Organisations are going full circle with their unstructured data

New research from Iron Mountain, published in partnership with FT Longitude, shows that in pursuit of AI readiness, organisations’ top ambition for unstructured data over the next 12 months is to enhance AI-powered decision-making and agility. Such a goal reflects a widespread recognition of the intrinsic value of analytic data.

The study, based on a survey of senior leaders at 500 large organisations worldwide, shows that over the past 12 months, they achieved an average of $1.9bn in revenue growth as a direct result of their current information management systems and strategies.

The volume of data flowing into organisations, including unstructured data ranging from emails and social media posts to images, audio and video, is ever-increasing. “Some unstructured data may even still be stored in physical form and so must be digitised for AI models,” says Narasimha Goli, CTO of Iron Mountain.

But the research suggests that over three-quarters of organisations struggle to use unstructured data to consistently generate value and competitive advantage.

However, Iron Mountain’s research also reveals a cohort of leaders already demonstrating superior results, including greater revenue and profitability uplifts, because of how they manage their data. This group of high performers provides valuable learnings on enhancing the AI readiness of unstructured data.

The first lesson is the importance of securing unstructured data expertise, both in-house and from third parties. Encouragingly, non-leaders recognise this and are planning to upskill their workforce to offset talent shortage issues, with a focus on improving data literacy and addressing bad habits. Common bad habits include inconsistent processing methods, poor governance, inadequate security measures and a failure to address data silos.

Second, organisations plan to build on the successes they have achieved so far from using AI-powered intelligent document processing, quality assurance and quality control tools in their unstructured datasets. With AI readiness being the top objective for unstructured data, many organisations’ data strategies are going full circle. AI itself is helping organisations make their unstructured data more AI ready.

Leaders are not afraid to let AI analytics guide them

The leadership cohort ranks AI-powered analytics as one of the two most effective strategies to date. Moving forward, most of these leaders (55 per cent) plan to employ these tools to interrogate the accuracy and reliability of unstructured data, compared with only 36 per cent of other organisations.

Using AI to address integrity flaws in unstructured data will have tangible commercial benefits. On average, organizations in the research lost US$389,000 in the last 12 months due to data integrity issues.

AI-powered analytics can also assist with the efficient identification and analysis of unstructured data, another vital concern of the leader group. At Lloyds Banking Group, an AI-powered knowledge management tool used by the customer service team is already showing strong results: “It makes our processes easy for employees to understand by providing the ability to ask questions in their own words for a personalised response that guides them through processes step by step,” explains Rohit Dhawan, Director of AI and Advanced Analytics at the bank. The tool has halved the time agents spend finding and digesting information, and, as a result, the bank has “seen a reduction of up to 30 per cent in the amount of time customers are put on hold.”

Consolidation and classification are vital

The two other main unstructured data strategies leaders plan to use, cloud-based consolidation and automated classification, will also support their efforts to use AI to make their datasets more AI ready.

Data consolidation is a priority for leaders: 40 per cent plan to prioritise the use of cloud-based platforms to centralise data, compared with only 26 per cent of other organisations. This unification allows AI tools access to the necessary data pools to assist with the standardisation and quality checks that will ensure the reliability of the data being fed into the AI models.

Another priority for leaders is to automate the classification of the ever-increasing volume of incoming data. Accurate curating at speed will accelerate the correct allocation of that data to AI tools.

Organisations can also guide employees to use AI outputs responsibly by classifying unstructured data according to the extent to which it can be trusted to inform decisions. For instance, AI nutrition labels can distinguish authoritative data from any unverified data that should only be used for contextual purposes. When fed into AI models, these metadata labels can help users assess the quality of data an AI system is using, which allows them to exercise the right level of caution when using the AI’s outputs.

Explainability and human oversight have to stay at the core

Explainability is a critical principle. “AI tools can be incredibly helpful for extracting and structuring the insights hidden in unstructured datasets,” says Andrew Chin, Chief AI Officer at investment management company AllianceBernstein. “But sometimes, AI gets it wrong, it takes time to calibrate AI correctly and fine-tune it. To trust AI, humans must be kept in the loop to check that AI is responsibly sourced in terms of the credibility of the data and instructions it is working from.”

AI nutrition labels are especially helpful in this regard, providing users of an AI model with clear information on the data that generates its outcomes. Almost all the leaders in the research use AI nutrition labels (98 per cent), compared with 89 per cent of other organisations.

Even in an automated AI-powered environment, manual intervention, with human beings conducting checks and assessing data quality, is an essential element of the workflow, says Iron Mountain’s Goli. “For instance, by adding a human review into the automated tagging process, you will drive trust.”