It's All About the Data

Over the last seventy years organisations have purchased and deployed communications and computing technologies to enhance their productivity and competitiveness. As we enter the next phase of technology evolution, we will see increasing use of cloud-based storage and computational services, including AI as a service capabilities.

As technology becomes increasingly available on an “as a service basis,” an organisation’s data and the ability to derive insights from it, will become a critical differentiating factor.

One of the learnings, from the AI projects we have undertaken to date, is that many enterprises could enhance the return they derive from AI projects if they approached the management of data as though it was a critical business asset. I believe there should be a Chief Data Officer, who is accountable for the integrity and effectiveness of the strategic and operational aspects of data management including:

1. Implementing a suitable data storage and management regime that can cater for the high volumes of data and processing required. This should encompass all forms of data; structured (probably already in relational databases), semi-structured data (e.g. emails) and unstructured data (e.g. video clips, photos, content from digital media sites and transcripts from call centres, chat bots and collaboration tools).   
2. Managing regulatory compliance; obtaining permissions for processing and storing data, ensuring processing and storage policies are fully compliant with regulation and permissions and the security of the data.
3. Identifying potential insights which includes:

• Estimation of an unknown value based on historic data; an example is the price of a property given a textual description.
• Prediction of a future state based on analysis of historic data. For example, predicting footfall in a retail centre based on public transport schedules, traffic and weather information and the timing of other events in the area.
• Classification. For example, assign a positive/negative sentiment to a string of text.
• Detection; an example is text analysis where the AI develops an understanding of sentence structure and therefore can differentiate between types of names. i.e. it can determine if a name is a person, organisation or location without the use of look up lists. 
• Comparison; returns a quantified comparison according to some abstract criteria; such as indicating if two paragraphs of text are written by the same author. This could be applied in determining plagiarism, for example.

It may take many months to accumulate sufficient data to develop helpful AI solutions. Therefore, it is necessary to implement the data strategy as soon as possible. The key steps to take are:

1. Identify sources of trusted data.
2. Consider approaches to extract items of interest from the data into a structured format. There are commercially available platforms to facilitate this process.
3. Review how the data could be used in analysis that will be of business value
4. Consider how the data could be enriched with data from other sources to gain differential commercial advantage.

I am not advocating developing the data strategy in isolation from the business, but, based on years of practical experience I have concluded that It isn’t difficult for businesses to identify opportunities for AI. It is imperative to have data ready and usable; hence why I advocate the role of the CDO and a pro-active strategic approach to data management.

For practical advice on developing a data strategy please get in touch.