Most mornings at an asset management firm start the same way: a stack of overnight broker notes, a few earnings transcripts, internal research emails, and a portfolio that may or may not be affected by any of it. Mpho Montsho has spent the past few years building software to change that.
Montsho, co-founder of Techbridge Innovations, a US-based software development house, came to the problem from the banking side. As an investment banking analyst in Deutsche Bank's Financial Institutions Group, he worked on transactions including the initial public offering of digital asset platform Bullish and the sale of insurtech company Vouch to Hiscox. A lot of that work, he says, was assembling information before anyone could form a view on it.
"You'd spend days pulling numbers into a model, checking them, formatting them, and then the actual discussion about whether the deal made sense might be an hour," he says. "I wasn't complaining about it at the time. It's just how it worked."
He left banking to co-found Techbridge, where he leads product, operations and the delivery teams. The company designs and builds AI and software platforms for its clients, many of them in investment and asset management.
The system Montsho talks about most is an AI analyst agent built for a client's portfolio management team, running on Google Cloud's Vertex AI. Montsho was the product manager on the build: he worked with the firm's portfolio managers to define what the agent should and shouldn't do, set the product roadmap, and steered the engineering team through delivery. It connects to the firm's research library, email and portfolio data, and portfolio managers ask it questions the way they would ask a junior colleague: what changed in this company's guidance, which notes this week mention a particular holding, whether anything contradicts the original reason for owning a stock. The first version was delivered in three milestones, and the client's portfolio managers now use it daily.
"It's become part of how we start the day," said a portfolio manager at the firm, who asked not to be named because the firm does not publicly discuss its technology vendors. "Before, you'd skim everything and hope you hadn't missed something. Now I'll ask what's changed on our holdings overnight and go from there. It doesn't make the call for us, but it means we're having the right conversation sooner."
The design principle was Montsho's, drawn from his time in banking: the agent supports the decision rather than making it. "Portfolio managers don't want something that makes decisions for them. Honestly, they'd stop using it," he says. "What they want is to not miss things. If there's a paragraph on page 40 of a filing that matters, they want to know about it before the meeting."
He is now leading the second phase, which pushes further into decision support: flags when new information appears to undermine an investment thesis, tracking of how conviction in a position shifts over email threads, structured notes generated from recorded calls, spreadsheet and quantitative analysis inside the chat interface, and benchmark tracking.
Montsho expects the analyst agent to be the start of something larger. He sees asset managers moving towards bespoke AI agents built around their own investment process: agents that know a firm's holdings, its research and its house views, and that can test how new information should change portfolio composition. "Every manager invests differently," he says. "A generic tool doesn't know your process. The value is in building agents around how a specific team actually makes decisions, and then helping them make those decisions faster and with more of the evidence in front of them."
Montsho is currently leading the second phase of the analyst agent, alongside new client projects in the investment sector.