Newsletter

Work with me

Guides

AI Operations for Mid-Market and PE-Backed Companies

What an agentic operating system answers inside a company, how to score AI readiness in a deal, and how to keep AI cost below the value it creates. We run one inside DevriX, so the examples come from our own tasks, hours and invoices.

Mario Peshev

First released

2 deep dives · 15 short answers · 3 chapters · newest October 25, 2026

Every article in this guide

How it works

AI Ops Hive: An Agentic Operating System for Mid-Market and PE

The layer we run inside DevriX: what it reads, what it answers every night, and what it cost to build.

AI in a deal

What AI Readiness Tells You in a Deal, and How to Score It

How to judge a target’s AI claims against its data and its costs.

The economics

AI value creation ranks inference cost against revenue per user

Why inference cost has to be ranked against what a user is worth before anything ships.

Agentic operating systems

An agentic operating system is a set of AI agents, each owning one part of the operation, connected to the company’s own systems with mostly read-only access, a shared memory of what has been found, and a person approving anything that changes the outside world. It works on a schedule and reports what changed, which is the difference from a chatbot that waits for a question.

Inside DevriX it reads tasks, hours, code, invoices and servers every night. On one multi-site program over six months it showed that 48% of 1,390 logged hours went to planned build, 16% to coordination and 14% to rework after client feedback, a split nobody on the team would have guessed and one that changed how fixed scopes are written.

Terms used in this guide

Agentic operating system: AI agents connected to a company’s own systems that answer operating questions on a schedule, with people approving actions.

Inference cost: what each model request costs to run; it grows with usage, like a variable cost of goods.

Token: the unit models are billed in, roughly three quarters of a word.

AI readiness review: a score of the data, systems, team and costs behind a company’s AI plans, and the uses worth funding first.

Question bank: the fixed set of operating questions an agentic system answers for every company, every night.

Deep dives in this chapter

7 short answers on this

AI cost and value

AI cost behaves like a variable cost of goods: every request has a price, and the price grows with usage. Ranking inference cost against the revenue a user is worth, before the roadmap goes live, decides where the expensive capability belongs and where a cheaper or static answer will do.

Deep dives in this chapter

AI adoption in a portfolio

The portfolio question is where to start. One company, one readiness review and a fixed bank of operating questions answered from live data proves the value; the same question bank can then run across the rest of the portfolio with each company keeping its own data.

The rules that keep it safe

  • Agents read, draft and check; people review and decide.
  • Nothing writes to a live system without approval.
  • Every answer carries its source and how much of the source it read.
  • Token cost is forecast before deployment and measured after.

Deep dives in this chapter

8 short answers on this

Common questions

What is an agentic operating system?

A set of AI agents connected to a company’s own systems that answer a fixed bank of operating questions on a schedule, with people approving anything that changes the outside world.

How do you score AI readiness in a deal?

By the data the target can use, the systems and team that would run it, the cost of inference against the revenue it supports, and the uses worth funding first.

Do AI agents make decisions?

No. Agents read, draft and check; people review and decide, and nothing writes to a live system without approval.

Where should AI start in a portfolio?

With one company and a readiness review, then the same question bank across the rest once it proves its value.

Every article in this guide

Agentic operating systems

7 short answers on this

AI cost and value

AI adoption in a portfolio

8 short answers on this

Start a conversation

Bring the asset, the deal stage and the decision in front of you

A short note is the fastest way to a useful first call. Sponsors, operating partners and leadership teams of $30M to $500M companies.

Work with me
Contact