AI in Fund Administration - Lessons From Private Market Gladiators

By Ivo Schmiedt (Vestlane) · Stevan Marković (AQVC) · Davendra Patel (Alter Domus)

Private Market Gladiators: Episode 3 - AQVC's fund perspective, Vestlane's tech‑legal expertise, plus rotating insights from top investors, regulators, and influencers.

When it comes to artificial intelligence, fund administration is where talk meets reality. It is the place where subscription packs, KYC evidence, waterfall models, and NAV workbooks either move smoothly - or grind the entire vehicle to a halt.

In Episode 3 of Private Market Gladiators, host Ivo Schmiedt and co‑host Stevan Marković sat down with Davendra Patel, Head of AI & Automation at Alter Domus, to find out how AI is actually being used inside one of Europe’s largest fund administrators.

Patel isn’t theorising; he’s shipping. His team runs 65 live bots and an in‑house GPT (“Domus AI”) with 3,200 sign‑ups and 2,000 active users. Together, the trio is dissecting what works, what doesn’t, and what every GP or administrator can copy tomorrow.

Why Fund Admin Is Ground Zero for AI

"I have qualified people doing Excel work. Come on.” - Davendra Patel

Private‑market operations face a brutal equation:

Regulations multiply (EU AML package, EU AI Act).
LPs demand faster, cleaner data (capital‑call notices in minutes, not days).
Head‑count budgets stay flat.

Without automation, the back office becomes a choke‑point. As Stevan Marković put it:

My job here at AQVC is to eliminate friction between LPs and GPs … Fund admin is where that friction shows up first.

Three Real‑World Wins From Alter Domus

1. KYC False‑Positives: From 6 Hours to <30 Minutes

“I can’t believe people actually type the name into Google … The bot does it all in seconds now.” - Patel

Problem FinScan hits and Google scrapes consumed up to six analyst‑hours per investor batch.
Automation A UiPath bot ingests an Excel list, runs FinScan + adverse‑media searches, and exports a PDF decision pack.
Impact 4‑6 hours saved, false‑negative risk slashed, audit trail generated automatically.

2. NAV Reconciliation: 11 Sheets → 20‑Minute Cycle

“We replaced 11 spreadsheets and three weeks of work with a 20‑minute pipeline.” - Patel

Problem NAV teams juggled FX sheets, capital‑call trackers, and fee models, all linked by fragile VLOOKUPs.
Automation Python + Azure Data Lake validate headers, cross‑check FX tables, and feed a Power BI dashboard; an RPA wrapper triggers the run.
Impact NAV cycle cut from 3 weeks to ≈20 minutes. Reviewers now focus on anomalies, not copy‑paste chores.

3. Domus AI: A Private GPT for 3,200 Users

“We’ve got 3,200 people signed up - about 2,000 use it actively each month, and it’s changed the company from within.” - Patel

Architecture Azure OpenAI in a tenant‑isolated environment; prompts never leave EU data centres.
Everyday Uses Draft investor letters, extract waterfall clauses, convert PDF capital‑calls into JSON, and even debug legacy C# code.
Cultural Win “People are happier when they’re not doing mind‑numbing work until midnight.” - Patel

Flywheel Effect: “Once one bot goes right, they give us more work.” - Patel

Governance & Guardrails - Patel’s Playbook

“Compliance is the cost of entry. Build guardrails first - then innovate fast inside them.” - Patel

Guardrail	AI Committee
How Alter Domus Does It	Legal, risk, security, and ops meet monthly.
Guardrail	Human‑in‑the‑Loop
How Alter Domus Does It	No output reaches a regulator or LP before dual human review.
Guardrail	Audit Logs
How Alter Domus Does It	Every prompt & response stored immutably for 10 years.
Guardrail	Data Residency
How Alter Domus Does It	EU data centres; no public SaaS for PII.

Guardrail	How Alter Domus Does It
AI Committee	Legal, risk, security, and ops meet monthly.
Human‑in‑the‑Loop	No output reaches a regulator or LP before dual human review.
Audit Logs	Every prompt & response stored immutably for 10 years.
Data Residency	EU data centres; no public SaaS for PII.

Ivo Schmiedt underlined the point:

We’re not replaced by this. AI should lift the friction away.

Talent Shift - Hiring for 2025, Not 2015

“What a data scientist was before ChatGPT is not a data scientist today.” - Patel

Patel now screens for:

Curiosity over Credentials Low‑code prototyping beats academic papers.
Explain‑Ability Can they translate embeddings to a fund accountant?
Ownership Small teams, zero micromanagement.

Upskill first: train existing accountants and KYC analysts in prompt engineering before chasing six‑figure ML résumés.

Fund administration is ground zero for AI’s impact on private markets. The tech is already chopping cycle times and error rates while keeping human judgment front and centre. Firms that start piloting now will reap those gains; laggards will still be reconciling 11 spreadsheets by hand in 2026.

What Is Private Market Gladiators?

Private Market Gladiators is a no-filter discussion series by Vestlane and AQVC, featuring raw conversations with the people shaping private markets—from GPs and LPs to regulators, tech builders, and lawyers.