AI, funding and research culture: are we solving the right problem?

As artificial intelligence tools become more widely embedded in research processes, attention has largely focused on how they are changing the way research is conducted. But a parallel shift is now emerging: AI is beginning to shape how research is assessed and funded.

Faced with a growing volume of grant applications—driven in part by the ease with which AI tools can support application writing—UK research funders are beginning to consider the role of AI in processing and filtering submissions.

On one level, this is a pragmatic response to a capacity challenge. On another, it raises a more fundamental question: what happens when the technologies that make it easier to produce applications also become part of the system that determines which of them succeed?

A system under pressure

The challenge facing funders is clear. Application volumes have been increasing for some time, and AI is accelerating this trend:

• drafting tools reduce the time and effort needed to prepare proposals

• applicants can iterate and refine bids more quickly

• teams can submit to more calls, more frequently

This creates a structural tension. As the cost of producing applications decreases, the volume increases. The burden then shifts to funders and reviewers, who must process, assess and differentiate between a growing number of bids.

Introducing AI into this part of the system can be seen as a logical step—an attempt to manage scale by increasing efficiency.

But this is not just a technical adjustment. It is a systemic shift.

Filtering is not neutral

Any system that filters applications shapes what is visible—and what is not.

If AI is used to triage or prioritise applications, it will necessarily rely on signals:

• past performance

• language and structure

• alignment with expected formats

• patterns derived from previous funding decisions

  
  

These signals are not neutral. They reflect the existing definitions of quality within the system.

There is a risk, therefore, that AI does not simply manage volume, but reinforces existing norms:

• privileging well-established applicants over emerging voices

• favouring polished, well-optimised bids over more exploratory or unconventional ideas

• amplifying patterns of past success rather than enabling new directions

This is not unique to AI. But automation has the potential to scale these dynamics quickly and quietly.

AI as both cause and solution

What is particularly striking about this moment is that AI is both contributing to the pressure and being proposed as part of the response.

• AI makes it easier to write applications → more applications are submitted

• More applications place strain on funding systems → AI is used to manage the volume

This creates a feedback loop:

The more efficient application writing becomes, the more selective and automated application filtering is likely to become.

Over time, this risks turning the funding system into an optimisation challenge, where success depends not only on the strength of the idea, but on how effectively applicants can navigate both AI-assisted writing and AI-influenced assessment.

AI and uneven advantage

Alongside questions of efficiency and scale, the rise of AI in research funding raises important questions about equity.

At recent discussions on equitable partnerships, including at the conference in Pretoria, there was growing recognition that AI has the potential to exacerbate existing inequalities in the research system. Access to more advanced tools is increasingly shaped by the ability to pay for premium subscriptions, integrate tools into workflows, and build capability around their use.

This creates a familiar pattern: what begins as a broadly accessible technology can quickly become stratified.

At a global level, this risks reinforcing existing imbalances between Global North and Global South researchers. Where institutions in higher-income systems are able to invest in paid tools, training and infrastructure, others may be limited to default or lower-functionality versions. Over time, this shapes who can produce competitive applications—and at what speed.

These dynamics are not only global. Similar disparities are already emerging within institutions in the UK. In some cases, well-resourced teams have access to multiple paid tools and dedicated support, while others rely on standard institutional provision.

The result is not just a difference in tools, but a difference in capacity:

• to produce applications quickly

• to refine and iterate proposals

• to respond strategically to funding calls

If AI is both enabling the production of applications and contributing to how they are processed and filtered, then access to these tools becomes more than a productivity issue. It becomes part of the infrastructure of opportunity.

Efficiency, or redistribution?

The introduction of AI into funding processes is often framed in terms of efficiency:

• reducing reviewer burden

• speeding up decision-making

• managing scale

But it is worth asking a more nuanced question:

Is AI reducing administrative burden—or redistributing it?

For applicants:

• the expectation to produce more, faster, and more polished applications may increase

• the need to engage with AI tools becomes part of the “hidden work” of applying

For institutions:

• new systems, guidance and governance around AI will need to be developed

• disparities in access and capability may need active management

For funders:

• oversight, transparency and assurance around AI systems introduce new complexities

Efficiency gains in one part of the system may create new pressures in another.

What problem are we solving?

This raises a deeper question for the sector.

Is the core challenge:

• how to process increasing volumes of applications?

Or is it:

• why application volumes are increasing

• how incentives shape behaviour

• whether the current model of competition is sustainable at scale

AI may help manage the symptoms. But it does not address the underlying dynamics that produce them.

A research culture question

At its heart, this is not just about AI. It is about research culture.

Funding systems shape behaviour:

• how often researchers apply

• how proposals are written

• how risk is framed

• what kinds of work are prioritised

Changes to those systems—whether through policy, process or technology—have cultural effects.

If AI becomes embedded in both the production and processing of applications, it will influence not only efficiency, but:

• what kinds of research are proposed

• who feels able to compete

• how success is defined

A final reflection

The use of AI in research funding is understandable. Funders are responding to real pressures, and exploring tools that may help manage them.

But this is not just a technical shift. It is a moment that requires careful reflection.

• What behaviours are we enabling, and what behaviours are we reinforcing?

• Who benefits from these changes, and who is left at a disadvantage?

• Are we solving the right problem—or simply making the current system more efficient?

Because ultimately, how we design funding systems shapes research culture.

And if we are not careful, tools introduced to manage scale may quietly reshape the system in ways that are efficient—but not necessarily equitable.