Why data culture is blocking agentic AI success

In partnership with Domo, data leaders delve into the cultural challenges business and people face in the era of agentic AI.

Overview

​

• AI is stuck in pilot phase
• Businesses need a new data culture
• Understanding the value of data
• AI exposes the weaknesses in data

​

Artificial intelligence (AI) is such a new way of working that you would expect there to be problems in moving beyond experimental projects to realistic business outcomes. However, as a recent HotTopics Food for Thought debate uncovered, the biggest blocker to adoption is data - with much of this tying back to the organisation’s culture.

​

Digital and data leaders gathered with HotTopics’ partner, Domo, the data ecosystem technology provider, to discuss these topics.

Whether working in scientific research, financial services, private equity-backed or listed companies, housing associations, or the technology industry, all attendees revealed that cultural challenges are causing data issues. Those data issues are then blocking organisations from realising the potential for agentic AI to improve processes, unlock new efficiencies and modernise the organisation.

​

Field CTO of Domo, Jamie Morrison, posed the table the burning question:

“The average organisation runs four proof of concepts (POCs), pilots for AI use cases at any one time - but only 20% actually go into production and start delivering the intended value. This poses the question; why do these pilots fall over? Why do they fail?”

​

The CTO believes the main problem is not technological. Despite their relative youth, Morrison said the large language models (LLMs) that underpin agentic AI are “very capable,” but when it comes to moving an agentic AI pilot into a genuine business use, the wheels can fall off.

Morrison was not alone in this observation, with the chief data officer (CDO) of a scientific research organisation, adding:

“We've got a lot of data, a lot of value to extract from that data from a scientific perspective. I cover operations and research data, and from that perspective, I think culture is the block.”

​

Enterprise data hurdles: The operational culture blocker

​

Over the course of the debate, it was clear that there are two types of cultural blockers, one that is on the enterprise culture and another issue that is more people-based.

For instance, a director of data and insights said data-rich organisations such as hers are now finding that they don’t have the necessary context in that data to gain real business benefits from agentic AI. She said there are often issues with the metadata, which creates sourcing issues for AI, and that data may not have the right permissions attached. These data governance and management problems can exacerbate the likelihood of AI hallucinating, or not being able to tackle tasks effectively.

​

A number of other organisations have reported that the productivity benefits of AI are not being realised, as their teams are now spending time validating the work that AI does. Validation is vital and will always remain so, but to date, too many organisations are finding that AI is making validation harder, and in some cases, slower than existing, manual human-based processes.

​

That business culture was inhibiting AI in the private sector was a surprise to some in attendance. Data leaders from sectors such as housing associations thought the private sector would be ahead, but HotTopics debates have long revealed that the cultural challenge can be a problem across all sectors.

​

Morrison recognised these challenges and said many organisations have a wealth of data, but the data is not in a good enough state to be used by AI, as it lacks the necessary context.

​

Data democratisation: Turning teams into data champions

​

Across all sectors, it was clear that team members have developed a set of distinct cultures towards data.

This may be because they are under pressure and trying to quickly get to an outcome, or it could be down to an organisation or function with limited data training, skills or literacy. What this debate really showed was that organisations must now work with their people to develop a culture towards data and AI that benefits the individual, the organisation and therefore the customer.

As one CDO observed: “How do we support our staff to transition into their next phase, and can we alleviate some of the fears seen in the conversations that I've had recently?”

​

Others shared stories of widening data siloes - how staff members take their own data extracts and do their own analysis work, which causes friction in the organisation, as a debate, or worse, a dispute begins about which data is correct. The issue can often be that different business lines have different metrics; if they are then creating their own data sets, then these will suit their metric. That may not be the metric of the organisation. As one attendee said: “Historically, you figured out how to define a key performance indicator (KPI), you then figured out how to calculate the KPI.”

​

Creating a single source of the truth, sometimes called a ‘golden’ data set, is seen as one way to resolve this.

​

Other organisations are looking to democratise data. Rather than the data team building a data workflow for the marketing team, data ownership is given to those departments that need to use the data.

​

To succeed with people-driven solutions, like data democratisation and a single source of truth, requires a new culture towards data by team members, the debate found. Attendees said there needs to be a change in the mindset, so that everyone considers how they treat data, and how it may be used by them or the organisation 10 to 15 years from now. This will ensure the data remains useful. As one attendee said, everybody has to understand how data impacts the organisation, and that it is part of their function to curate data.

​

Overcoming AI FOMO: Building the foundations for agentic AI

​

AI has highlighted that organisations need to improve their data, which is not a new problem. Unlike previous changes in working methods and technology, AI has generated a lot of pressure from boards, but also employees, as attendees said, there is a fear of missing out (FOMO).

​

No matter the sector, all are struggling to quantify the value AI can deliver for the organisation, which demonstrates why the right approach to data is necessary. AI can extract value from good data, but without good data, it is hard to realise that value - and generate true return on investment (ROI).

​

AI has a wealth of potential to rethink ways of working and business models, but needs to be underpinned by data that is reliable and high quality.

Learn more from Domo’s industry experts on why AI’s impact hinges on the data, context and workflows behind it.