For Leaders In Private Markets
AI Window is the AI advisor to private equity, venture capital, family offices, and the portfolio companies they send us into.
Independent advisory, based in Austin and working with firms across the United States.
Request An IntroductionFifteen years at NASA JPL as a Systems Engineer on the Curiosity and Perseverance Mars rovers, then a year at Blue Origin working with Jeff Bezos on a lunar lander, then built and exited a company of his own in three years.
"AI training went in and a delivery metric moved. That is adoption you can see in the numbers."
AI Window runs on the same systems it installs. The daily briefs, the pipeline checks, and the follow-ups are all built the way a client build is.
Systems That Sit On Top Of The Deal, Not Beside It.
AI inside a private equity workflow usually stops at a demo. AI Window builds on the data a firm already holds: data rooms, CIMs, models, board packs, and portfolio reporting. Baseline first, so value creation is measured, not assumed.
CIM Analyzer
Reads the business and the transaction together, produces the firm's standard report, and flags the pages a person should still read.
Portfolio Reporting Tracker
Knows what each company owes and when, chases the late ones, and rolls what arrives into the LP update draft.
Executive Enablement
1:1 coaching for partners, operating partners, and portfolio CEOs on where AI changes the plan, and where it does not.
Where AI Adoption Stalls In A Portfolio.
AI adoption in a portfolio company stalls in four predictable places. Each one is operational, and each one is fixable.
No Baseline
The tools are licensed and nobody can say what changed, because nobody wrote down the before number.
Pilots That Never Leave The Pilot Group
A few people use AI well. The other ninety percent of the org was never trained.
A Policy Nobody Can Find
People hold back because no one has told them plainly what is allowed with which data.
A Gain With No Owner
The process improves for a quarter, then slides, because the hours to maintain it were never budgeted.
What Changed, And Who Measured It.
Measuring AI adoption takes a before number, and almost nobody writes one down. Putting one on record is the valuable part. Every figure below came out of the client's own systems, and each one says who produced it.
Cycle Time, Weeks Down To Days
An engineering team of under 20 went from a 15 day baseline to under 3 by the second sprint. Weekly throughput moved from about 3 items to more than 20. Their PM and director ran the change. The org was trained on Claude alongside it.
Writing Took Days, Review Took Weeks
Median code writing was about 6 days. Code review took more than 20. Both came out of the client's own Jira, and the VP who owns delivery data had never seen the split. It is measured monthly now.
Clear Rules On What AI Can Touch
The AI use policy was out of date and nobody could say which tools were allowed with which data. The rewritten decision path was accepted with the CISO in the room and now sits with policy governance.
The Cost Of Keeping The Gain
Manager time to maintain the new process, plus about the same from the person running it. A gain nobody budgets to maintain does not last, so the upkeep is counted with it.
Three Services, Bought On Their Own.
A menu, not a bundle. A one-time diagnosis, monthly executive coaching, or an ongoing Chief AI Officer mandate. Each is bought on its own.
AI Opportunity Diagnosis
Every workflow mapped by role, tool, and time, then scored for replaceability, impact, and speed to build.
- —Every workflow step scored for ROI
- —Ranked roadmap, not a wish list
- —For operators who cannot say where AI belongs
AI Executive Coaching
Private working sessions with one executive, weekly or every other week, building agents, briefs, and workflows into the real week.
- —Built into your own inbox, calendar, and week
- —Weekly or bi-weekly cadence
- —For founders, partners, and family offices
Chief AI Officer
Ongoing ownership of AI adoption across a whole organization, run as five pillars: methodology, metrics, tools, rollout, and security.
- —Baselines taken before anything changes
- —Standing weekly session with tech leadership
- —For CTOs and CEOs at mid-market companies
Frequently Asked Questions
What does a fractional Chief AI Officer do?
Owns AI adoption across a whole organization, run as five pillars: methodology, metrics, tools, rollout, and security. Baselines get taken before anything changes, and there is a standing weekly session with tech leadership. It is a monthly engagement. It exists because a mid-market company usually needs the ownership without the full-time hire.
How do you measure AI adoption in a portfolio company?
Start with a before number out of the company's own systems, whether that is Jira, the CRM, or the finance stack. Measure the same number every month after that, and count the hours it takes to keep the gain running. A gain nobody budgets to maintain does not last. At one client, engineering cycle time went from a 15 day baseline to under 3 days by the second sprint, and weekly throughput went from about 3 items to more than 20.
Can AI be used on deal data safely?
Yes, once somebody writes down what is allowed with which data. The problem is rarely the tools. It is that the AI use policy is out of date and nobody can find it. One client's policy was rewritten with their CISO in the room, as a decision path an engineer can run in under a minute.
What is a CIM analyzer?
A system that reads the business and the transaction together, produces the firm's standard report, and flags the pages a person should still read. It does not replace the diligence read. It finishes the first pass and shows where every figure came from.
Who does AI Window work with?
Private equity, venture capital, and family office firms, and the portfolio companies they send us into. 14+ investment firms so far: 9 PE, 3 VC, and 2 family offices, plus 3 PE-backed portfolio companies.
Who is Gregory Villar?
Founder of AI Window. Fifteen years at NASA JPL as a systems engineer on the Curiosity and Perseverance Mars rovers, then a year at Blue Origin working on a lunar lander, then built and exited a company of his own in three years. He has delivered 148 one to one sessions and coached 30+ executives.
Why does AI adoption stall after the pilot?
A few people use AI well and the rest of the organization was never trained. The tools get licensed, the pilot group gets good, and nothing crosses into the wider team. The fix is training the organization rather than the pilot group, and putting a number on it that leadership already approved.
How does an engagement start?
With a 30-minute call scoped around one live problem. Introductions are by request, and you can expect a reply within two business days.
Bring The Deal Data. We Will Build The System.
Introductions are by request. Expect a reply within two business days and a 30-minute call scoped around your challenge.
Why I Built AI Window
I spent 15 years building spacecraft, including two Mars landings. Now I build the AI systems that private equity and venture capital firms and their portfolio companies actually use, and I measure whether they work.
I Failed My Way Into NASA.
I grew up in Long Beach, California, went to high school in the Philippines, and started college in the US at age 16. I failed my first upper-level math class. I scored in the bottom 10% on the Physics GRE and was rejected by every graduate school I applied to.
I kept going anyway. A NASA scholarship, a master's at USC, and a summer internship at JPL in 2008 turned into a 15-year career.
Fifteen Years, Two Mars Landings.
I built spacecraft at NASA's Jet Propulsion Laboratory (JPL) and Blue Origin. At JPL I worked on the Curiosity and Perseverance Mars rovers: verification and validation lead, operations lead, and test conductor for a parachute test in the world's largest wind tunnel.
Both landings were surreal. Each one was five to ten years of work by thousands of people, all for a single moment you get exactly one shot at. That taught me how big, high-stakes projects actually succeed: you define the outcome, you test everything, and you measure whether it worked.
Then The World Went Remote.
During the pandemic I realized the work could happen from anywhere, and I had accomplished what I set out to do at NASA. So I left to try something new.
I built and sold a company, Profit Pandas. The lesson I took from selling it is the one AI Window is built on: build a business that runs without depending on one person, from day one.
Why AI, And Why Now.
In 2025 I was spending my time helping business owners, and it was clear AI was about to hit mass adoption. I wanted to be ready, and positioned, before that happened rather than after.
My first AI Window client was Jess Mah. She finished high school at 15, became the youngest woman ever accepted into Y Combinator, and grew inDinero from zero to more than 250 employees. She introduced me to managing partners across private equity and venture capital, and that is who AI Window serves today.
What I Do Now.
I'm an engineer first and a strategic advisor second. I build AI systems the same way I built spacecraft systems: simple, reliable, and measured against a baseline. Almost no one is doing this for funds and their portfolio companies.
That means deal flow and CIM analysis tools for investment teams, and AI rollouts inside portfolio companies. At one, an engineering team cut a 15-day delivery cycle to under 3 days by its second sprint.
Request An Introduction