6 hours Ago
Imran Aftab, CEO and Co-Founder, 10Pearls —driving AI innovation and creating meaningful opportunities that make an impact. getty Widely heralded for its " proactiveness ," agentic AI is an autonomous, goal-oriented technology that’s able to take decisive action without much, if any, human supervision. AI agents are taking the world by storm.
KPMG found that 51% of companies are exploring using them, and Gartner predicts that agentic AI will be a part of over a third of enterprise software applications by 2028, autonomously making at least 15% of day-to-day work decisions. However, are business leaders trying to run before they can walk? Generally, yes, as many currently lack a strong enough data strategy. In fact, according to McKinsey research, about 20% of organizations still consider data to be the biggest challenge holding them back from capturing value from AI.
As opposed to previous AI tools, including generative AI (GenAI), agentic AI isn’t just about churning out responses—this latest innovation is about taking action. Specifically, agentic AI is designed to make complex decisions with minimal human involvement by learning from its environment and the real-time information available to train it. Agentic AI tools refine their decision-making abilities from the experiential data they’re exposed to.
This concept is known as reinforcement learning and is a core driver for agentic AI to deliver tangible business results. However, because of how agentic AI behaves with its environment, the efficacy of its decisions and actions is linked to the quality of the data they’re trained on. Business leaders can’t expect AI agents to behave reliably if they’re training these tools with data that is fragmented or siloed, for instance.
It’s widely known that any AI solution is only as good as the data it’s fed. Just like Amazon’s résumé-filtering tool favored men due to a biased dataset with considerably more male hires over 10 years, data errors like silos and fragmentation feeding agentic AI would send bias and unreliability off the charts. Weak and compromised data also means organizations are running a higher risk of data poisoning and manipulating how the agent yields responses via prompts.
Picture a logistics company relying on agentic AI to optimize delivery routes to speed up delivery times and cut down on fuel costs. Its teams make sure to regularly synchronize and update their operational data. As a result, the agentic AI tool efficiently assigns drivers so they’re not overworked and optimally plans routes, speeding up deliveries, boosting customer relations and keeping staff happy—leading to more profitable operations and better resource management.
Fortunately, many business leaders understand data’s crucial role in AI adoption. According to Deloitte, 75% of early adopters of GenAI are prioritizing data management in their IT investment. Before diving into revamping your entire organization’s data management, your first step should be to holistically assess the current state of your data.
Ensure data accuracy and completeness and identify where inconsistencies—such as incorrect formatting, missing data, duplicates and silos—occur. This involves taking all the data quality assessments from departments across the organization, looking at how to unify them as much as possible and implementing data standards. Once these inconsistencies have been addressed, organizations should also look at how that data is stored, accessed and tested for training and informing models (like LLMs) that are used to power AI agents.
Data isn’t just a concern for the IT department. Championing quality data needs to happen across the entire organization, starting with leadership. Leaders shouldn’t just be ensuring that AI goals align with data infrastructure but also need to be involved in implementing strong strategies from the ground up.
Achieving the structured, well-governed data needed for proficient and dependable AI agents takes an enterprise-wide effort across departments, too. For that to happen—and to get a robust data strategy off the ground—here are a few factors to consider. 1.
Define how AI-driven decisions are going to support and transform measurable business outcomes. For example, that could mean mapping out how agentic AI can optimize operations by predicting failures or automating workflows to reduce costs or improve customer relations. 2.
Reinforce your data infrastructure to address any data inconsistencies. Siloed and fragmented data cause bottlenecks since agentic AI needs instantly accessible, real-time information to make intelligent decisions. Withheld data can result in faulty predictions, which undermine agentic AI’s proactiveness and reliability for autonomous decision-making.
Moreover, poor data hygiene can cause cascading failures as agentic AI continuously operates and learns from its environment. 3. Regularly clean your data rather than wait for something to go wrong with their agentic AI.
One method is using AI tools to automate real-time data cleaning and detect any anomalies like duplicates in existing databases. This can also help standardize data formats across systems to overcome silos and allow agents to consistently access structured and unstructured datasets. When data isn’t regularly updated and validated, old and inconsistent information can mislead agents.
Schedule frequent updates so your data is in excellent shape for maximizing the potential of AI agents. For instance, by using reliable, strong and individualized financial data to develop an AI assistant, a fintech firm can slash countless hours of manual work reconciling accounts while providing intelligent insights and task-specific recommendations and assistance. 4.
Introduce a governance framework to maximize security and trust around data-sharing and use via AI agents. Cyberattacks compromise data and can lead to massive problems like data poisoning, which sabotages agentic AI’s performance and productivity. Compliance is an integral part of a governance framework, and agentic AI needs trustworthy data that is permitted to be shared.
Organizations have to enforce policies around data access and sharing plus security measures like encryption as part of complying with regulations like HIPAA and GDPR. Data sets the tone for agentic AI’s success. Complete, consistent and continuously updated data is key to AI-driven transformation that unlocks tangible business value.
With the right data infrastructure, agentic AI can help businesses achieve a competitive advantage and maximize operational efficiency. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?.
