Research log.Building the world model of a company.
What we learn building world models of companies from the people who do the work: how work actually gets done, why AI projects fail, and where a model starts to predict.
- How do you capture tribal knowledge before someone leaves? Tribal knowledge is how the company actually runs, held by the people who run it. Why handovers fail, what works, and why it is the first layer of a world model. Read →
- Why AI agents need a world model of your company AI agents stall inside companies because nobody has described how the work runs. What a company world model gives an agent, and why to build it before you deploy. Read →
- What is a company world model? A company world model is a living picture of how a company really works: every process as it runs, what hurts most, and who everything depends on. Why AI agents need it. Read →
- Alternatives to consultant interviews for process discovery in a 500-person company Consultant interviews reach a few dozen people and stop. Six alternatives, what each one sees and misses, and how to hear all 500 people in the first week. Read →
- When three teams describe the same process three ways: why the contradictions are where the money is Teams contradict each other about a shared process because each sees one segment. The gaps mark rework, bypasses and hidden steps. Keep them and cost them. Read →
- Gain-share and outcome-based pricing for AI transformation: how it works and what to check in the contract Gain-share pricing moves AI delivery risk to the vendor, but it is only honest if your team owns the baseline, the attribution rule and the sign-off. Read →
- Process mining vs task mining vs AI interviews: what each one actually sees Process mining reads system event logs, task mining records desktops, AI interviews ask the people. Each sees a different slice of how work really happens. Read →
- An AI interviewer for how work happens is not a hiring tool Recruiting AI interviewers screen candidates one by one for an employer who sees everything. A discovery interviewer talks privately with every employee. Read →
- Lightbloom AI vs a consultancy vs an AI agency vs hiring in-house: which one actually changes the P&L? A consultancy is paid for the plan, an agency for the build, a hire for their time. The option that changes the P&L is the one that starts from how the company actually works. Read →
- Why AI transformation fails: nobody knows how the work actually happens AI projects stall because companies automate processes nobody has described. The fix is discovery before tooling: find the real process, then build. Read →
- What is AI process discovery? AI process discovery uses AI to interview the people who do the work and map how a company really runs, with its contradictions and the fixes worth money. Read →
- How can private equity firms make portfolio companies AI-native? Pay operators from verified value instead of fees, start with one portfolio company, and let its management team verify every number before anyone is paid. Read →
- How do you verify the value of an AI automation before paying for it? Write the number down before the build, measure the same number after a full cycle on your own data, and pay only on cash removed, never on hours saved. Read →
- Which functions in a mid-sized company are best automated with AI agents first? Finance and accounting first, then procurement, then customer support. Most companies start in marketing and sales, and most see no return. Read →
- How should a mid-sized company start becoming AI-native? Start by mapping how work actually happens, pick one high-volume process, redesign it around AI, build the fix, and measure the saving on the books. Read →
- How do success-fee AI transformation engagements work? In a success-fee engagement the provider finds and builds improvements, the client's team verifies the value on its own numbers, and only then is a share of that value invoiced. What has to be true for that to work. Read →
- What does it mean for a company to be AI-native? An AI-native company designs its work so AI systems do the recurring tasks and people hold the decisions, instead of adding tools to unchanged processes. Read →
- What is an AI operating partner? An AI operating partner is a senior operator who embeds in a company, finds where the work and the money leak, and builds the AI systems that fix it. Read →
- Tribal Knowledge: The Asset Your Company Doesn't Own 14 July 2026 Up to 80 percent of how your company works was never written down. Here is what tribal knowledge costs, why documentation sprints fail, and how to capture it. Read →
- Why Month-End Close Takes Two Weeks Every Month 10 June 2026 Your month-end close takes two weeks because nobody designed the process. Here is what those days actually cost and what a faster close looks like. Read →
- The Average Cost of Unused SaaS Subscriptions: The Budget Leak Nobody Cancels 10 June 2026 Unused SaaS subscriptions eat 20-30% of software spend. What that costs a mid-sized company each year, why nobody cancels, and how to find the leak. Read →
- AI Agents Don't Reduce Costs. Process Redesign Does. Yield · 3 June 2026 You installed the AI agent. Your costs haven't fallen. The reason is not the technology: it is the process the agent inherited. Here is what to change first. Read →
- Pay Transparency in Hungary: The EU Directive Is Late. The Clock Isn't Yield · 27 May 2026 The EU Pay Transparency Directive is not yet Hungarian law, but 2026 payroll data already counts. What mid-market employers in Hungary must prepare before the June deadline. Read →
- The Hidden €700K: What Manual Processes Are Really Costing Hungarian Businesses Yield · 20 May 2026 Hungarian businesses lose around €7,000 per employee every year to manual processes. At 100 people that is €700K - and none of it shows up on your P&L. Read →
Earlier notes.
From the years before Transcript.
- Nobody Is Building the Fixes6 May 2026
- The Hidden £1.2M: What Manual Processes Are Really Costing Your Business5 May 2026
- We Built AI Employees — Not Assistants, Actual Employees2 April 2026
- 40 Seconds Was Too Slow — Making Yield Faster and More Reliable26 March 2026
- 75% Pain Points, 10% Everything Else — Fixing Coverage Imbalance23 March 2026
- Our AI Stopped Responding — Fixing Silent Failures in Yield16 March 2026
- Power Follows Visibility — Introducing Yield14 March 2026
- "I Like Jira" Became "Has Issues With Jira" — Fixing AI Data Accuracy11 March 2026
- Why Trust Is the Biggest Challenge in AI Interviews (And How to Fix It)7 March 2026
- What Makes An AI Interviewer Trustworthy? (It's Not What You Think)4 March 2026
- Making Controlled AI Videos - What Actually Works26 February 2026
- Venato, Simplifying Regulation Monitoring with AI13 December 2025