Emblem's August Kiles on how AI turns 100-hour data room reviews into seconds-long answers, why funds reach a no faster, and why the model beats the tool.
AI is turning the dozens or hundreds of hours associates spend combing through a data room into questions answered in seconds. That is how August Kiles, head of product at Emblem, described it on Episode 31 of Built This Week. Emblem indexes every file in a fund's data room and handles search, financial extraction, financial models and reports, so private equity and growth equity teams can rule deals out in hours instead of a hundred hours.
Before the interview, Sam showed a demo he built with Emblem in mind: a portfolio scenario tool. It assumes a portfolio of 30 seed companies. You pick a macro scenario, such as rising regulatory pressure, then set how fast it changes, whether it is temporary, how market sentiment moves and a time horizon of six, twelve or twenty-four months. AI sorts each company into groups that look unaffected, in an advantage zone, in a stress zone or worth watching.
Sam was clear that it is a demo, something Emblem could layer on later. See the portfolio macro scenario simulator build page.
Watch August Kiles on AI due diligence in Episode 31 (starts at 6:31) · Episode 31 notes and transcript
August says the goal is a one stop shop for anyone in growth equity, private equity or family offices, tracking deals across the whole life cycle of a fund. It starts the second a data room comes in the door: you sync it with Emblem and it indexes all your files. From there it offers search inside a single deal and across every deal ever uploaded, report generation that exports to Word, PowerPoint and Excel, help building financial models, and extraction of the financial data in the data room.
Jordan asked which part of the market it fits best. August says it is useful across the board, but its value rises with stage. Seed investing runs on instinct and simpler models. Private equity data rooms can hold thousands of files and complex financial models with years of data, which he calls the ideal use case.
August described the old way. A private equity firm considering an acquisition gets a data room from the CEO with everything needed for due diligence. Before AI, the only option was to have associates comb through all of it, which he says meant dozens if not hundreds of hours of work.
The level of detail does not change with AI. What changes is the time it takes. If you need a specific revenue number and do not know which spreadsheet holds it, you ask Emblem something like where the 2017 revenue is for a particular product, and you have it in seconds. Reporting goes a step further: give it a list of metrics and the analysis you want, and the results come back just as fast. August says associates still do the last 5%: deciding what to ask and presenting the answers inside their firm.
Sam asked whether a large fund would still need as many analysts. August comes down on the side of AI freeing analysts to do more. What prospective customers like early in the sales process is getting to a no faster.
So with Emblem, they can rule out investments way faster than before.
— August Kiles, Episode 31
Deals a fund actually does may take just as long, because firms stay detail oriented when real money goes in. The gain is in the pass decisions:
But if you can say no in, say just a few hours of work instead of a hundred hours of work, then you can look at that many more companies.
— August Kiles, Episode 31
His reasoning is about fund economics. Junior employees are not very expensive relative to management fees and do not move fund returns. What moves the needle is finding the right deals, and reviewing more companies means more chances to find a few diamonds in the rough. More on this industry on our AI in finance page.
Emblem runs a multi-agent setup that routes each task to the right model. August says it uses Claude and Gemini the most in the product, plus Grok and OpenAI. He later added that xAI sits mainly in a dropdown for users who prefer it, not in the backend.
On coding, the team leaned on AI too early. Around the start of 2025 they relied on models more than they should have, shipped too many bugs during a big refactor, and rolled it back over the summer. They spent most of the rest of 2025 relying on AI less. With the model updates available at the time of recording in February 2026, August says a majority of Emblem's code is again AI generated in some capacity, and the newest Opus had become very good at generating a financial model and outputting it to a spreadsheet, a big performance bump for the product.
The news segment started with reports that Amazon engineers wanted Claude Code instead of the internal tool, Kiro. Jordan said friends at Amazon told him a few months earlier that they were several OpenAI models behind, and one friend told him Kiro runs an older version of Opus in the background. At the time of recording, developers wanted Opus 4.6 because of the jump from 4.5. Jordan thinks the gap between a model launch and enterprise adoption is shrinking, and that enterprises may soon get new models before the public.
August called that Opus release an inflection point in his own work. When he does a good job on architecture and prompting, he rarely needs to adjust the code the model writes, which he says was different a year or even four months earlier. His contrarian view was about the tool:
At the end of the day, everything out there is really good and using, say Cursor or Copilot with Opus 4.6 is going to be so much better than using Claude Code with Opus 4.5 that it's not even funny.
— August Kiles, Episode 31
He still uses Cursor for its in-editor browser for UI tweaks and its debug mode. He applies the same logic to Emblem's competitors: everyone uses the same models, so the product and the user interface decide who wins. Sam agreed that the model, not the IDE, changed his workflow. We break down the tools on our Claude Code vs Cursor page.
August's guess is that Elon Musk is less interested in being an API that companies rely on for basic workflows, and more interested in a full stack digital human that can do any work a person can, the digital Optimus he has talked about. Jordan's view is more practical: if developers at xAI are asking to code with Claude or Codex instead of their own tools, that alone could be enough to prompt a management change. He added that we do not run Grok models in production. August said Emblem does not use xAI on its backend either.
August walks through Emblem and Sam demos the scenario tool on screen. Watch Episode 31 of Built This Week, where the full transcript is also on the page, or read more about August Kiles. For another take on AI due diligence, see Episode 47.
