Replacing the research pile with a briefing agent

District 3 is a publicly funded startup incubator at Concordia University in Montreal, processing roughly 200 to 250 applications per month. When an application lands, an operations team member spends one to four hours on background research: scouring LinkedIn, checking founder websites, mapping the competitive landscape, looking for patents and press. They write hasty bullet-point notes and assign a program stream. Stream leads walk into panel meetings and spend their first ten minutes just orienting themselves.

The research is the bottleneck. It's tedious and inconsistent. Two people reviewing the same application can produce dramatically different notes. The quality of the research dictates the quality of the panel's decision, yet it's treated as rote admin work.

This is the kind of process that evolved before modern AI. It wouldn't exist like this if built today.

I built an AI system that eliminates the research bottleneck. When a new application is submitted, an agent autonomously generates a research brief: founder profiles with verified sources, competitive analysis, SDG alignment assessment, stream classification with reasoning, a scored evaluation rubric, risk flags, and interview questions for both operations staff and panelists, all cited to real URLs or application fields.

The human can now open an application that's already deeply researched. Operations shifts from doing research to reviewing research. Stream leads arrive at panel meetings with context instead of spending their first ten minutes catching up.

Version 1 of this system was a scripted pipeline, a chain of prompts with predetermined steps and retry logic. It broke constantly. If a LinkedIn page was down or a founder didn't have a website, the whole chain derailed. Wrong output was worse than no output, because it created more work for the ops team to correct.

For version 2, I replaced choreography with agentic design. The agent isn't given a task list. It's given an environment: knowledge to reference, tools to use, a goal to achieve, and guidelines for quality. Within that environment, the agent decides what to research, in what order, and how deep to go.

The agent operates in a loop of research, assessment, and output. It loads D3's knowledge base, reads the application, creates a research plan, then executes that plan phase by phase. After each phase, it self-assesses its confidence and decides whether to proceed, retry with different sources, or flag for human review.

Brief sections are emitted incrementally as the agent completes them, not batched at the end. This means the brief builds in real time, and observers can watch the research unfold through a live server-sent event stream.

Trust calibration. The brief is good enough that ops might rubber-stamp instead of reviewing. The self-assessment loops and human review flags resist this. The system tells you when it's uncertain, but the real safeguard is organisational. Keeping humans accountable for the decisions the system explicitly refuses to make.

Other scaling concerns include API cost management at 200+ briefs per month, rate-limiting from external sources like LinkedIn, and the need for a proper database as the file-based storage approach won't hold under concurrent load. But the trust calibration problem is the most interesting because it's not a technical problem. It's a human one. The system is designed to make human oversight easier, but it can't force people to actually exercise it.

I had been circling this idea since 2022, when it became obvious that a lot of legacy workflows were about to look embarrassingly pre-AI. District 3's admissions research process was one of them. High-volume, repetitive, cognitively messy, and still dependent on humans to stitch together context from scattered sources. It was exactly the kind of system I wanted to rebuild, but not the kind of internal project you can casually decide to spin up on your own.

The Wealthsimple application prompt was what actually made me build this; it asked for a real system, not a speculative deck, and it explicitly cared about where AI should take on responsibility and where it should stop. That was the kick in the pants I needed to finally build the thing properly.

What interested me most was the pattern, and how it can be applied to other use cases. This kind of high-volume review work with fragmented evidence, real operational pressure, and a decision boundary that still belongs to a human. That same pattern shows up in places like KYC and AML. I would not hand final judgment to the model, but I would trust it to assemble the case, surface inconsistencies, rank risk, and give a human reviewer a far better starting point with valuable time saved.