Why PE-Backed Software Now Has to Beat the Build-It-Yourself Option

Published by PrivateEquities.AI

🧭 a16z Just Set a New Test for Every Software Company You Own

On November 2, Andreessen Horowitz growth partner Tugce Erten published a short, uncomfortable essay: "How AI Applications Earn Their Premium."

Her argument is simple. Better models and easier build tools keep improving the customer's other option.

If a customer can reproduce your core output with a general-purpose model and a little effort, your premium is exposed.

The new test: buying the product has to beat the customer's best credible alternative on total cost, quality, speed and risk.

She cites Repsol's legal team, which ran a blind evaluation of Harvey, Microsoft Copilot and its own internal prototype across an eight-week pilot with 50 lawyers in 12 jurisdictions. The vendor had to win against the in-house build, not just against other vendors.

Here's the catch. Private equity owns more software than almost anyone, and much of it was bought before customers could build.

Carta counts a nearly 500% increase in software companies held in PE portfolios since 2006, and U.S. buyout software holdings were already marked down 8.9% from Q4 2025 to Q1 2026.

A note from the team behind this newsletter: we are building OpenMap, The Agent Network, the connective tissue for the world's AI agents, starting with every person and every local business. We are raising a £5M seed. Meet us at openmap.ai/investors.

💡 From "Nobody Would Build This" to "Why Wouldn't They?"

For a decade, the software thesis was easy to defend.

Switching costs were high. Building in-house was slow and expensive. Seats grew as customers grew.

That moat is getting shallower every quarter.

Erten puts it plainly: moving even one high-volume task in-house can be enough to challenge the premium.

PwC's dealmakers saw the same shift earlier this year. Software, they wrote, is moving from selling access to delivering outcomes, and seat-based pricing is under pressure.

The painful truth: your portfolio company's real competitor may be its own customer's AI team.

🔍 Why Most Funds Can't Answer the Test

Ask an operating partner which portfolio companies would pass a "best credible alternative" review. Most will give you a feeling, not a number.

The evidence exists. It is just scattered.

Usage sits in product analytics. Renewals and discounts sit in the CRM. Support load sits in a ticketing tool. Cost to serve sits in the ERP.

PwC flags the trap: healthy net revenue retention can hide AI-driven seat contraction underneath. You only see it when you cut retention by cohort and module.

And that cut requires data from four systems that were never designed to talk to each other.

This isn't a product problem. It's a context problem.

You can't prove a product beats the build option if you can't see where customers get value and where they quietly leave.

⚙️ How It Works: Pressure-Testing Every Software Asset

PrivateEquities.AI gives your fund one connected, always current view of the portfolio, so you can see which software companies are earning their premium and which are living on it.

1. Connect, Don't Collect
   The platform plugs into the systems your portfolio companies already run: CRMs, ERPs, data rooms, warehouses and BI tools. No data migration. No integration project per company.

2. One Context Graph
   Usage, renewals, seats, discounts and cost to serve are linked across the portfolio, so retention and churn mean the same thing in every company.

3. Agents on Top
   ProfileAI, CleanseAI and other in-house agents validate data and harmonize formats before a figure reaches a diligence memo, a board pack or an exit deck.

4. Always Live
   When a source system changes, the graph updates. You see seat contraction in the month it starts, not when a buyer finds it.

📊 Software Value: Before and After Context

The software is the same. Whether you can defend its premium is what changes.

💸 Price the Unit of Work, Not the Token

Erten's third rule is the one operating partners should tape to the wall: prove the value, then price to capture it.

That means knowing the full cost of one completed unit of work, a contract reviewed or a ticket resolved, including retrieval, model calls, retries, human review and failed attempts.

It matters because AI bills are not shrinking. In a16z's August survey of 100 enterprise LLM decision-makers, 94% expected total LLM spending to rise and 37% expected it to at least double.

A portfolio company that runs every step on a premium model will struggle to price below the build option and still keep its margin.

The PrivateEquities.AI Inference Cost Calculator shows up to a 35x price gap between frontier and open-weight model classes for the same workload.

The capital is committed. The overpay is optional.

🧩 A Day in the Life of an Operating Partner Using PrivateEquities.AI

8:00 AM - The Headline
You read the a16z essay over coffee. Your first question: which of our software companies could a customer rebuild this year?

10:30 AM - Reading the Signals
You open the context graph. Two companies show rising usage per account. One shows steady net retention but shrinking seats in its largest cohort.

2:00 PM - Finding the Cause
The agents line up CRM, billing and support data. The shrinking cohort leans on one reporting module, the easiest piece for a customer to rebuild with a general model.

5:00 PM - Acting on It
The CEO gets a clear brief: deepen the workflow around that module, reprice it on completed work, and move routine AI tasks to a cheaper model class.

That's how conviction compounds.

Every month the alternative gets better. Your evidence has to get better faster.

🚀 Ready to See Which Software Assets Earn Their Premium?

If your software thesis still rests on switching costs and a single retention number, the build option is closing in.

PrivateEquities.AI helps funds turn portfolio chaos into operating clarity.

👉 Book a Demo
See how connected context shows where each software company wins and where it is exposed.

👉 Run the Inference Cost Calculator
Find out what your portfolio is overpaying for AI.

🔔 Key Takeaways

1. The Bar Moved. a16z says AI apps keep their premium only if they beat the customer's best credible alternative, including building in-house.

2. PE Is Exposed. PE portfolios hold far more software than in 2006, and buyout software marks already fell 8.9% in early 2026.

3. Headline Retention Can Mislead. Healthy net retention can hide AI-driven seat contraction.

4. Context Comes First. You can't prove a product beats the build option without connected usage, renewal and cost data.

5. Cut the Overpay. Knowing the true cost per unit of work, and matching tasks to the right model class, protects margin.

🧠 Final Thought

The funds that keep their software premiums won't be the ones with the stickiest contracts. They'll be the ones that can show, company by company, why buying still beats building.

PrivateEquities.AI is making that possible.

Warm regards,

PrivateEquities.AI Team