Skip to main content

What CIOs need to know about introducing AI

(Image credit: Image source: Shutterstock/PHOTOCREO Michal Bednarek)

The number of enterprises implementing artificial intelligence (AI) has grown 270 per cent over the past four years and has tripled in the past year alone. And as companies become more comfortable with using AI to analyse data, a growing number are turning to acquisitions to gain a data advantage.

Google and Salesforce provide two recent examples of this strategy. In June, Google spent $2.6 billion to acquire Looker, a data discovery and analytics startup. Looker’s goal was to be an end-to-end analytics platform, able to connect, collect, analyse and visualise data across various databases and applications. In August, Salesforce completed its $15.7 billion acquisition of Tableau Software, a provider of data visualisation and analytics. The goal is to deliver powerful AI-driven insights across all types of data and use cases for people from every background and skillset.

Obtaining data resources is only half the journey. The other half, arguably even more crucial, is applying AI effectively to achieve the desired insights and outcomes. No one understands this challenge better than CIOs. But many aren’t sure how to introduce AI in a cost-effective, low-risk way that will deliver measurable impact and lay the foundation for broader use.

Getting started

The first step is to develop an AI strategy that identifies the key business problems/data questions that AI is intended to address. This might seem obvious, but a surprising number of companies fail to take this critical first step.

Quick wins are paramount to show progress, so early AI projects should address well-understood, well-scoped problems that can be implemented quickly, with clear milestones for success. This quick-wins approach offers CIOs and their teams several advantages, including:

  • Hands-on experience working with AI ‒ fast
  • Upskilling internal resources AI skill, which will strengthen in-house expertise in both the short and longer-term
  • Delivering faster results and operational impact, which will build confidence for future projects

As C-suite executives and leaders see early returns on their investment, confidence across divisions will increase. This alleviates fears commonly associated with the adoption of new technologies that require substantial investment. As trust in the AI program and its ability to achieve results grows, CIOs will find it easier to win agreement for additional, follow-on investment.

Taking the reins

As with any new technology, there is a learning curve for AI to make accurate predictions on data. Once the strategy is in place, the next step is to find the right talent. AI should be designed not just by technical teams, but by people who understand the underlying data and what actionable insights it can yield. Ultimately, any mistakes in AI are down to human error in training and programming. Getting this wrong could have serious impacts for your customers and brand, for example, Amazon’s AI ruling out female job applicants.

Next, take a rigorous approach and a robust training methodology to tag and curate the data. For example, items like structural material strains, reactor temperature readouts, aircraft engine performance data must be properly identified, categorised, and labelled. CIO’s need to work with their AI leads to design effective training regimes and identify a level of accuracy that is appropriate for the business and specific use case. For example, in consumer marketing, AI might find that people who tweet about shoes are more responsive to shoe ads. If targeting shoe tweeters consistently leads to a 10 per cent uptick in sales, that might be considered a success. On the other hand, when using AI to identify engine fail in a plane, certainty is paramount.

Avoiding pitfalls

When looking to incorporate AI into the business process, CIOs should be aware of and avoid the following pitfalls.

  • Guard against overblown expectations about what AI will accomplish when it’s incorporated into the enterprise. To manage this, start small, with an eye toward ensuring that the AI plan isn’t beyond the organisation’s capabilities.
  • Remember that AI requires a significant investment in both time and money – not only to start a project but also to train algorithms correctly and to continually re-train them with the real-world, operational data accrued during its use.
  • Involve internal users when starting an AI project. Doing so gets the buy-in from people throughout the organisation and enables the CIO to avoid long-term organisational issues.
  • Finding the right talent to execute an AI strategy can be challenging, because successfully implementing AI is more about the expertise of the team as it is about the technology.
  • Massive capital injection into a shiny AI platform is not enough to be successful. Deploying AI requires the right blend of problem-solving, business domain understanding, and technical expertise, all steered by a clear AI strategy.

Beware of biases – even those you cannot immediately see

Applying AI to data will continue, especially as technologies become easier to integrate, costs decrease, and the benefits of doing so become more obvious. CIOs at Global 1000 enterprises can learn from digitally native organisations that have gone to great lengths to build strong relationships with other vendors to complement their internal teams, increasing their performance and capabilities almost overnight.

Compare that approach with large legacy companies in traditional sectors that are trying to transform into digital businesses. They tend to focus their investment and effort on platforms and infrastructure. All too often, they overlook people and culture. Those biases tend to be inherent in how they approach procurement, organise data, and respond to outcomes.

Bias in AI is a reflection of the biases in the humans involved in their development. AI bias can adversely impact the business. For instance, Microsoft released Tay, a social AI chatbot, touting it as a symbol of AI’s potential to grow and learn from people on Twitter. Tay was designed to converse with people on Twitter. Tay’s ‘personality’ developed based on these conversations. But Tay’s inability to ignore negative comments led people to overload it with a barrage of racist and sexist comments, which Tay quickly incorporated and responded with similar sentiments. In just 16 hours of life, Tay was quickly taken offline by Microsoft.

Enterprises that effectively deploy AI have CIOs that apply insights from data analytics throughout their organisations and consider every facet of the business. The CIO’s role in driving AI adoption will become increasingly important as the technology matures. As such, current and aspiring CIOs should identify ways in which AI can deliver business value, incorporate lessons learned from failed AI initiatives, develop strong partnerships with vendors that have successfully implemented AI, and develop their long term strategy for future AI deployment.

Matthew Jones, Lead AI and Data Science Strategist, Tessella

Matthew Jones is Lead AI and Data Science Strategist at Tessella, an international data science, analytics, and AI technology consultancy that is part of the Altran group.