Trusted, transparent AI is fundamental to next-generation decision making

Thursday, 01 February 2024 15:43

Trusted, transparent AI is fundamental to next-generation decision making

By Ray Greenwood

Ray Greenwood, Director, ANZ Customer Advisory, SAS

GUEST OPINION: Transparent, explainable AI is more than a story of customer confidence. It’s about having trust in strategic decisions your business makes based on your AI’s analytics.

Trustworthy AI is quickly becoming one of the hottest business debates on the planet. While many of the discussions are about delivering confidence to the wider public, it’s important to see the value of transparent, explainable AI for truly transforming business processes.

We're dealing with global disruption that is only increasing. Information overload is far exceeding human capacity. So, as a business, you have to get beyond being data-driven. You have to become outcome-driven, leveraging AI to increase human productivity and scalability.

When AI gives you an answer that is intuitive, everyone agrees and moves on. The hardest thing is when you get an answer that challenges your assumptions. If you’re going to take that to your executive leadership and tell them to do something that’s counterintuitive but correct, you better be able to show the data and the proof of work behind that. Something transparent and explainable.

AI can often be presented as an inscrutable black box. But it's not that complicated. AI is rooted in mathematics. It simply studies the past to make predictions about the future. The reality is that if the data that you're studying is biased or underrepresents a particular group, I can guarantee the AI you create will also be biased and possibly have unintended consequences. SAS has a history of providing transparency to ensure financial organisations can meet their regulatory obligations for defensible decisions and medical organisations can maintain human oversight of decisions directly impacting patient wellbeing.

As leaders in AI, we must make sure decisions made using AI analytics are made on behalf of people. There's not some board-creating AI. We create AI. We choose AI models to apply to our own data sources. So, we have to make sure the decisions that are coming out of AI are explainable, transparent, and fair. At SAS, we call this responsible innovation.

Building transparent AI

SAS is proudly a cloud-native AI and analytics company. Our aim is to enable people to skill up and become more mature in their use of AI, so they’re learning and understanding more by working through this process. If you don’t have a data scientist or developer to build models for you in Python or SAS, it’s hard to compare models. But our AutoML tools enable people to target a variable, run models, and even take the pipeline and edit or extend it. We never hide that code.

The key here is that at the end of this process, you will see transparent, responsible AI scorecards that reveal details of the underlying training, assumptions and potential biases. Every model is given a natural language explanation in the context of the model. We give them performance benchmarks, and we give them explanations.

By using AI tools with this level of transparency, you can build trust in the AI analytics process because you can open it up and see how it works. If someone asks, “but why did it decide that is the right path of action?” we can do a lot more than shrug.

A constant qualification cycle

We need to be clear that AI holds the most potential when it is working with your data to understand your business’ specific needs and challenges and help make strategic observations around your future. At SAS, we constantly hear from around the world that data management is still a difficult problem. That often means that organisations spend 80% of our time preparing data and 20% deriving value from it. We are committed to bringing AI capabilities to absolutely flip that script and to do so in a trusted way.

Through SAS Viya we are automating the process of data discovery across your enterprise. We will analyse your data, build a data catalogue, and create an inventory with metadata and quality metrics. We can automate the entire analytics lifecycle. From data pipeline to model development to deployment, to model management.

Perhaps most importantly, that last step will become the most important over the long term. Data changes. Your business needs to change. So, we need to keep that lifecycle operating as a true cycle, returning to the data, again and again, to qualify its utility and ensure we are using the right models to achieve the best possible outcomes for the business.

Generative AI demands ever-greater trust

Trust is a story of vigilance. We don’t earn it once and keep it forever. We must constantly earn it and prove it to our stakeholders.

We also see generative AI becoming part of the analytics story. We’re already demonstrating tools that make our queries with AI data models begin to look like social media interactions. This will absolutely lead to arguments over what is true and what is not. So, it will be even more important to bring all of the above virtues – transparency, explainability, and trust – to the fore. We will give you the data insights that show our work so that everyone in the organisation can believe what the AI is telling you.

AI-driven analytics hold extraordinary potential for leading your organisation toward outcomes that transform your business and the lives of your customers. It just requires explainability and transparency to give you the confidence to take the insights seriously.

We’re seeing SAS Viya deliver massive productivity gains while also driving toward trustworthy insights that lead our customers to make better decisions that every stakeholder can believe in. We are orchestrating everything in our AI lifecycle to do that. And that’s really exciting.

Read 844 times

GARTNER MARKET GUIDE FOR NDR 2022

DOWNLOAD NOW!

PROMOTE YOUR WEBINAR ON ITWIRE

MORE INFO HERE!