The Operating System Show

Mishka Kofman has been writing software for 28 years, from Madden at EA to AWS. Today he runs Paracorp and builds ARC, a platform that modernizes legacy software by pulling out every business rule and rebuilding the system around them. His warning for owners: AI made code cheap to write, and it did nothing to make code cheap to keep. Teams are shipping faster than ever and creating debt faster than ever, and most leaders cannot see what it is costing them. His rule for build versus buy is simple. Buy what is not central to you. Build what holds your business rules and gives you an edge. ## In this episode - Why he left steady employment to build Paracorp - What AWS and EA each taught him about how teams should run - Why every decision should be written down, and what that does for AI - Whether a product owner should read the team's retro notes - Technical debt in the AI era, and why cheap code is not cheap to keep - The house that keeps adding rooms while the foundation moves - Rebuilding software from its business rules - What he finds in every legacy codebase he opens - Build versus buy, and keeping what holds your business rules - How he wins over engineers who do not want AI touching their code - Who should not modernize right now ## Timestamps 00:00 Intro 01:08 Why he started Paracorp 03:19 What AWS and EA taught him 05:30 Teams that run without top down orders 07:12 Why every decision gets written down 08:57 Should the product owner read the retro notes? 12:36 Technical debt in the AI era 14:20 Cheap to write, not cheap to keep 16:48 The house that keeps adding rooms 20:04 What ARC does 23:27 Business rules are the software 24:37 How you avoid breaking things 27:40 What he finds in every legacy codebase 29:40 Build versus buy now 32:17 Merging three apps into one 33:43 Adding features during a modernization 35:37 Dev teams that do not want AI touching their code 38:57 Who should not modernize right now 40:44 What is next for Paracorp and ARC 44:05 Where to find Mishka ## About Mishka Kofman Mishka Kofman is the founder of Paracorp and the builder of ARC, the Architecture Refactoring Coordinator. He has spent 28 years in software, including Madden at EA and AWS, and runs workshops on using AI to write predictable, maintainable code. Paracorp: https://paracorp.io/ ARC: https://try-arc.com/ Listen: https://feeds.transistor.fm/the-operating-system-show Be a guest: https://crm.frontrangesystems.com/discovery

Jordan Fleming bootstrapped smrtPhone from Poland, with the technical team in Romania, to just under 10 million dollars and nearly 60 people. Along the way it went from a phone system built for Podio to what he now calls an agentic conversation platform. He has since moved from CEO to chairman. His argument is that most companies spending on AI are not seeing growth from it. They drop very capable agents into processes designed around human limits, and those processes are full of invisible work that people quietly cover. Only about 5 percent of the companies he has worked with have mapped their processes the way the work actually runs. That is the subject of his book, The AI Workforce: redesign the business, one process at a time, around work shared between agents and people. ## In this episode - Four owners, distinct roles, and disagree and commit when they can't agree - Why they built in Eastern Europe, inside the EU legal and IP framework - The happy accident of building a phone system for Podio, and growing with almost no marketing - How every product update now flows into a company brain that updates help docs and agents - An agent that made 220,000 calls in three and a half weeks - Why AI spend is not turning into growth - The invisible work that never makes it onto the process map - Business structures built for one kind of labor, and what changes now - Hiring a role based AI agent and onboarding it to your company - Why voice agents built software first sound great in demos and fail in execution ## Timestamps 00:00 Welcome 01:01 Four owners, and stepping up to chairman 02:15 Splitting the roles so nobody deadlocks 03:43 Why build the company in Eastern Europe 05:18 From a PhD in music to Poland 06:08 Bootstrapping a phone system for Podio 08:09 From VoIP company to agentic conversation platform 08:45 How AI changed the way they build 09:40 220,000 calls in three and a half weeks 10:15 Why AI spend is not turning into growth 12:08 Only 5 percent map their real process 13:52 Redesigning process after process 14:42 The AI Workforce, his book 16:08 The four zones of AI 16:39 Business structures built for one kind of labor 18:47 The next 24 months 20:06 What broke as they scaled 22:05 Hiring role based AI agents 23:27 Onboarding an agent: property maintenance 25:01 Latency, and why telecom first matters 26:43 Who smrtPhone is for 28:14 Where to find Jordan and the book ## About Jordan Fleming Jordan Fleming is the cofounder and chairman of smrtPhone, an agentic conversation platform for voice and text, built for companies of 10 to 500 users. He is the author of The AI Workforce, a framework for redesigning a business around shared execution between AI agents and people. The AI Workforce, free executive briefing for listeners: https://thefutureworkforce.ai/p/os-show smrtPhone: https://smrtphone.io Listen: https://feeds.transistor.fm/the-operating-system-show Be a guest: https://crm.frontrangesystems.com/discovery

Bryan Clayton started mowing yards in high school and grew it into a Tennessee landscaping company with about 150 people and 90 trucks. Fifteen years in, it was organized chaos every day, and he was the scaffolding holding it up. He never took a vacation longer than two days. After he sold it, he set out to build the opposite: a business with a system for almost every problem. That became GreenPal, a marketplace for lawn mowing that has handled about 300,000 transactions with fewer than 50 people, no offices, and no lawnmowers. We get into what he refused to repeat the second time, the difference between hard work and difficult work, how GreenPal keeps quality high with vendors it does not employ, and where he draws the line on what AI should never touch. ## In this episode - Why his first company was built as scaffolding around the founder - Every problem you solve unlocks five more, and why that is the job - Running a marketplace as a never ending series of experiments - From fragile to antifragile: weekly loops that make the business better on their own - Hard work versus difficult work, and the partner who wore 15 hour Sundays as a badge - What breaks in a remote first company, and why even Zoom has an office - The carrot and stick system that keeps independent lawn pros showing up - How AI gave a state school founder an MIT data scientist at the table - Why the founder should still do an hour of customer support every day - "Stop being lazy and ask Claude," and why AI makes everyone busier ## Timestamps 00:00 Intro 00:49 Under 50 people, 300,000 transactions 02:35 What he did differently the second time 05:45 Landscaping versus software 08:21 A never ending series of experiments 10:01 150 employees, 90 trucks, and a mechanic shop 11:44 From fragile to antifragile 14:19 Hard work versus difficult work 16:57 Running a remote first company 18:46 Quality from vendors you do not employ 21:53 How GreenPal uses AI 24:42 Who should not build a marketplace 26:25 The four zones of AI 27:59 Why the founder should do customer support 29:28 Messy data and "stop being lazy and ask Claude" 32:16 Why AI makes everyone busier 33:26 What is next: a million users 34:11 Where to find Bryan ## About Bryan Clayton Bryan Clayton is the cofounder and CEO of GreenPal, a marketplace that connects homeowners with local lawn care pros. Before GreenPal he built and sold a landscaping company in Tennessee. GreenPal: https://www.yourgreenpal.com/ Listen: https://feeds.transistor.fm/the-operating-system-show Be a guest: https://crm.frontrangesystems.com/discovery

Jay Bourgana took a real estate brokerage from zero to 350 million dollars in sales volume in four years, then sold 70 percent of it to a much larger firm in 2018. Since then he has been on both sides of about 30 acquisitions, buying plumbing, HVAC, and IT companies. Today he runs Foresight Partners. His favorite deals are what he calls flywheel businesses: no website, no marketing, customers who keep calling anyway, strong margins, and a founder who carries the whole operation in his head. Before he even signs a letter of intent, he starts pulling that knowledge out. His warning for owners racing toward AI: fix the data first. Most companies run on duplicate records, spreadsheets stitched together by hand, and reports that arrive a month late. Put AI on top of that and it just moves faster in the wrong direction. ## In this episode - How a manufacturing mindset built a brokerage that answered new leads in seconds - The cold email he almost deleted, and selling 70 percent of his company 30 days later - Why he buys businesses that only work because the founder is inside them - The three engines he maps before closing: clients, talent, and capital - Why owners grow to five million, hate it, and shrink back to three - People, then process, then technology, and why speeding up the machine shakes everything loose - Data before AI: closed loop companies and the cost of lagging numbers - Why middle managers become the human in the loop for AI agents - Who should sell to him, and who should not - Kids Business School, teaching entrepreneurship to kids 7 to 17 ## Timestamps 00:00 Zero to 350 million in four years 00:52 Bringing manufacturing thinking to real estate 03:22 How owners can use data better 06:41 Speed to lead, then and now 09:11 The cold email that sold the company 10:43 Retiring, then getting back in 12:02 What he does first when he buys a business 14:02 When everything lives in the owner's head 16:11 Buying to own forever 17:12 Why owners shrink back to three million 20:00 People, process, technology, and the shake 21:01 How he uses AI and why data comes first 23:07 The four zones of AI 24:53 Why middle managers become the human in the loop 28:13 75 states in a 50 state CRM 31:46 If you need Excel, something is wrong 32:35 The company that could not tell him its churn 34:06 Who should sell to him, and who should not 37:50 What is next, and Kids Business School 39:39 Where to find Jay ## About Jay Bourgana Jay Bourgana is the founder of Foresight Partners, which buys and operates founder led home services and professional services businesses. He built and sold a real estate brokerage, has been on both sides of about 30 acquisitions, and is building Kids Business School, an entrepreneurship program for kids 7 to 17. Acquisitions Collective: https://acquisitionscollective.biz/ Kids Business School: https://kidsbusinessschool.com/ Listen: https://feeds.transistor.fm/the-operating-system-show Be a guest: https://crm.frontrangesystems.com/discovery

Craig Wahl coaches CEOs and founders whose companies got bigger, faster than they expected. He runs Njorth, works as a fractional COO for tech companies, and is finishing his first book, The Walk. His core reframe for CEOs: your job isn't to make all the decisions. It's to make sure they get made, by the right people, and that someone is accountable for them. He learned most of it scaling a team fast. He chased A players and found a team of solid B and C players was just as strong. He watched long documented processes go unused and swapped them for simple checklists the team actually ran, then tracked everything and kept iterating. ## In this episode - Why CEOs need a safe place to talk, and who they can't talk to inside the company - Your job isn't to make the decisions, it's to make sure they get made - What founders refuse to hand off, and what the CEO should be doing instead - Why hiring only A players is a fool's errand when you're scaling - The Sherpa on Everest: stop yelling down the mountain - A process nobody follows isn't a process - Trust is boring, and how leaders shortcut it - The Walk, his practical guide to the long middle of leading people - Why "use AI more" is one of the most dangerous things an executive can say - The question to ask anyone selling you AI: have you ever run a business? ## Timestamps 00:00 The show open 00:30 Meet Craig Wahl 01:14 The three things Njorth does 02:20 You don't call an electrician for a leak 02:55 Why CEOs need a safe place to talk 06:30 Make sure the decisions get made 07:33 What founders refuse to hand off 08:55 The leadership triangle 10:25 Can anyone sell as well as the CEO? 12:45 What breaks when the team gets big 13:23 Why hiring only A players is a fool's errand 14:49 The Sherpa on Everest 17:25 A process nobody follows isn't a process 18:36 Simple checklists, track everything 20:45 Trust is boring 23:10 How leaders shortcut trust 25:06 The Walk 27:14 A master's in leadership and a junior manager 29:38 What do you want people to remember about you? 30:20 How he uses AI in his own business 32:25 Why "use AI more" is dangerous 34:24 Have you ever run a business? 35:10 The four zones of AI 35:46 What error rate will you accept from AI? 38:43 Ten salespeople, selling 80% of the time 39:36 Who should hire Craig 39:59 What's next for Njorth 40:39 Where to find Craig ## About Craig Wahl Craig Wahl is a leadership architect and CEO advisor at Njorth. He coaches founders and CEOs, runs leadership alignment workshops, and works as a fractional COO, mostly for tech companies scaling from Series A to Series C. His first book, The Walk, comes out in spring 2027. Website: https://njorth.ai Listen: https://feeds.transistor.fm/the-operating-system-show Be a guest: https://crm.frontrangesystems.com/discovery