Council uses AI to strengthen grants program

A fast-growing New Zealand council is experimenting with artificial intelligence (AI) across its grants programs, using the technology to classify data, summarise applications, verify information and model funding decisions.

Queenstown Lakes District Council distributes around $2.25 million in grants annually across programs including its Community Fund, Events Fund, arts schemes and heritage grants.

Community investment advisor Giovanni Stephens told Grants Management Intelligence that grants formed part of a broader community investment approach, which also included community leases, venue hire and fee waivers.

“Grants are like one of many tools in the toolbox,” he said.

The council’s grants processes were once fragmented, with applications arriving via handwritten forms, emails, web forms and spreadsheets. About two years ago, it centralised its grantmaking through SmartyGrants, coinciding with the emergence of accessible large language model (LLM) tools.

Since then, Stephens and his colleagues have explored a range of AI applications while maintaining strict privacy safeguards.

One of the council’s earliest uses of AI was to convert years of unstructured application data into usable categories.

“We used LLMs to classify that qualitative data and turn it into categorical information,” Stephens said, “effectively backfilling our standard fields.”

AI has also proven valuable for summarising applications.

Faced with processing around 100 applications for assessment panels, Stephens developed a reporting system that generated concise summaries, one-page overviews and assessments against funding criteria. What once required significant manual effort is now much more efficient.

The council also uses AI-powered research tools to gather publicly available information about applicants, helping staff verify claims, enrich records and provide additional context for assessors.

The most sophisticated use case involves funding allocation modelling. Using assessors’ scores, historical funding data and a range of constraints, including budget limits and regional balance requirements, Stephens developed a system designed to maximise overall community impact.

When tested against a previous Community Fund round, the model’s recommendations closely matched the council’s final funding decisions.

Despite the successes, Stephens said AI had not performed well in merit-cased grant assessment.

“In our experimentation, we’ve found that it’s not particularly strong on that front,” he said.

“There’s a lot of latent information that comes into an assessor’s mind – relationships they hold, things they’ve seen from the group’s work.” He said human assessors remained central to evaluation.

The council also avoids relying on a single AI platform, noting that the technology is evolving rapidly and leadership among providers changes frequently.

Privacy remains a key consideration. Applicant data is anonymised before being processed, and consent is sought when AI-assisted research is used.

Stephens believes many organisations are still grappling with the legal and ethical implications of AI.

“I suspect quite a few organisations in New Zealand might be violating the Privacy Act with what they input into an LLM, and it is an underdeveloped area of legislation.”

For grantmakers reluctant to experiment with AI, he recommends starting with low-risk testing using synthetic data.

“You can use invented data, invented applications, and use it to fine-tune your processes anyway,” he said.

The council has also observed widespread AI use among grant applicants.

Stephens estimates that well over half of applications show signs of AI-assisted drafting. While some submissions are clearer and more concise, others have become longer without providing meaningful information.

“You get a lot of words that say very little,” he said.

In response, the council runs workshops to help community groups use AI effectively and avoid obscuring genuine achievements behind generic language.