Investor research is the most universal task in capital raising. Before any meeting, any email, any LinkedIn message, a fundraiser has to know who they're talking to: how the firm allocates, what they've committed to recently, which mandates are open. That research used to take 20 minutes per prospect. With the right AI setup, it now takes 20 seconds.
That shift is the point.
Investor research isn't disappearing; it's collapsing in time. And it's happening at a moment when asset-management AI adoption has moved past the pilot phase. Mercer's 2026 AI in Asset Management Survey of 131 managers (February-March 2026) found that 55% have integrated AI into at least one investment process, with idea generation and research the most common use case. SimCorp's 2026 InvestOps Report (200 buy-side executives) put front-office AI adoption at 70%, up from roughly 10% a year earlier.
This post covers what changed, what the new research workflow looks like, and what it means for fundraising teams.
What Investor Research Used to Look Like
Until recently, a fundraiser building a target list followed a familiar sequence…
Log into the allocator database. Filter by metro, AUM band, asset class preference. Export the results. Open the CRM to check existing coverage. Open LinkedIn to map relationships. Open news sources to check for recent commitments. Open a spreadsheet to dedupe and prioritize.
Each step was straightforward. Stacked together, they ate the better part of half an hour per query. For a fundraiser running 20 to 30 targeted queries a week, that adds up to eight or ten hours, before a single email is drafted.
The work itself wasn't optional. Showing up to a meeting unprepared is the fastest way to lose an allocator.
But the workflow had built up enough friction that the prep often happened in the margins: late at night, between calls, on the train. Research wasn't the bottleneck because it was hard. It was the bottleneck because every step required a different tab, a different login, a different export format.
What Changed
Two things happened in parallel.
The first is that large language models got good at structured data tasks. Asking an AI assistant a question and getting back a clean, filtered list stopped being a parlor trick.
The second, and more consequential, is that Anthropic introduced an open standard called the Model Context Protocol (MCP). MCP lets an AI assistant connect directly to live data in real time, then return answers in plain English. There's no integration code to write, no developer required, no batch export to manage. The data stays in the source system; the AI assistant queries it on demand. This is a different thing from an API, though APIs are still doing the heavy lifting underneath.
The combination of these two shifts, capable LLMs plus a standard for connecting them to live data, is what makes 20 seconds possible. A fundraiser opens Claude, ChatGPT, or any compatible AI assistant. They type a question in plain English. The AI assistant queries the allocator database through an MCP server and returns the answer.
The New Workflow in Practice
A worked example. A fundraiser at a private credit fund is preparing for a Texas trip and wants a target list of family offices likely to commit.
The old workflow: log into the database, filter by location (Texas), filter by allocator type (family office), filter by asset class (private credit). Export to Excel. Cross-reference with CRM for existing relationships. Open LinkedIn to find investment-team members. Search news for recent private credit commitments. Build the list manually in a spreadsheet.
The new workflow: open the AI assistant. Type "pull a list of family offices in Texas that allocate to private credit, with investment-team contacts." Get the list in seconds. Then, in the same chat: "draft an intro email to the top five." Then: "add these contacts to my CRM."
The fundraiser never leaves the AI tab. The research, the drafting, and the next step are all in one place. Total time elapsed: under two minutes.
That's what the 20-minute to 20-second shift actually looks like. The query that used to require multiple platforms, multiple exports, and a manual list-building step is now a single sentence. Learn how to connect Dakota Marketplace to your Claude here. For more examples of what to ask, see How to Use AI to Research Institutional Investors.
To walk through the workflow live with your own queries, book a demo of Dakota Marketplace.
What This Means for Fundraising Teams
The research itself isn't going away. Relationships still have to be built, meetings still have to happen, trust still has to be earned in person. What's changing is the ratio.
When research is 30% of the working week, fundraisers spend most of their time at desks. When research is 3%, they spend most of their time in front of allocators. That's a meaningful shift in how time gets allocated and how teams get structured. The most-cited measurable benefit of AI adoption in Mercer's 2026 survey was enhanced operational efficiency, named by 69% of respondents. Faster, higher-quality insights came in second at 55%. Translated to fundraising, that's more conversations, better-prepared conversations, and a tighter feedback loop between research and outreach.
For teams built around research-heavy junior fundraisers, this also changes the math on staffing. Less of the day is data-gathering; more of it is relationship-building. The skills required at the entry level shift accordingly.
The broader move, and this is what asset managers report seeing across the industry, is from AI as experiment to AI as workflow. Dakota's view: this is the direction allocator data and fundraiser workflows are heading, and an MCP server is how investment firms get there without writing code or rebuilding integrations.
Whether you're pulling family offices in Texas, endowments in the Northeast, or pension funds raising private credit allocations, you can query the data the same way you'd ask a colleague: in plain English.
Book a demo to see how Dakota fits into your AI-driven fundraising workflow.
