Your business needs AI. Here’s how to overcome decision paralysis

A 2021 report by McKinsey found that more companies than ever before are adopting artificial intelligence (AI), with 56% of respondents saying that they use it in at least one business function.

By automating the mundane, AI frees up knowledge workers’ time to focus on loftier, transformational projects and optimise internal operations.

However, there are still many companies that are taking a cautious approach.

Despite having more data and greater access to advanced software, decision overload paralyses many businesses from adopting this new disruptive technology.

In a 2019 NewVantage Partners survey, technology accounted for 5% of AI failures, and process and culture accounted for 95%. Three years on, not much has changed.

For companies looking to unlock business transformation through AI, identifying the common decision points where AI stalls is the first step.

From compliance to infrastructure, personnel to timing, organisations must carefully weigh many factors. 

Are you plagued by decision paralysis?

1. Deciding when to take the plunge

   The most common refrain among business leaders on what is stopping their AI transformation is that their data is not ready for AI. Business leaders mention their data needs to be centralised. They may fear the introduction of data bias through the use of data sets that reflect the historical biases of the organisation. They may even have concerns about their own data shortcomings and the potential impact on the accuracy and efficiency of AI results.

   Business leaders must recognise that what they need is quality data — not perfect data — to reap the benefits of AI. Data is a compass; it is not always perfect, but it points us more often than not in the right direction. Data allows better-informed business decisions to be made. Regardless, companies can overcome any hesitations around data by conducting frequent checks on their AI outcomes, especially in the early stages of implementation.

2. Deciding what infrastructure to leverage

   When incorporating AI for the first time, businesses with legacy systems face the challenge of transferring their operations onto a modern analytics infrastructure that is agile, flexible, and scalable, with enough processing power to accommodate large volumes of data. Then comes the decision between building versus buying. Should an organisation buy off-the-shelf AI products that trade higher upfront costs and lower flexibility for faster implementation?

   Conversely, a leader building their own AI stack means having to toss up between trading IT risk, long-term upkeep, and maintenance for potentially lower costs and more flexibility in implementation. Both these approaches have their champions and detractors and the fight between both camps has paralysed many organisations. Many conclude that their organisation does not have the financial or talent capabilities to build out their own platform and look to supplement with off-the-shelf platform tools.

   Either approach to AI investments will come with a sizable price tag, either financially or in company resources. To succeed, leaders must achieve quick wins, plan for the long term, and limit the technical debt when deploying AI within their organisation. So how might this look in practice.

   Projects not in production have a lower inherent value; projects that can operationalise start generating returns and create wins for the team. Quick wins build organisational credibility, allowing the team to be strategic and take on a portfolio of AI projects. These projects should vary in scope, risk, and reward from easy automation wins to higher risk transformational projects. Constructing an AI portfolio this way ensures a steady set of wins with high potential upside. Last, the AI data and software landscape changes every six months. Today’s hot tech could be tomorrow’s abandoned project. Minimising the risk of a poor technology investment by minimising the technical debt carried by the organisation gives freedom to change the AI stack if or when the landscape dictates it. A great example of this philosophy is the use of cloud-based services. Small and medium-sized businesses can circumvent the need for new servers and powerful processors with this approach.

3. Deciding which segments of the organisation should have access to data

   According to LinkedIn, machine learning engineers and data engineers are the second and fifth fastest-growing job titles in Australia. Employees with these talents serve as the “tip of the spear” of AI transformation. As businesses raise their data and AI capabilities, demand for these skilled practitioners grows, making them harder to find, harder to hire, and more expensive. This may leave business leaders in a tough spot when balancing ROI, budgets, and timelines for projects.

   However, the days of data being used by data science teams alone are behind us, and the time has come to usher in a new era — everyday AI. Everyday AI is AI that is so ingrained and intertwined with the workings of the day-to-day that it’s a part of the business and not something that’s used or developed by one central team. By empowering all employees to use AI, the whole organisation can make better decisions, transform its work, and propel the company forward.

4. Deciding how to work with compliance rules

   As late as three years ago, compliance and regulatory issues were perhaps a third or fourth concern when introducing AI transformation. In the intervening years, regulatory agencies and industry watchers have become attuned to companies’ use of AI technologies. This has led to stricter regulations and safeguards put into place by government and industry bodies and more watchful third parties like the media. One only needs to read articles about Amazon, Zillow, Apple, and others to realise AI transformation isn’t just a financial risk. Business leaders looking to deploy AI at scale may find themselves crippled with building compliance processes, requiring central control while still pushing for more agility at the edges of the organisation.

   Overcoming regulatory paralysis means building a culture and mindset of compliance and respect for data regulation across the organisation. Creating opportunities for education and breaking down knowledge silos allows teams to uphold compliance across the board instead of weighing on business leaders’ minds. At a minimum, conducting frequent training and developing employee handbooks helps ensure you’re protecting your team from making compliance errors. That said, go further — incorporate data and AI checks into your projects and create an Institutional Review Board. Adding practices such as these can help to de-risk your organisations’ AI projects.

Businesses getting their hands dirty with data and AI for the first time may find themselves overwhelmed when trying to drive strategic outcomes. Leaders may even fear that tapping into AI may lead to redundancies in the company, especially at the decision-making level.

With a recent report predicting that AI colleagues will become a core part of business teams by 2030, business leaders must overcome this mindset to make the most of AI. By systemising the use of data and AI throughout the organisation, business leaders will see the decision-making responsibility spread among the entire team, allowing for a smoother rollout of AI, with company-wide collaboration and investment rewards.