Implementing AI in an organisation is a complex task with many dependencies. What follows are the most common roadblocks.

Skills.

The people who create AI solutions are increasingly in demand – in the UK, for instance, there are more than two jobs available for every qualified person[i]. In the US, according to IBM annual demand for the fast-growing new roles of data scientist, data developers, and data engineers will reach nearly 700,000 openings by 2020. By 2020, the number of jobs for all US data professionals will increase by 364,000 openings to 2,720,000 according to IBM[ii].

[i] https://www.computerweekly.com/news/450427634/UK-at-risk-of-AI-skills-crisis

[ii] https://www-01.ibm.com/common/ssi/cgi-bin/ssialias?htmlfid=IML14576USEN&

Leadership.

According to PWC[i] “There’s no single technology leader in the enterprise anymore” because increasing amounts of IT spending happens outside the CTO’s budget. This can make it difficult to assign leadership responsibility for AI In addition, departments across organisations need to be in sync to think through data access and management, cybersecurity, regulatory compliance and other issues. And in organizations with many, well-established data and analytics teams, somebody in the ‘C-suite’ needs to bring their AI efforts together.

[i] https://www.pwc.com/us/en/advisory-services/digital-iq.html

Data.

Data is the lifeblood of AI. Without access to meaningful, clean and high-quality data, the organisation will not be able to move forward with AI projects. The organisation needs the capacity to capture, store, clean and prepare data before models can be fitted against the data to obtain meaningful predictive outcomes.

Infrastructure.

In most real-life scenarios, companies have accumulated an eco-system of IT systems over the years, most of which would not have been designed to access all data in the finest granularity in near-real time all the time. System and data engineering is therefore required to ingest, pre-process and connect the right data to relevant AI processes. Once AI predictions have been outputted, the results need to be fed into the operational layers of a business. Infrastructure that supports operationalised AI processes, has to run reliably and smoothly, and be fault-tolerant.