AI Strategy: Focussing AI where your business needs it most
Note: This post was adapted from a version originally written for versor.com.au
Lots of businesses have tinkered with implementing AI and machine learning projects, some of which have actually come to fruition!
However, the norm, which is a disturbing trend, is that many data science-related projects either fail or do not provide any business value at all.
In a 2019 paper, Gartner predicted that “through 2022, only 15% of use cases leveraging AI techniques (such as ML and DNNs) and involving edge and IoT environments will be successful.”
In the same paper, they also predicted that “through 2022, only 20% of analytic insights will deliver business outcomes.”
This implies that only 20% of the insights generated by the 15% of successful AI projects are useful. Or in other words, that only 3% of the effort and cost expended in the process of running data science and AI projects produce a useful outcome!
A Forbes article early last year (Jan, 2020) claimed that less than 15% of firms have deployed AI capabilities into production, with the main reasons found to be limited available expertise, increased data complexity, siloed data, and lack of AI-development tools.
It seems that these reasons are just symptoms of a general inability of companies to pinpoint their priorities and develop a path towards development, implementation and maintenance of valuable machine learning and AI pipelines.
Companies really need to develop and then follow a comprehensive AI and data science strategy. Such a strategy should involve discussions with stakeholders at all levels of the enterprise – from analysts, floor managers, HR personnel, C-suite executives, middle managers and everyone in between. The aim of these discussions is to identify points of technical weakness all throughout the enterprise where improvements using automated and AI solutions may improve decision-making, business processes and customer satisfaction.
Once a comprehensive list of potential projects is identified, they need to be prioritised in order of potential value and ease and cost of implementation. Often the biggest gains will be made by relatively easy-to-implement projects that represent the low-hanging fruit.
The strategy should identify and set implementation timelines, so that the company has specific goals to work towards, with the more important projects aiming for earlier completion dates. Often the decision will be made to develop projects that will have more of an impact on customer satisfaction than internal employee efficiency, merely due to optics. This may or may not be the best decision strategically, but the decision to prioritise one project over the other really needs to be made dispassionately so that constant value is returned from the AI strategy over its 3-5 year time-frame.
The reasons cited previously for project failure are addressed by investing in both internal and external expertise. Internal staff who are subject matter experts need to be trained in advanced analytic techniques, so they become a point of communication with external consultants who are brought in to marry the data with the technology. The IT department will need to be across deployment and maintenance of the AI and machine learning pipelines. Whilst data scientists should be engaged to develop the solutions, working closely with the business to rapidly produce value according to the strategy that has been developed.
It is important to establish an AI and cognitive strategy creation process designed to ensure that while working intensively with you, a vendor will ensure all points of value in your business are pin-pointed and appropriately prioritised. Their expert analysis will provide a comprehensive strategy for building up the automation and machine learning capabilities in your business, ensuring that the most appropriate and cost-effective technologies are recommended. In this way you can be assured that your projects will end up as part of the 15% of those branded as a success!
