Generalist LLMs like ChatGPT got us excited. But investment research needs something purpose-built — workflow solutions that understand your process, protect your data, and deliver outputs your analysts can actually use.
How We Got Here: From Macros to AI Tools to LLMs
If you have been in this industry long enough, you remember the early days of automation. It started with Excel macros. Someone on the team figured out how to auto-populate a comp table, or pull pricing data into a model with a single keystroke. It was not glamorous, but it saved hours.
Then came the AI wave. Firms started building or buying AI tools: scripts that could extract financial data from PDFs, parse SEC filings and transcripts. These were narrow, task-specific tools. They did one thing and they did it well.
We are now in a phase of generalist LLMs like ChatGPT and Claude, and suddenly, a single model could do most of what those individual tools did, and more. Analysts are using them as a Swiss Army knife — summarising transcripts, drafting notes, brainstorming theses, doing quick calculations. And it works. Sort of.
The Problem With “Sort Of”
If you are an analyst, you already know this feeling. You open ChatGPT, paste in a transcript, get a decent summary, then spend the next 15–20 minutes reformatting it into your firm’s template, cross-checking the numbers against the filing, fixing the hallucinated data points you almost missed, and copying everything into the right Excel tab. The AI saved you an hour of reading, but cost you an hour of cleanup.
If you are a decision maker, the problem looks different. You gave your team access to ChatGPT Enterprise or Claude. But three months in, you are asking: what is the actual ROI? Your analysts are using it, sure, but the deliverables still take time. Quality has not measurably improved. And your compliance officer just flagged that someone pasted client financials into a public model.
The Next Step: Purpose-Built Workflow Solutions
The real shift should be from generalist AI to purpose-built investment research workflow solutions. A generalist LLM takes a prompt and gives you a response. A workflow solution takes your process end-to-end, and builds AI into every step of it. The difference is not cosmetic — it is structural.
Consider what happens when an equity research analyst needs to produce an earnings update. With a generalist LLM, the analyst copies the transcript, asks for a summary, then manually pulls numbers, compares estimates, drafts in Word, and formats. With a purpose-built workflow solution, it ingests the transcript, extracts key metrics, compares against your model, flags beats and misses, updates estimates, generates tables in your template, and drafts the report — all in a connected pipeline. Every claim is cited back to the source.
The Data Security Problem Decision Makers Cannot Ignore
When an analyst pastes client financial data into ChatGPT or any public LLM, where does that data go? Into a model hosted on someone else’s servers, governed by someone else’s data policies, potentially used for training future models. Banning AI outright puts you at a disadvantage. Allowing unrestricted use puts your clients at risk.
A purpose-built system can be deployed on your own infrastructure or in a private cloud environment. The data never leaves your controlled environment. Client financials stay within your walls. NDA compliance is built into the architecture, not left to individual analyst discipline.
The Bottom Line
The journey has been clear: macros to task-specific tools to generalist LLMs. Each step was a leap forward. But the next step is not a bigger, better chatbot. It is a purpose-built system that understands investment research workflows from end to end.
If you are an analyst, the tools are about to get dramatically better, and the ones who learn to work within structured AI workflows will have a significant edge. If you are a decision maker, generalist LLMs were the proof of concept. The question now is whether you move to the next stage or watch your competitors do it first.