Latest Blog

2019: The Year of Successful AI Implementation

29 March 2019 | Kate Perevoshchikova

2018 was the year of the artificial intelligence (AI) frenzy. An avalanche of articles, blogs and conferences sought to weigh up the benefits and perils of AI, which led to a media hype that often distracted from the real-life uses and benefits the technology can bring. If we look beyond this hype and focus on implementation, many companies still fail to effectively deploy AI technology and extract business value from it.

According to a McKinsey & Company survey published in 2018, around 50% of companies worldwide have embedded AI capabilities in at least one function or business unit. However, the same survey notes that ‘most had not yet adopted the complementary practices necessary to capture value from AI at scale’. Put simply, companies are spending time and money to adopt these technologies but are not implementing them due to a lack of existing infrastructure in place.

What is behind this AI procrastination? I have investigated a number of reasons which are costing organisations a key competitive advantage:

1. Failure to Identify the Pain Points

In order to create a competitive advantage using AI, the firm has to consider the question: ‘What area of my work needs improving?’ It should be a daily pain point which the company dedicates money and human resources towards, so that the solution is felt widely enough to drive further adoption of AI. As AI strategist Ankur Agarwal put it: "Capabilities must be secondary to opportunities". Therefore, business leaders should seek the most impactful AI use cases for their companies to springboard future use.

2. Long implementation periods

Companies have grown used to implementation periods of at least a year through using legacy technologies that require huge amounts of time and resources in order to function. This limits an organisation’s openness to new tools on the market. Businesses need technology that immediately fits into their workflow. For example, Luminance provides significant time savings by presenting data in a visual dashboard from day one, without the need for training or configuration of the system to produce results. Luminance is true machine learning and therefore eliminates the need for training and constant configuring of the system.

3. Needs to come from the top

As Andrew Ng, the prominent AI expert, pointed out in the AI Transformation Playbook, one of the best strategies for adopting AI is to “develop company-wide platforms that are useful for multiple divisions [or] units”. In large organisations, technological solutions often tend to be consigned to small parts of the business. An early adopter of Luminance pointed out that people move around within an organisation frequently, so it is essential to implement technology that is flexible enough be used by multiple teams. Luminance works seamlessly regardless of language, jurisdiction or discipline and consistently proves value in new investigations and data sets it has never been exposed to before. Keeping AI locked within a certain department as one person’s side project means failing to capitalise on the tremendous value it can bring. Having a dedicated AI task force and a clear, top-down AI strategy will ensure the benefits of the technology can be felt across an entire organisation.

4. Failure to Manage their Own Data

It is common knowledge that data is the fuel of AI systems. Google and Facebook have accumulated an extensive amount of data through their users. Smaller companies can harness their data in a similar way through the use of AI as they often possess rich repositories of business data which are niche and specific to the area in which they compete. One example is the beer brewery analytics group, Weissbeerger, who have installed measuring equipment in taverns and bars to track their beer drinkers’ behaviour patterns – this information is too specific to be useful for the vast majority of organisations but is incredibly valuable if used properly by that company or their competitors.

By shifting from a focus on ‘Big Data’ to ‘Smart Data’, companies can build in-house ‘enterprise AI’ and train their own models on relatively small data sets. The more differentiated and specialised these models are, the more competitive advantage they can provide to the company. However, many companies fail to comprehend the hidden value of their data.

To overcome the data challenge, companies need first to create a unified database across the whole company. As Andrew Ng argues, it is very difficult to build efficient AI models if you store your data in 50 different databases in 50 different departments. Secondly, companies need to separate the highly valuable data from the low-value information. Bad inputs produce bad outputs and no AI technology can magically transform bad data into something constructive. To efficiently access and maximise the benefits of AI, data management should be recognised as a high priority across an organisation.

5. Lack of AI Education and Training

Significant technological shifts often bring uncertainty. People tend to wait until the technology is ‘good enough’ or has sufficiently penetrated the market, meanwhile relying on outdated methods and processes that move the organisation further and further behind. In their reluctance, they fail to capitalise on innovation and fall behind the savvy early adopters.

People often lack awareness of what technology can do for their organisation. Increasing awareness of the AI tools available across the whole company by providing user training and showcasing the platform is an invaluable way of driving adoption and excitement around new technology. Importantly, key leaders within the company need to understand AI and champion AI strategies, as a key part of the overall vision and goals for the organisation.

Andrew Ng recommends a minimum of 4 hours of AI education for executives and senior and 12 hours for mid-level managers to enable successful implementation of the technology. This is not just ‘how to’ sessions and demos, but broader planning and strategising, as well as attempts to shift reluctant attitudes to new technology. Sometimes implementation requires the tweaking of existing business processes but these small changes unlock countless potential benefits.

For many organisations, successful AI adoption means staying competitive in the market. I have identified several bottlenecks that prevent a fully-fledged transformation of business through AI and these stumbling blocks requires courage and business acumen to spot them within your organisation. However, successfully adopting AI holds the potential for an exceptional boost in productivity, job satisfaction and the ability to outpace the competition.