Unstructured data is like a massive library where the books have no titles, no catalogue, and no shelving system; all the knowledge is there, but impossible to use until it’s organized.
While enterprise AI investments have increased over recent years, less than half of those AI projects in the past two years have delivered measurable business outcomes. While implementing the technology remains an ambition for organizations across the world this is a clear signal that value realization remains the industry’s biggest challenge.
Mixed results though have provided a valuable upside. This trial and error has uncovered the real issue that has been plaguing enterprises for years: they simply don’t have the foundations for any digital innovation, let alone AI. And organizations are finally waking up to this reality.
Latest Videos FromTechRadar
Chief Product Officer at Hyland.
A key piece of this foundation is found in an enterprise’s unstructured data. These are the emails, documents, PDFs, images, videos, chat logs, forms, reports, clinical notes, customer interactions that live outside of structured databases, and make up anywhere from 80% to 90% of an enterprise’s content.
This unstructured data is critical to the successful deployment of AI and agentic automation in the enterprise because it provides context needed for enterprise AI agents to act in a way that aligns with the business and enables agents to adapt to continuously changing business context. When done properly, with scalable human oversight and governance, enterprise agents act on a combination of AI-ready data, deep business context and industry expertise.
Agentic AI acts less like code, and more like a co-worker. It’s exactly this growth of capability that means charging ahead without a plan is risky. Recent reports show that the UK Public Accounts Committee found poor-quality data trapped in legacy systems is a major barrier to AI, further proof that fragmented data, shaky governance and fuzzy objectives will derail agentic AI, unless the foundations come first.
So, before committing time, effort and budget, every organization needs to ask two crucial questions: is your infrastructure ready for agentic AI and are the outcomes you’re aiming for clearly defined?
Prepare, then deploy
Unlike assistive AI that responds to prompts, enterprise agents are designed to act. Because enterprise agents carry out tasks rather than just acting as an assistant, introducing them requires a thorough audit of your workflows and data. To deliver value, they must move through the same terrain people do: the systems where work is recorded, the channels where customers interact, and the rules that keep everything compliant.
This requires a deliberate strategy and clear governance approach. If you treat agents like plug-ins, they could potentially duplicate effort, create bottlenecks, or carry out incorrect, unwanted actions that ultimately result in more work for the people you were trying to help by implementing AI tools in the first place.
That is why the work must start before deployment. Before even thinking about implementing AI, businesses need a platform that allows them to connect and leverage the terabytes of unstructured data that have historically been difficult to access and even harder to govern and activate at scale.
Content like emails, photos, and videos can all provide valuable information and context for agents to operate more effectively. However, until now, much of this data has remained siloed and underutilized, gathering digital dust instead of training smarter systems.
Strategy before scale
A common mistake when onboarding any new technology is that organizations often try to run before they can walk. While it’s enticing to envision how AI, particularly agents, will transform the entire business, the first step is understanding where automation can deliver real impact.
Take Accounts Payable (AP) for example. In AP, agents are well-suited to work that is repeatable and rule bound. They can capture invoices from any channel, convert them to clean, structured data, validate the key fields, and perform three-way matching against purchase orders and goods – before posting to the finance system.
Crucially, because they sit inside the existing stack, approvers see what they need without hopping between apps, and mobile sign-off keeps cycle times tight. Every step is logged, so audit trails and fraud checks aren’t bolted on later, but instead, a part of the flow.
While this automation saves hours of effort, human involvement remains the glue that holds it together. An agent can field most of the routine workload, but keeping human judgment where it changes outcomes is essential.
Aim first, automate next
In 2026, AI is no longer going to be an experiment. Leading enterprises are already putting this technology to work at scale, and those that keep treating it as a trial run will be left behind by competitors who are building on strong foundations. The experimentation phase is over, and organizations must start learning, both from their own mistakes and from those who have succeeded.
Getting started means getting the basics right. That means curating and refining your existing data and implementing connected infrastructure that scales alongside innovation. Importantly, these investments are far more than just ‘AI upgrades’. They’re changes in how decisions get made and how your organization future-proofs itself.
The winners in this new era won’t be those who were first, but instead those who take the time to strategize, determine where automation will deliver real impact, and give their AI systems the foundations and resources they need to help your business flourish.
What’s to come in the year ahead
Over the next 12 months, enterprise agents will shift from assistive tools to governed operators that interpret, validate, and act on unstructured content at scale. Policy‑encoded autonomy, human‑in‑the‑loop controls, and federated content intelligence will make multi‑agent collaboration a standard workflow pattern.
Triage, validation, compliance, and orchestration working in concert across finance, HR, customer service, healthcare, government, and insurance will be commonplace. Natural‑language automation will become the universal interface, letting teams design, simulate, and deploy processes without specialized development, while closed‑loop feedback continuously improves cycle times, exception handling, and straight‑through processing.
At the same time, verticalized agent skill packs and repeatable blueprints will accelerate adoption, and advance production‑grade operating models. Agents will observe work, propose optimizations, and safely execute under role‑based autonomy with full audit trails and rollback protection.
The backbone is a unified, federated content layer, connecting repositories and applications without forced migrations, so agents can reliably gather context, link evidence to source documents, and trigger downstream actions that deliver measurable outcomes.
Bottom line: the trial phase is over. Enterprises that get their data foundations, governance, and enablement in place now will run faster, smarter, and safer with agentic automation. Those that wait will spend 2026 watching competitors turn unstructured content into an action engine — and get left behind.
We’ve featured the best AI chatbot for business.
This article was produced as part of TechRadar Pro Perspectives, our channel to 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/pro/perspectives-how-to-submit
https://cdn.mos.cms.futurecdn.net/h8ZQHernNUVpnGYX7QnxVM-2560-80.jpg
Source link
Michael Campbell
