read here

The IBM GM went on to introduce the steps that a company can make towards implementing AI successfully:

1. Modernise: Company data needs to be housed in a multi-cloud environment. This would fulfil the idea of an Information Architecture that can give “choice and flexibility” to users and reduce the need for constant assembly.
2. Collect: Data then needs to be collected in order to establish a base from which it can be simply accessed.
3. Organise: A foundation for data is nothing without a corresponding foundation for analytics, which also must be organised through integration, cleansing, filing and then governing.
4. Analyse: Once organised, data must then be scaled in order to develop advanced analytics and AI model management.
5. Infuse: The final step involves implementing established AI models into company practices, including trusting AI to make decisions, explain those decisions and to detect bias.

How AI can improve quality assurance: six tips

Quality assurance (QA) is an essential area of concern to companies. If the public thinks a company or service does not care about keeping its quality level high, they’ll likely do business elsewhere. AI for quality assurance can help. Read here

2. A lack of trust

A lack of trust is a significant obstacle for the implementation of AI.  A study by KPMG found that just 35% of business decision-makers trusted the use of (AI) and analytics within their company.

To amend this, three steps were suggested:

1. Gain understanding of the data’s origins.
2. Learn how AI models make their decisions.
3. Identify model flaws, e.g. bias, anomalies.

3. An ever present skills gap

According to Thomas, another AI adoption pitfall is a supply and demand mismatch when it comes to data skills.

He stated that while there is a demand for data scientists, “supply is not where it needs to be”, with a skills gap of 200,000-500,000 strong being present in this field.

Has the cloud skills gap become a cloud skills crisis?

Due to the rise of automation and artificial intelligence, many believe that human insight is less important, but is this the case? With the cloud now underpinning business transformation, thanks to a growing technology skills gap, organisations are struggling to stay ahead of the curve. Read here

Automatic AI

Automatic AI was touted as a potential solution to this skills gap — this way of working can reduce trial and error from weeks and months to a matter of hours.

Putting forward Watson’s AutoAI as an example, Thomas said that automatic AI can undergo human machine learning and data engineering duties within minutes. The software can sift through many algorithms to select the correct strategy (or strategies) to use for a certain task. Further data is generated in order to provide insights into the program’s decision logic, and the program also allows for hyper-parameter optimisation.