The philanthropic arm of a global asset manager partnered with Promptworks to apply AI to one of its most demanding internal workflows: the twice-yearly process of writing disbursement recommendations to its board.

The foundation funds high-impact organisations across Asia, with a focus on vulnerable and marginalised communities. Its small team produces up to 15 partner reports each cycle, each one informing a board-level recommendation on funding. The work is high-stakes, judgment-heavy, and concentrated into two peak seasons in Q1 and Q4.

The team came into the engagement following the departure of a team member. With one fewer set of hands and a busy disbursement cycle approaching, the team wanted to test whether AI could lift some of the load before hiring a replacement, and build practical AI fluency across the team at the same time.

Each disbursement cycle required the team to read, review and synthesise a large body of partner material before drafting concise, board-ready recommendations. The bulk of the work was research and analysis, comparing current performance against prior proposals and historical context. The writing was comparatively short, but it had to be precise, tightly argued, and grounded in years of relationship knowledge.

Three connected pressures shaped the engagement. Reduced capacity following the team departure left the group operating with less bandwidth than usual. The twice-yearly disbursement cycle created sustained workload spikes in Q1 and Q4. And the team had a wider ambition to move from ad hoc AI use into something more confident and habitual, with shared standards and shared language across the team. A senior team member had also picked up responsibilities outside her existing skill set, sharpening the case for systems that could compress review time and surface insight more efficiently.

Promptworks combined two strands of work: hands-on AI training and the build of two purpose-fit AI agents anchored to the disbursement workflow.

The team went through a tailored prompt bootcamp covering structured prompting, output evaluation, and the foundations of treating AI as a thought partner, not a content factory. The training built shared standards and the habits the team needed to embed AI in how the team works day to day.

Alongside the training, the engagement delivered a multi-agent framework built around the disbursement workflow itself. A synthesiser agent took on the rigorous review and analysis of partner reports, including year-on-year comparisons against prior proposals. A disbursement writer agent drafted the one-page board recommendation off the synthesised outputs. The two agents were designed to work in sequence, mirroring the team’s natural workflow. The synthesiser does the heavy reading and surfaces the headline findings; the writer translates those findings into the structure of a recommendation. The split made the underlying tasks, research and writing, clearer to manage and easier to improve over time.

The clearest result was speed and rigour at the research stage. The end-to-end disbursement process moved from three days of work to two, a 33% time saving. More important than the raw hours saved was the value added from using these agents. The synthesiser agent handled the volume of comparison and review that the team would not realistically have completed at the same depth manually.

The team described the synthesiser as more rigorous than they could be on their own under time pressure, noting that the agents can take in volumes of data that would otherwise take days of human review. That mattered especially for newer team members coming in without years of historical context, who could now get up to speed on partner relationships and prior funding decisions quickly.

The disbursement writer agent provided drafts that were valuable starting points for the team to enhance with their institutional knowledge and individual tones of voice. It’s a classic use of generative AI to get a project started for the human to add critical knowledge that isn’t deliberately captured or easily visible to the tooling. As one team member put it, the agent was “a really good thought partner in what the final piece would be.”

Beyond the disbursement workflow itself, the engagement shifted how the team thinks about AI. After the prompt bootcamp, AI use became habitual rather than experimental. The team began evaluating other workflows for AI suitability and is now exploring AI capacity-building workshops for its partner organisations, extending the value of the work into its broader network.

– Director of Operations & Projects

For this philanthropic team, the Promptworks engagement turned a high-stakes, twice-yearly bottleneck into a faster, more rigorous workflow, without removing the human judgment the work depends on. A 33% time saving on the disbursement cycle is the headline number. The deeper outcome is a small team that now uses AI confidently and habitually, with a clearer view of where AI adds leverage and where human judgment remains essential.

For organisations weighing AI adoption in judgment-heavy work such as philanthropy, advisory, research, and policy, the case shows what a pragmatic intervention looks like. Structured training, agents built around a real workflow rather than a generic one, and a clear understanding of what should stay human. AI in the lead where it earns its place; humans in the lead where they always should be.