How a leading APAC asset manager scaled research across 1,000+ companies with Pascal.

A leading APAC asset manager overseeing more than $70 billion in assets came to Pascal with a problem that was surprisingly straightforward to describe but difficult to solve. Their investment team of roughly thirty analysts was responsible for covering well over a thousand listed companies between them. That leaves every analyst maintaining a live view on something like thirty to fifty businesses at any point in time.

Maintaining a live view is far more demanding than it sounds. In practice, the analyst in charge needs to constantly update their understanding of the business, its industry, competitors, what management is saying, and whether every new piece of information changes the original investment thesis. Every listed company leaves behind a growing paper trail, and every quarter adds another layer to it through reports, results, investor presentations, conference call transcripts, and exchange disclosures. Before AI, the only way to keep up was to read, extract, condense, and document everything by hand. That was the problem they came to Pascal with.

The firm's research process relied entirely on manual workflows. Analysts spent a disproportionate amount of their time on undifferentiated work: reading filings, pulling numbers, updating models, and documenting changes by hand. The team was experienced and well-resourced. The actual problem was one of coverage density. With each analyst responsible for so many names, the volume of information that needed to be processed to keep every view current was simply too large for any person to keep up with in real time.

The firm had experimented with consumer AI tools, but without a structured approach to data, citations, or workflow design, the outputs were inconsistent and difficult to trust. They needed a system that could handle the mechanical work of research without compromising on the rigour their investment process demanded.

Pascal centred the solution around the natural rhythm of research inside a public market fund, which falls into two distinct seasons: earnings season and the off season.

Earnings season

If you work in finance, you know that earnings season is one of the most intense periods of the year. On a heavy day, an analyst might have five or six portfolio companies reporting simultaneously. The day can quickly turn into a race between updating financial models in real time, sitting through back-to-back earnings calls, reconciling reported numbers against expectations, and juggling a hundred other things that suddenly need attention.

Pascal built an end-to-end agentic workflow that runs alongside the analyst throughout the earnings cycle.

Before results

The evening before earnings, every analyst receives a briefing on the reporting companies they cover and the specific variables worth paying attention to.

As results are released

Once the numbers land, the agent reads every filing against the firm's existing research, the analyst's financial model, and prior expectations. Within minutes, the analyst has a clear view of what actually changed and whether any of it is material to the investment thesis.

After the conference call

Once management finishes speaking, the agent synthesises the discussion against the existing investment view in a way that captures the key takeaways automatically and lays out the obvious follow-up work.

The undifferentiated work disappears, and analysts can now spend their time on investment judgement.

The off season

The off season is when most of the primary and secondary research happens.

Primary research

This is when analysts spend time understanding businesses in far greater depth. Most of the travelling happens during this period: management meetings, plant visits, ground checks. Before an analyst travels, an agentic workflow synthesises everything available in the public domain alongside the firm's own internal context on that company. It then produces a short list of the things that are actually worth travelling to find out, and more importantly, it surfaces the questions that management has not already answered somewhere in print.

Secondary research

Secondary research is largely desk-based work, so it includes things like studying in depth how a business works, understanding its industry, and studying its comparables. Pre-AI, this meant working through a large pile of documents manually. Today, that same work can be turned into an agentic workflow. The analyst runs a company primer workflow that maps the business and surfaces the key information needed for deeper analysis.

Beyond the core research workflows, Pascal also automated several other manual processes across the firm that were creating a disproportionate amount of operational overhead.

Proxy voting

Large mutual funds own meaningful stakes across hundreds of listed companies, and once those holdings cross certain thresholds, board resolutions requiring shareholder votes start arriving on a near-daily basis.

Every request arrives with a substantial amount of supporting material, including long explanatory documents and independent proxy advisory opinions. For this customer, that often meant reviewing five to ten resolutions a day, with the analyst spending two to four hours simply reading through the material before making a decision. With Pascal, an agentic workflow does the reading: summarising each resolution, laying the proxy advisors' views side by side, and flagging anything contentious. The analyst spends their time on the decision.

Inbound email workflow

The firm also had an inbound email workflow that was consuming a significant amount of time. It was largely mechanical work that required little judgement, so Pascal replaced it with an end-to-end agentic workflow that handles the entire process automatically.