A new PE firm just raised $885M on one promise.

Not cheaper. Bigger.

Aldaron, built by Palantir and Carlyle alumni, closed its debut fund and bought a majority of Oak Point Partners. The company thinks revenue could rise as much as 5x in 4–5 years (WSJ). That's a growth claim, not a margin claim. And it's the test case for whether AI is a revenue engine in PE, or just a better cost-cutting slide.

TL;DR · The 30-second version
  • The deal: Aldaron closed ~$885M (AUM close to $1B) and took a majority stake in Oak Point, the remnant-asset buyer, in March (WSJ, press release).
  • The claim: up to 5x revenue in 4–5 years. That's roughly 38–50% a year (my math). The baseline isn't public.
  • The trend: 41% of PE firms aim portfolio AI at revenue vs. 24% at cost. But most target 5–10% gains, not 5x (FTI).
  • The catch: only 24% of PE leaders report revenue growth from AI so far (Grant Thornton). Underwrite the levers, not the label.
$885M
Aldaron debut fund
5x
Oak Point revenue goal, 4–5 yrs
41% / 24%
PE AI aimed at revenue / cost
24%
PE leaders seeing AI revenue
Section 1 · The story
Deal Intelligence

The lead: Aldaron buys Oak Point, embeds engineers

Aldaron calls itself "transformation equity." A Deployment Team of engineers and investors works inside each company. The team that underwrites the deal builds the tech and is "measured on whether it succeeds" (press release).

  • Founders: Jay Grewal co-founded Palantir's commercial business and led 600+ forward-deployed engineers. Evan Mendez was on Carlyle's US buyout tech team and an early member of its AI team. Will Oberndorf ran private markets at Oberndorf Enterprises (same source).
  • The target: Oak Point buys remnant assets, meaning intangible assets and payment rights left over at corporates and bankrupt estates. It has done 2,000+ deals since 2006 (Oak Point).
  • The AI plan: speed up discovery of remnant assets across Oak Point's asset channels and add capacity. The release says 1,000+ channels; Oak Point's site says 250+.
  • Already tech-heavy: Oak Point says its platform was built by ex-Palantir engineers and mines 1B+ data points a day (Oak Point).
  • Not disclosed: price, terms, Oak Point's revenue and the fund's LPs (citybiz). Founder-CEO Eric Linn stays, and existing investors rolled.

Why the 5x is plausible, and why it isn't proven. Remnant assets are a discovery business. If the bottleneck is finding and pricing assets, better search lifts volume directly. But 5x in 4–5 years means ~38–50% a year compounded (my math), with no public baseline to check it against.

Why it matters: this is a revenue thesis in a niche where data is the moat. If it works, the returns come from volume. If it doesn't, you've paid a growth multiple for a services business.

The services version: General Catalyst and Teddy AI

Last week's roll-up story is the mirror image. GC's $1.5B Creation Fund and Teddy AI's $60M seed buy services firms and add AI to the margin (General Catalyst, Axios). Aldaron is pitching growth first.

Section 2 · Take it
Operator Toolkit

1. How to underwrite an AI revenue case in IC. What to believe, what to haircut.

  • Start with the baseline: no audited starting revenue, no multiple-of-revenue target. A "5x" with no denominator is a slogan.
  • Find the bottleneck: AI revenue is credible when demand is there and the constraint is discovery, capacity or speed. It isn't when the market is the constraint.
  • Separate proprietary from off the shelf: pay for levers that run on data competitors lack. Off-the-shelf uplift gets competed away.
  • Check the evidence tier: shipped and measured (believe), piloted (haircut 50%), slide only (haircut 100%).
  • Run the implied CAGR: compare it to the company's 5-year history and its market. Anything well above both needs a mechanism, not a narrative.
  • Price the build: engineers, data work and time to first revenue. Embedded teams cost money, and often fees too.
  • Set kill criteria: one KPI per lever at day 100 and day 365. If a lever misses, it comes out of the model.

2. The benchmark. 65% of PE revenue-focused AI programs target 5–10% improvement; 4% target more than 20% (FTI). Only 14% of PE-backed CEOs say AI has raised revenue and cut costs (PwC). A plan far above that range is the exception. Underwrite it like one.

Deflections are findings. If management can't name the lever, it isn't in the model.

Section 3 · Paste it
AI Workflows

Both plays work in any AI tool: Claude, ChatGPT, Grok Bot or your firm's approved assistant. Paste the prompt, add your data, and save it as a Project, custom GPT or routine so you can rerun it.

Play 1: The AI revenue opportunity scan

Paste a portfolio company's basics. Get ranked revenue levers, each tagged proprietary or off the shelf, and a 100-day plan for the best one.

You are an operating partner building an AI revenue plan
for a portfolio company. Revenue only. No cost cuts.

COMPANY DATA:
- Business / what it sells: [ ]
- Revenue (last 3 yrs) and split by product/channel: [ ]
- Customers: count, top 10 as % of revenue, churn: [ ]
- Sales motion: reps, pipeline, win rate, cycle length: [ ]
- Pricing: how set, last increase, discounting: [ ]
- Data the company owns that competitors don't: [ ]
- Capacity limits (people, hours, backlog): [ ]

Tasks:
1. List up to 8 AI revenue levers in four buckets:
   A) Find more (lead/asset discovery, new segments)
   B) Win more (conversion, cycle time, pricing)
   C) Keep more (churn, upsell, cross-sell)
   D) New products built on the company's own data
2. For each: the mechanism in one line, the metric it moves,
   a low/base/high revenue impact using ONLY my numbers,
   time to first revenue, and what must be built.
3. Mark each lever "proprietary" (depends on data/process
   competitors lack) or "off the shelf" (anyone can copy).
4. Rank the top 3 by value x speed x defensibility.
5. Write a 100-day plan for the #1 lever with owner, KPI
   and a kill criterion.

Rules: show the math for every estimate and label it an
estimate. If an input is missing, write [VERIFY] and do not
guess.
  • Feed it: the latest board deck, a CRM export or the 100-day plan.
  • Run it across the portfolio: one company per chat, then ask for a ranked list.
  • Keep it private: use an enterprise or firm-approved workspace for company data.

Play 2: The IC stress test

Paste the AI-driven revenue plan. Get the implied CAGR, evidence-based haircuts and a downside case where the AI levers deliver zero.

You are a skeptical IC member reviewing an AI-driven
revenue plan. Your job is to find what breaks.

THE PLAN: [paste the model summary or deck section]
BASE CASE: revenue today [ ], target [ ] in [ ] years
HISTORY: revenue growth last 5 yrs [ ], market growth [ ]
AI LEVERS CLAIMED: [list with revenue attributed to each]

Tasks:
1. Compute the implied CAGR and compare it to the
   company's history and its market. Show the math.
2. Decompose the target: how much is volume, price, new
   product, M&A, and "AI." Flag any plug.
3. For each AI lever, rate evidence: shipped and measured /
   piloted / slide only. Haircut: 0% / 50% / 100%.
4. Rebuild a haircut case and a downside case (AI levers
   deliver zero). Does the deal still clear our hurdle of
   [ ]x MOIC / [ ]% IRR at entry price [ ]?
5. List the 5 diligence asks that would most change your
   view, and 3 KPIs to put in the 100-day board pack.

Rules: use only my inputs. Label estimates. If something is
missing, write [VERIFY]. Do not soften the conclusion.
  • Use it before IC: run it on your own model first, so you hear the questions before the committee asks them.
  • Use the output: the haircut case is what you pay for. The rest is upside.

Run Play 2 on the next deck that says "AI-driven growth." If the deal only works with the AI levers, it's a venture bet.

Sources: