GenAI as a challenger to human decision-making at the organizational level
Ryan Cox
Global Head of AI
Synechron
Every business has critical decisions to make. Every business has risks to manage. Before the digital era, organizations primarily relied on historical data and the intuition of seasoned experts to make predictions.
These methods demanded significant time and were prone to errors. Techniques such as the Delphi method, scenario planning, and SWOT analysis, even when performed by experts, could have elements of subjectivity and bias.
Things changed with the advent of computers and tools like VisiCalc came to facilitate faster and more intricate forecasting – but, drawing back to the point we made before, these tools didn’t eliminate the inherent uncertainties and the inaccuracies and fallacies in human decision-making.
Currently, we have more access to data than ever. Artificial intelligence (AI), particularly generative AI (GenAI) has dramatically transformed the approach to handling and analyzing vast, unstructured data sets, converting complex information into lucid, actionable insights. And the thing is, this transcends simple productivity improvements: It’s fundamentally about elevating the calibre of decisions made.
Let’s explore.
AI's separation of prediction from decision-making is critical.
Humans have a long history of combining prediction and decision-making, beginning as early as agriculture. For example, they predicted the right time to plant crops based on natural events like solstices, often marking the periods with rituals to signify the change of seasons. They even passed down the historical data in the form of stories to inform future generations and preserve essential knowledge.
However, in these early systems, the same people making predictions – priests or village elders – also made the decisions, giving them a unique power over society’s actions. And many times, the results were often subjective and biased.
The application of AI empowers a broader range of decision-makers with unbiased, data-driven predictions. From an organizational perspective, separating prediction from decision-making means predictions become more accurate and explainable, while leaving the human to steer and make the final decision.
AI is a powerful tool for prediction, but a lot of work goes into ensuring the data is normalized and in good shape.
For decades, traditional AI, particularly machine learning, has excelled at analyzing structured data. Data scientists have been inputting vast, organized data sets into these systems to uncover patterns and generate predictions. This has allowed AI to provide predictions that inform – but do not dictate – human decisions.
Handling missing values, duplicates, outliers, and inconsistencies, as well as making sure data is free from biases, is all part of it. In fact, data scientists spend up to 60% of their time cleaning and preprocessing data.
AI processes large data sets swiftly, identifying patterns and trends that human analysts might miss, while decision-makers benefit from these insights without being overwhelmed by data complexity.
It’s clear, traditional data cleaning can be labor-intensive, but what if we could optimize this step?
GenAI excels in handling imperfect data.
While predictive models have traditionally depended on structured data – organized tables within SQL databases, Excel spreadsheets, and CSV files – GenAI thrives on the challenge of unstructured data too.
Using advanced algorithms, GenAI adds structure where there is none, transforming verbose and unwieldy information into accessible, organized data that's primed for prediction and analysis.
Take Amazon customer reviews as an example – these free-text summaries vary widely and can’t be easily categorized, making it difficult to forecast outcomes like sales performance or project success based on traditional machine learning techniques. GenAI can sift through them, flagging instances where a positive review might oddly be paired with a low star rating and challenge inconsistencies. This, in turn, improves data quality and empowers more accurate predictions.
For data scientists, this means they can redirect their efforts from data cleaning to more strategic tasks. The end-to-end process becomes faster, more accurate, and less biased, giving businesses an advantage over competition with better decisions.
Allow GenAI to be a ‘devil’s advocate’ – it’s more than happy to fill this role.
Once the data is polished and we obtain our predictions from the traditional AI models, GenAI can question existing assumptions and introduce alternative approaches. This encourages a more comprehensive evaluation of potential strategies. As a result, decisions are informed by the best available information and the broadest array of explored options. For instance, when it comes to strategic business decisions, GenAI can suggest unconventional strategies that go against the grain of conventional wisdom. No matter the industry, it could:
· Sift through customer inquiries and feedback to detect an unmet need for a service that no current data points to, leading to the creation of a first-to-market offering.
· Analyze unstructured feedback in project managers' notes to recommend prioritizing a feature in the development pipeline that aligns with an unforeseen shift in user interest.
· Use social media trends to recommend products or services that are gaining online interest, even if they're new to a company's traditional offerings.
These are just some examples, and there’s a ton of possibilities to explore in this space. To sum up, GenAI streamlines data refinement and reveals insights that might otherwise go unnoticed.
But it doesn't make decisions for us. That's our job. Traditional AI has long provided us with valuable predictions, helping us understand patterns and trends from structured data. GenAI enables us to take things further by incorporating unstructured data, and it's this combination that allows us to challenge and refine our strategies. This is how we can steer our businesses towards smarter, more innovative futures.