Last year McKinsey conducted a survey on the adoption of Artificial Intelligence (AI) across several industries. More than 75 percent of respondent companies have either adopted AI or are piloting the use of AI into at least one standard business process. Although the survey identifies multiple applications for AI across different sectors which can bring significant value, few companies are extracting the full potential of AI.
A common factor amongst the challenges and barriers for AI adoption identified in the survey is a lack of trust in AI by executives and functional teams as the integration with business processes is not intuitive and transparent.
One possible explanation is the fact that the majority of companies adopted turn-key AI solutions. Although they may reduce lead time of AI initiatives, turn-key solutions often require modifying the existing business processes and are generally perceived as disconnected black boxes.
At Savvy Data Insights, we believe that successful adoption of AI requires a seamless integration with existing decision-making processes using explainable and transparent insights. To this end we have designed a flexible end-to-end innovation framework that covers the different phases of adopting AI and other innovative technologies.
The initial pillar (Engage) of the Savvy Innovation Framework brings companies closer to AI, Machine Learning (ML) and Blockchain. We empower executives with the right knowledge to lead internal initiatives and train functional teams on how these technologies can be integrated with business processes. We also help companies identify the main internal challenges and opportunities that can benefit from these technologies.
The next service pillar (Build) covers the development of innovative bespoke solutions over multiple stages to ensure full alignment with business users. Our bespoke solutions follow a savvy approach that amplifies the expertises of business users by combining their expert knowledge with insights extracted by AI/ML algorithms from both internal and external data sources.
“No man is better than a machine, and no machine is better than a man with a machine.”
Paul Tudor Jones, Founder of Tudor Investment Corporation
In our bespoke solutions, the role of the AI/ML algorithm is to reduce the complexity of the business problem by identifying the most relevant actionable insights. For example, from a large number of client transactions to the most suitable cross-selling opportunities; from an extensive amount of expense claims to the ones that are potentially fraudulent. The role of the decision-maker is to analyse the insights extracted by the AI/ML algorithm and select which instances need to be acted upon.
Could AI/ML algorithms fully replace the decision-maker?
“All models are wrong but some are useful.”
George Box, Statistician
Although AI/ML algorithms extract relevant insights from vast amount of data, not all the information associated with a business problem is in digital form or can be easily processed. For example, Customer Relationship Management systems do not contain all the information account managers have on clients; expense claims do not embed the full context of business trips.
On the other hand, decision-makers find it hard to identify non-obvious relations in large datasets. For example, sales managers of an office supply company can easily define rules to find complementary products: customers who buy printing paper should also buy toners. However, identifying cross-selling opportunities across thousands of clients and hundreds of products is a much harder task that requires a significant amount of time and effort.
Nonetheless, human decision-makers are extremely proficient at validating a reduced selection of relevant cases using their experience and knowledge. For instance, when a recommendation algorithm identifies cross-selling opportunities for a particular customer, the algorithm does not know why that customer has not yet bought the products/services identified. An account manager can easily select the most suitable products/services suggested by the algorithm using his/her own knowledge about the customer.
Our bespoke solutions deliver a seamless integration with existing decision-making processes by presenting actionable insights extracted by AI/ML algorithms using custom interactive dashboards. This approach allows decision-makers to intuitively understand and decompose the reasons behind every insight and extract the full potential of AI.
Our savvy approach can be applied to both simple and very complex AI/ML algorithms. Consider for example the application of Deep Learning to identify animal species based on photographs. Although it is extremely difficult to interpret the general Deep Network model used, it is possible to highlight which parts of the image were considered relevant when identifying the animal species for a particular image.
A similar approach can be applied to an AI algorithm that detects cancer from x-ray images. However, if a tool only presents the diagnosis produced by the AI algorithm and a correspondent probability value, radiologists will find it difficult to trust and use the these results. By highlighting which part of the x-ray was used to produce such diagnosis, the insights extracted by the AI algorithm are presented transparently to a radiologist thus potentially improving the analysis of x-ray images.
If you want to know more about integrating AI into your business processes in a transparent way and improve your efficiency and effectiveness, please do not hesitate to contact us (email@example.com)