We are already one quarter into 2018, and many organisations and professionals are working hard to meet their innovation targets for this year. Adoption of AI can be a key contributor to achieving innovation. Whether your aim is to bring high degrees of automation through AI, or use AI as a self-standing service internally or externally with your clients, there are a number of challenges that need to be handled appropriately. In this blog we share some of the hurdles we have encountered. We will also share how best to bring about the necessary speed in adopting or scaling AI automation in your organisation.
As a truthful machine-learning company we are always trying to cut through the noise and hype around AI and to avoid disillusionment and fatigue. An organisation adopting AI can typically be displayed on the following hype cycle:
Peak of Inflated Expectations
Organisations at this stage have inflated expectations about what AI can do for them; they are in the peak of the hype cycle. In the main these organisations and individuals have been overwhelmed with exaggerated sales pitches and the general noise in the market around AI and it’s promise of instantly changing business for the best. Many of the sales messages are around creation of an unbelievably beneficial human-machine interaction, which in some cases doesn’t even require a machine learning algorithm to be trained, while at other times it requires a completely new computing system to be developed specifically.
Engaging with organisations who are at this stage requires careful consideration as the gap between expectation and reality can lead to disappointment. We find that engaging at this stage requires us to follow some fundamental steps. First we engage with them thoughtfully to educate them about how ML exactly works, followed by an open and honest opportunity assessment that allows us to understand what it would take for an implementation of AI to succeed in terms of data access and structure, operational outcomes to be achieved and the organisation’s change capabilities and culture. We basically try to open their eyes so they can see clearly through the noise and skip the disillusionment towards a productive use of AI.
Trough of Disillusionment
Those who find themselves in this phase have mostly become a victim of the hype, noise and over promises. One-off-studies have most likely been performed using some kind of ML algorithm but the results were poor or unrepeatable. The truth is, for many outcomes, the answer may be nowhere to be found in the data. We know of many examples where engaging in expensive AI studies with poor results was an enormous waste of resources. This has led to increased skepticism, and for people to actively distrust AI results, preferring to follow their gut instincts instead.
In this phase the motivation of the team is poor, and the distrust in AI capabilities leads to delay or even avoidance of the productive adoption and use of AI.
For these organisations, we focus on rebuilding trust; winning back confidence and trust in AI capabilities and shift the mindset from AI being used in one-off-studies to a continuous learning cycle with AI. It is key that in this learning cycle we dare to acknowledge when our algorithm is not confident (i.e. low probability) based on the available data, but also dare to provide confidence (i.e. high probability) the answer when sufficient evidence is available. By continuously learning from new data, also the algorithms builds up confidence in the results. Getting this right is a skill that requires deep knowledge of machine-learning and a high degree of awareness of domain evidence.
Triggered by technological need
Our favourite type of organisations are the ones who have had limited or more technical exposure to AI as they are still unbiased. This group is not affected by the hype and also does not suffer any disillusionment in the application of AI. Typically they have been ‘triggered’ by a need to apply AI. To engage with these organisations we find it works best to have a straightforward conversation to assess the opportunity to apply AI, including necessary data access, operational and structural requirements. With this we build their AI system for productive use and take them straight to the productive use of AI, skipping the unnecessary hype.
If you are interested in learning how AI can help you improve outcomes and bring your team to the next level through innovation, improved efficiency and ROI by getting real-time insights from your data? Subscribe to get our regular content