AIRON

AI agents are quickly becoming one of the biggest topics in artificial intelligence.

Every week brings announcements of increasingly capable systems that can search for information, complete tasks, coordinate workflows, and interact with multiple applications with growing levels of autonomy. It is understandable that many research organisations are beginning to ask whether they should invest in AI agents.

For many institutions, however, the better question is not “How do we build AI agents?”

It is “Where should we start?”

Successful AI adoption is rarely about implementing the newest technology first. It is about building the right foundations, developing capability across teams, and introducing AI in ways that improve research support while maintaining trust, governance, and human oversight.

Step 1: Build AI capability before building AI agents

It can be tempting to jump straight into AI agents. However, for many organisations, the greatest opportunity lies in first helping research-supporting professionals become confident using AI in their everyday work.

Today, AI can already support activities such as drafting communications, summarising documents, analysing information, identifying funding opportunities, reviewing policies, and preparing meetings. These are often low-risk ways to explore what AI can do while building confidence and practical experience.

This stage is not simply about learning how to use AI tools. It is about developing AI literacy, understanding where AI performs well, recognising where human oversight remains essential, and establishing appropriate governance.

Without these foundations, introducing more advanced AI systems risks automating poor practice rather than improving it.

Step 2: Optimise before you automate

Once organisations begin using AI, the next step is not necessarily more technology.

It is understanding the work itself.

One of the biggest lessons from digital transformation is that automating an inefficient process rarely solves the underlying problem. AI provides an opportunity to rethink how research support is delivered, identify unnecessary complexity, reduce duplication, and redesign workflows around the needs of researchers and institutions.

The objective is not automation for its own sake. It is better research support.

Step 3: Introduce workflow automation

Once teams understand how AI can support their work and processes have been reviewed, workflow automation becomes a natural next step.

Rather than individuals repeatedly moving information between systems, routine activities can begin happening automatically. Funding alerts, proposal tracking, document routing, reporting, and research information updates can increasingly become connected parts of a wider workflow.

Importantly, people remain responsible for reviewing outputs and making decisions.

Automation should reduce administrative effort while allowing research-supporting professionals to focus their time where it creates the greatest value.

Step 4: Introduce AI agents where they create clear value

Only after organisations have developed AI capability, improved workflows, and established appropriate governance does it make sense to introduce AI agents.

Unlike AI assistants that respond to prompts, AI agents are designed to complete tasks. They can retrieve information, coordinate workflows, prepare reports, monitor opportunities, and interact with multiple systems as part of a larger process.

This does not remove the need for people.

Research management depends on judgement, context, ethics, collaboration, and institutional knowledge that continue to require human expertise.

The role of AI agents is not to replace research-supporting professionals but to enable them to spend less time on repetitive work and more time on activities that require human judgement.

Step 5: Build an AI-enabled organisation

Over time, AI becomes less visible as a collection of individual tools and more a natural part of how an organisation operates.

Research-supporting professionals spend less time moving information between systems and more time advising researchers, solving complex problems, building partnerships, improving services, and contributing to institutional strategy.

This is where AI creates its greatest value.

Much of the public conversation focuses on whether AI will replace jobs. For research management, however, the more useful question is how AI changes where people create value.

As AI takes on more routine execution, human expertise becomes even more important. Judgement, governance, ethics, collaboration, institutional knowledge, and strategic thinking remain central to effective research support.

The future is therefore unlikely to be defined by AI replacing research-supporting professionals. It is more likely to be shaped by organisations that successfully combine human expertise with responsible AI.

A practical roadmap

At AIRON, we encourage organisations to resist the temptation to chase the latest AI trend.

For many research organisations, the journey is unlikely to be a direct move from using ChatGPT or Microsoft Copilot to deploying AI agents.

Instead, it is more likely to begin with building AI literacy, followed by responsible AI use, workflow redesign, automation, and then AI agents. Over time, these foundations enable organisations to become genuinely AI-enabled.

Each stage builds the capability needed for the next.

The organisations that succeed will not necessarily be those with the most advanced AI. They will be the ones that understand where AI creates meaningful value, redesign work thoughtfully, develop the capability of their people, and keep human expertise at the centre of research support.

The journey towards AI agents does not begin with technology. It begins with understanding the work.