Financial Modeling, AI
From 370 Files To One IC Memo In 24 Hours: A Case Study In AI First Due Diligence
Most lean Series A and B funds are drowning in pitch decks, data rooms, and messy Excel models while trying to run “real” diligence with tiny teams. In this case study, we walk through how one investor used Acephalt to turn a 370 file data room into a full IC memo and model review in under 24 hours, with every claim traced back to source documents.
By combining multi agent AI, a specialized financial agent on top of Excel, and customizable workflows, Acephalt helps VCs and family offices save hours on each deal while actually improving the depth and consistency of their underwriting. The result is simple: less time playing detective in spreadsheets, more time building conviction and talking to founders.
From 370 Files To One IC Memo In 24 Hours: A Case Study In AI First Due Diligence
If you are running a Series A or Series B fund or a lean family office, your job is simple to describe and brutal to execute.
You need to find real product market fit signals, make conviction bets, and do it all with a small team that is drowning in pitch decks, data rooms, Excel models, and stakeholder expectations.
In a recent product session, Acephalt founder Winnicent Zuo walked through the platform with Josh, a partner at a US VC backed by a large insurance company. Josh invests across proptech, fintech, insuretech, auto, and enterprise software. His challenge is the same one many investors face today:
Too many deals
Too much data
Not enough time or headcount
And generic AI tools that are helpful for pitch decks, but fail miserably on real financial models
The Core Problem: Diligence Did Not Modernize When Deal Flow Did
Most lean investment teams still run diligence in a way that would look familiar in 2010.
Data is scattered across email, CRMs, shared drives, and vendor portals.
Analysts spend days inside Excel, clicking through formulas and trying to understand why revenue jumps 40 percent in December or why churn suddenly spikes in Q3.
IC memos take so long to produce that by the time they are ready, the deal has often moved on.
As Dr. Paul Newton summarized when profiling Acephalt, founders are moving at AI speed while investors are still operating at the pace of manual spreadsheets and document review. (UpNext.World)
The result is:
Slow deal cycles
Inconsistent depth of analysis from deal to deal
Teams forced to choose between speed and diligence quality
Josh described one of the most painful parts of that process, which every investor reading this will recognize. When he gets a model from a founder he spends hours as a “detective” in Excel:
Hitting F2 to see where every key number comes from
Tracing assumptions that affect revenue, margins, and headcount
Hunting for yellow flags like unrealistic margin expansion or churn that suddenly disappears in the forecast
ChatGPT can help summarize a pitch deck or generate a list of generic risks. It cannot reliably parse a live Excel model with complex formulas, edge cases, and messy real world data and then tell you:
“Here is where the assumptions break.”
“Here is why December margins jump 15 percentage points.”
“Here are the five lines in this model that you must ask the founder about.”
That is exactly the gap Acephalt is trying to close.
Redefining Product Market Fit For AI Diligence Tools
When Winnicent asked Josh what would convince him that Acephalt itself had product market fit, the answer was revealing.
It was not “20 signed logos” or “nice testimonials on a website.”
Instead, Josh said he would be excited if Winnicent could say something like:
“We got in with a 10 person fund. Over the last month they ran 20 deals through Acephalt. They started with 1 or 2 users, now most of the team is onboarded. Everyone touches the platform at least once every two weeks and half the team is using it every couple of days.”
In other words, for a tool like Acephalt the signal is not just adoption, but depth:
How many deals per month are actually being processed through the system
How many people on the team rely on it in their day to day
Whether it becomes part of the core toolkit, not a trial that quietly dies at renewal
This framing directly shaped how Acephalt thinks about success. It is not enough to auto generate a nice memo once. The platform has to sit in the real workflow across sourcing, screening, deep diligence, and post investment monitoring.
Why “Design Partners” Are Not Just Fancy Beta Users
In the last few weeks, Acephalt shifted from “any investor who is curious” to a very focused ICP:
Series A and B investors and US family offices
Teams with fewer than 10 people
Funds that lead or co lead rounds, write real checks, and receive hundreds of decks per year
These are the teams that are most squeezed. Their check size demands deep work, but they do not have a 30 person diligence engine.
Within that ICP Acephalt is intentionally working with a small group of design partners instead of a broad pool of casual beta testers.
As Winnicent put it in the call:
A beta user is a free tester. They will say things like “can you make this button bigger” or “I prefer this color,” which is useful but shallow.
A design partner says “even if you make this feature perfect, I will not use it. This is not my real problem.”
Design partners:
Run real deals through the platform
Share their actual models and data rooms
Give daily feedback on what creates value and what is noise
That is how Acephalt ended up with investors in consumer, AI, healthcare, and other verticals who are shaping the product around real workflows, not imagined ones.
Enter Acephalt: Multi Agent AI For Diligence You Can Trust
Acephalt is purpose built for investors, not for generic document summarization. At its core, the platform does three big things. (Acephalt)
Automates the heavy lifting in deal screening and due diligence
Upload a full data room and the system ingests financial statements, cap tables, contracts, and market reports.
Multi agent AI teammates clean, normalize, and link the data into a unified view.
The platform then drafts investment memos, risk matrices, and even draft term sheets aligned with your KPIs.
Acts as a specialized “financial agent” on top of Excel and raw data
Instead of a generic AI, Acephalt uses models tuned for investor tasks.
It can clean transaction level data, build revenue cohorts, calculate retention, and generate scenario models.
It highlights where the model behaves strangely and turns those into concrete follow up questions for founders.
Turns scattered information into traceable, auditable outputs
Every claim in an IC memo can be traced back to a specific file or source.
Risk flags are linked to underlying documents or lines in a model.
That auditability is critical when you are presenting to IC or LPs and need to show exactly where your numbers came from. (TechBeat Canada)
The goal is simple: let investors spend their time with founders instead of with file structures and lookup formulas.
A Real Deal: 370 Files To A Full IC Memo In Under 24 Hours
In the call, Winnicent shared a concrete example from a live Series B process.
The data room contained more than 370 files.
The team used Acephalt to ingest the full room.
The platform generated:
A structured IC memo covering problem, solution, market, product, team, IP, go to market, competition, traction, and scalability.
A visualization of the cap table showing top holders and their positions.
A set of diligence action items, such as “request the founder’s source and assumptions for market size” where the founder’s claims and third party data differed.
The key point is not that Acephalt produces a “perfect” memo automatically. It is that:
The first draft is ready in less than a day rather than a week.
The memo is anchored in the actual documents rather than hand written notes.
Partners and IC members can review a structured summary with live links back to the data room instead of sifting through folders.
For a lean team, that alone can be the difference between passing on a deal due to bandwidth and running a real process on it.
Solving The Excel Detective Work
Josh’s biggest time sink is the model.
His ideal outcome from Acephalt is not just a faster forecast. It is an automatic “yellow flag radar” that tells him:
Where historical data behaves in ways that warrant questions
Where forecast assumptions quietly drive big jumps in revenue or margins
Where headcount, COGS, or other drivers are out of sync with the story
This is where Acephalt’s “financial agent” comes in.
A typical workflow might look like this:
Upload the model
The financial agent parses sheets, formulas, and references.
It identifies core drivers (pricing, volume, churn, CAC, headcount, etc.).
Run pattern checks
Highlight months or quarters with unusual spikes or drops in revenue, costs, or churn.
Flag unrealistic margin expansion or silent compression of CAC.
Separate one time events from structural trends.
Generate investor ready questions
“Between November and December, gross margins improve by 15 percent. Please walk us through the underlying drivers.”
“Revenue in Q3 grows 60 percent while headcount only rises 5 percent. How is that achieved operationally?”
“Churn drops from 8 percent to 2 percent from this year to next. What changes in product, pricing, or customer success support this assumption?”
Feed this directly into your IC pack
Yellow flags and suggested questions are bundled into the IC memo or a separate “model review” section.
You walk into founder meetings with a targeted list rather than generic prompts.
This is exactly the kind of work that generic AI cannot do well today and where a specialized, model aware agent saves real hours per deal.
Custom Investor Workflows Instead Of One Size Fits All Templates
Not all funds work the same way.
That is why Acephalt includes a workflow builder that lets you design your own “playbooks” using Lego style blocks:
Triggers like “new company added to CRM,” “data room link received,” or “pitch deck emailed to a partner”
Actions like “create folder in Google Drive,” “enrich company data from PitchBook,” “run screening report,” or “draft IC memo section”
Integrations with tools such as Gmail, CRMs, portfolio databases, market data platforms, and cloud storage
An investor might build a workflow such as:
When a new deal is logged in the CRM at “screening” stage
Create a deal workspace in Google Drive
Pull public data on the company and market
Generate a one page screening memo with key metrics and initial risks
Send it to the deal partner by email and Slack
Or at “data room received” stage:
Ingest the entire data room
Run the financial agent on uploaded models
Generate a first draft IC memo
Prepare a list of founder questions based on flagged assumptions
Update a deal scoring dashboard based on your custom KPIs and sector benchmarks (Acephalt)
For a Series A or B fund with a 5 to 10 person team, this is effectively a digital analyst that never sleeps.
What This Means For Lean VCs And Family Offices
For Josh and funds like his, the value of Acephalt is not in buzzwords. It is in concrete outcomes that show up in day to day work:
Hours back per deal
Early adopters of Acephalt’s multi agent architecture report saving hundreds of hours of manual review across deals, with some estimating more than 1,000 hours of diligence time reclaimed. (UpNext.World)
More consistent underwriting
Every deal gets a full, structured memo and risk review, even when the team is stretched.
IC packs are built off the same backbone instead of ad hoc templates in each analyst’s folder.
Better founder conversations
Instead of asking broad questions the AI could have generated from a deck, partners go straight to targeted follow ups on assumptions that matter.
This improves founder experience and lets both sides spend their time on real strategic discussion.
A stronger story for LPs
When LPs ask “how do you systematize your process” or “how do you avoid missing risks,” having an AI infrastructure that logs sources, flags, and decisions is a real differentiator.
And at a strategic level, what came through Josh’s feedback is this:
If an AI tool becomes part of the core toolkit for a few lean, sophisticated teams and they use it every week across real deals, then it is no longer a nice to have productivity hack. It is part of the firm’s edge.
How To Leverage Acephalt In Your Own Investment Workflow
If you are running a Series A or B fund or a family office with a lean team, there are three practical ways to think about Acephalt.
Start with one or two live deals
Pick a current data room that is heavy on Excel and unstructured documents.
Use Acephalt to create the first memo draft and run the financial agent on the model.
Compare the AI output with your existing process and see where it saves time or surfaces new questions.
Standardize one repeatable workflow
For example, your initial screening process or “data room received” stage.
Build a simple playbook that triggers IC memo drafts, model checks, and risk flags.
Roll that out across the team and track usage.
Grow into a design partnership
If you see value and have strong feedback, work with Acephalt as a design partner.
This gives you influence over the roadmap and preferential economics while the product matures.
You help shape how AI diligence will work for funds like yours for years to come.
Final Thoughts: Moving Investors To AI Speed Without Losing Discipline
The conversation with Josh crystallized what Acephalt is really trying to build.
Not just a faster memo generator. Not an AI toy that analysts try once and abandon at renewal. But a multi agent system that:
Understands how investors think
Handles the messy, time consuming work behind every IC memo
Surfaces the yellow flags that truly change whether you wire or walk
In a world where founders are already using AI to move faster, investors who do not upgrade their own workflows risk being left behind.
If your team wants to spend less time clicking through models and more time talking to founders, Acephalt was built for you.
You can learn more about the platform and request access directly from the Features and About pages on the Acephalt site, or reach out to explore becoming a design partner while the team is still working closely with a small group of funds. (Book a Demo With Us)