Skip to main content

How to get artificial intelligence right

(Image credit: Image Credit: Geralt / Pixabay)

Due to the ever-increasing ubiquity of artificial intelligence (AI) in our day-to-day lives, and businesses’ growing reliance on it for a range of tasks, the stakes have never been higher for IT to get AI right. The consequences of successful or unsuccessful implementation are potentially profound. Used incorrectly, AI is capable of reinforcing harmful biases and increasing polarisation, as well as having many other damaging ramifications.

With the appetite for AI growing ever stronger and the hype surrounding its possible applications, it can be tempting to focus on the technology and coding disciplines — what one might call the “artificial” aspects. Nevertheless, the “intelligent” aspects of a digitally connected world cannot exist — let alone function — without data. Although the majority of executives and business and IT professionals are familiar with the people, process and technology capabilities of business models, they do not “speak data” fluently. Yet data is one of the cornerstones of any AI process.

In order to use AI accurately, organisations must advocate for data literacy becoming a new core competency for creators and consumers of AI alike. CIOs responsible for enabling AI initiatives are well-advised to follow a three-step approach: build AI right, use AI right, and, finally, keep AI right.

Build AI right

Before attempting to “build AI right”, it is imperative that those involved first establish the basic vocabulary of AI — a technical dialect used by people who “speak data”. CIOs should at least determine the primary terms employed to describe an AI system or solution, the reason the solution is being developed, and other key terms in relation to the different types of data that are used in and gathered from the solution.

Along with models and algorithms, data is foundational in any AI process. AI consumes and produces data. AI data design requires an understanding of and confidence in processing the datasets that will be parsed by the AI algorithm. CIOs and data and analytics leaders will be responsible for establishing and maintaining the data management aspects of AI. To succeed, developing data management expertise across the entire process is crucial.

Use AI right

Information language barriers can exist locally or systemically, regardless of the scope of a program or maturity of an organisation. Addressing a barrier demands a shift in mindset, as well as deliberate acknowledgement and intervention to “course correct”. While some may already appreciate the foundational role played by data in business transformation, the majority will not. Therefore, skilled leadership and purposeful change management discipline are required, beginning with the recognition of information as the new vernacular of the digital revolution.

Data literacy is the ability to read, write and communicate data in context. It includes an understanding of data sources and constructs, analytical methods and techniques applied, and the ability to describe the use case, application and resulting value. It is an underlying component of digital dexterity, which is an employee’s ability and desire to utilise existing and emerging technology to drive better business outcomes.

As organisations become more data driven, poor data literacy will become an inhibitor to growth. In order to hone data literacy across an organisation, CIOs should develop a data literacy program, implementing the following steps:

  • Identify fluent or native speakers who speak data naturally and effortlessly. Fluent speakers should be well-versed in describing contextualised use cases and outcomes, the analytical techniques applied to them, and the underlying data sources, entities and key attributes involved.
  • Identify skilled translators. Classic translators are often enterprise data or information architects, data scientists, information stewards, or related program managers.
  • Identify key areas where efficacy of data and analytics initiatives are being hindered by communication barriers. Pay particular attention to business-IT gaps, data-analytics gaps, and veteran-rookie gaps.
  • Actively listen for business outcomes that are not being clearly articulated in terms of explicit actions. What business moments are being enabled with enhanced data and analytics capabilities? What operational decisions are being improved?
  • Identify key stakeholders who require specialised translations. With a view to assessing data literacy levels, request that key stakeholders articulate the value of data as a strategic asset in terms of business outcomes, including enhanced business moments, monetisation and risk mitigation.
  • Compile and maintain a list of key words and phrases. Engage the data and analytics team in determining ways to articulate these phrases better.

Keep AI right

With all AI solutions, the concepts of “good AI” and “bad AI” arise. However, there is no singular or comprehensive definition of the terms. Choices made by different parties can have significant implications. Even the most successful companies should avoid falling into a false sense of security by believing they are immune to ethical mishaps implicated in AI. Extensive, dedicated discussions are necessary to discern different types of ethical questions and dilemmas that a company can be faced with versus the actual ethical position that it can take. CIOs should apply digital ethics and digital connectivism to guide and make explicit the business decisions and choices where AI is to be adopted.

The following steps should be considered:

  • Review the wider picture and absorb digital ethics and digital connectivism as a philosophy for the improvement of digital business (and digital society in a broader sense). The digital society is being shaped by interactions taking place between digital citizens. Rather than being designed per se, it is emerging organically. Nevertheless, digital leaders can determine considerably what it looks like.
  • Actively seek out ethical case studies that relate to the use of data in AI, as the ethical questions that businesses are confronted with are seldom new. Opportunities include competitive differentiation and a superior value proposition. Dangers include reputational risk, regulatory issues and financial losses.
  • Use AI algorithms and data exchange as enablers of digital interactions, and as a way to enable stakeholders to participate in an ecosystem, rather than as specific process controls. Encourage everyone who contributes data within the AI environment to be active participants in a mutually beneficial ecosystem.

Alan D. Duncan, Research VP Analyst, Gartner (opens in new tab)
Image Credit: Geralt / Pixabay

Alan D. Duncan is a Research VP Analyst at Gartner.