Invest in Responsible AI decision making at the outset – or lose its impact

This year, 72% of business leaders reported that one or more of their organisation’s business functions has adopted AI, according to a McKinsey survey.  

From the outset, this looks positive. It appears that businesses are increasingly realising the potential of AI decision making. 

However, I am not convinced. 

After all, the same research from McKinsey finds that only 8% of respondents’ organisations are using AI in five or more business functions. Roles like marketing and sales are using it more than those with a need for nuanced decision making, such as HR and risk.  

Introducing AI tools one at a time, without a company-wide strategy, is shortsighted.  

The potential of AI for complex decision making lies in responsible transformation – and it’s by understanding this now that you will be competitive in the long run. 

Invest in Responsible AI from the outset for robust results 

For some businesses, it may be the fear of getting AI ‘wrong’ that is driving slow AI adoption. With lawsuits like Mobley v. Workday Inc receiving lots of media attention, this is understandable.  

For other businesses, it could be that bespoke systems are not yet readily available, and so they must rely on off-the-shelf models that do not address their specific needs.  

Whatever the reason, they are missing out. AI tools can help organisations tackle gender inequality in recruitment and career progression, generate new customer insights, and support the development of business strategy. 

For the businesses that want to innovate, and see the largest returns on their investment, mapping a path to informed AI decision making is key.  

To achieve this, AI tools must be developed and deployed in a responsible way from the very beginning.  

My colleague, Dr Hayley Watson, sums this up in her recent article: 

“The way AI is designed matters, as it impacts the extent to which insights can be meaningful for those who rely on them to drive operations and decision making.”  

But what does a Responsible AI transformation look like? 

1. Create an AI strategy

To be candid, using a handful of siloed AI tools is not going to help you achieve your business objectives. Once you have decided which problem you would like to address with AI, you need to set about creating a paradigm shift – a fundamental redesign of how you make decisions. 

Imagine that you are an insurance company assessing health insurance claims for reimbursement. To address this challenge, you might need an interoperable AI system that consolidates data from multiple channels, identifies patterns without historical bias, and complies with new and emerging regulations. 

To get the best from this tool, you would need to embed responsible practices – such as transparency and explainability, human oversight, and privacy – from the start. Otherwise, it could generate spurious results, and harm the very people you are trying to help. 

2. Establish strong AI governance

Just as responsible AI approaches are not something that can be ‘added in’, AI governance also needs to be invested in from the outset.   

A core part of this will be creating policies to guide how tools are used internally. If your employees create harmful outcomes from their AI use, it is your organisation that will be held liable. This is true even for AI tools that you don’t ‘own’, such as ChatGPT.  

Alongside internal policies, it pays to build a culture of responsibility – as my colleague Dr Rachel Finn discusses in her recent article. When you have an organisation of people who understand the value of using AI tools legally and safely, you can be confident that they will work cohesively to achieve your business objectives.

3. Invest in ongoing AI assurance

AI governance must not be viewed as a tick-box exercise. It must be an ongoing practice to help you maintain outcomes that are relevant, profitable, and safe. 

As I say in this blog, the idea that you could remove bias from your AI tools just once is simply untrue. You need to invest in ongoing AI assurance as part of your transformation, or you could leave your organisation vulnerable to skewed results, poor decision making, and legal repercussions.  

Before you develop or deploy AI tools, make sure you have AI assurance frameworks in place. Whether you create these manually or use an AI governance tool, the point remains the same: be proactive or diminish your AI’s potential. 

With a multi-disciplined approach comes meaningful AI decision making 

In my experience, the businesses that benefit from AI decision making the most are those that invest in responsible approaches from the start.  

Of course, this process can be daunting. It requires a deep understanding of responsible AI principles, the legislation surrounding AI use, and the business case for AI tools to meet internal objectives. 

That’s why Trilateral Research has adopted an interdisciplinary approach to building and governing AI for our clients. With a team of experts in social science, law, data science, and many other fields, we create practical solutions that help businesses gather insights for robust AI decision making. 

Read more about our approach by exploring our projects, where we have supported on challenges related to health, data governance, defence, the climate, human rights, and much more.