How to eliminate digital friction to keep AI innovation on track

Generative AI – and now Agentic AI – is transforming the way we live and work on a scale that hasn’t been seen since the dotcom boom.

Organizations everywhere are looking for ways to use GenAI to gain efficiencies and boost their competitive advantage.

However, while advances in these technologies are enabling innovation, they are also increasing infrastructure complexity.

A recent BCG study found that, although 83% of companies rank innovation as a top-three priority, only 3% are ready to deliver on their goals. This discrepancy is driven by several key challenges:

• Untrustworthy data – Fragmented, siloed, uncategorized, and unsynchronized data will never deliver strong outcomes from AI initiatives
• Outdated systems or processes – Manual interventions, offline data dependencies, and slow or unwieldy processes can collapse under pressure unless well-orchestrated.
• Security risks – Data protection and access control are essential, especially where systems interface and data is exchanged.
• Governance complexity – Managing who uses what data, how, and for what purpose must be considered from the outset.

To keep their AI innovation plans on track, organizations need to overcome these challenges to eliminate friction between fragmented systems and data. The biggest struggle is often knowing where to begin.

Good data unlocks value, bad data creates chaos

Data lies at the heart of all modern technology, and acts as a force multiplier for transformation initiatives. The proliferation of applications, data fabrics, and AI agents is accelerating this trend.

However, organizations with legacy systems, technical debt, and manual processes risk being overwhelmed by the growing demands.

Even if current data is of high quality, scaling it with increased volumes and unmanaged APIs may introduce vulnerabilities and risks of breaches.

Systems that are not resilient will fail under pressure, eroding customer and supplier trust.

Processes that collect, transform, manage, and deliver data must evolve to handle greater volumes securely and with high accuracy.

When data is fed into AI frameworks, its quality directly impacts the outputs of those solutions. Poor quality data increases the likelihood of AI hallucinations, errors, and misguided decisions.

Organizations that rely on rigid data systems will also only amplify technical debt in the future as their AI initiatives progress.

A new data infrastructure is essential

To meet these challenges, organizations need secure, flexible, and adaptable systems that manage and connect data across devices, systems, and applications.

Low-code/no-code AI-driven integration strategies can drastically reduce the effort required to build and maintain these integrations.

By orchestrating automation through cloud-native workflows, organizations can also support centralized updates to their data.

Finally, agentic frameworks will lead to decentralized workloads becoming the norm, so strong governance and central control of data will be essential.

To succeed in their AI initiatives, organizations need confidence in their data.

This starts with a system that can aggregate, ingest, deduplicate, validate, and govern data across different frequencies and sources – forming a golden record.

That system needs to seamlessly adapt to new data sources and ensure safe, traceable access for all authorized consumers, from AI models to big data stores.

In this way, organizations can create a strong data foundation that enables their teams to:

• Generate operational insights in real time to identify inefficiencies or boost performance.
• Leverage low-code solutions to develop agents quickly using a robust pipeline of contextual data.
• Improve automated workflows by analyzing performance and scaling enhancements via a common interface.

Once a robust foundation is in place, safeguarding it and upholding responsible practices becomes critical, especially when enabling AI agents to access enterprise data.

APIs: The backbone of agentic AI

AI agents function by perceiving their environment, processing data, and completing tasks to reach defined goals. Advanced agents incorporate reasoning and learning, continuously refining their performance.

These agents rely on tools, perhaps most crucially including APIs, to act. Planning and reasoning agents can select optimal APIs to complete tasks efficiently.

While all organizations use APIs, not all follow robust security practices. APIs are now the top software attack vector, with shadow APIs particularly vulnerable.

To mitigate the risk of their AI initiatives, organizations need to ensure their APIs adhere to security best practices.

This includes implementing strong authentication and authorization, encrypted communication, and robust input validation and sanitization controls.

Appropriate rate limiting, as well as monitoring, reduced data exposure, and rigorous testing are also important.

Security policies should be applied in a decentralized way throughout the API lifecycle. This enables federated control that safeguards innovation while avoiding bottlenecks.

An agile API management system with central oversight, spanning internal and external APIs, can support this approach.

A modern solution for an age-old problem

Many digital transformation programs were initiated years ago under very different conditions.

As innovation cycles speed up, organizations can’t afford to lose pace because of longstanding issues in their data management practices.

Adaptability and composability are essential to riding the fast-moving waves of technological advancement.

Modern integration and automation strategies and platform-based approaches have made the challenges of the past infinitely more solvable.

It is now far easier to build a holistic, scalable, and secure data foundation to power AI initiatives.

Taking this more unified approach will enable organizations to turn their data into a real-time strategic asset, ensuring they can make the most of their ability to innovate today and long into the future.

We've featured the best AI chatbot for business.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro