Founder dependency score: what would stop without you
Answer 15 questions about pricing, approvals, customers, suppliers, follow-up and history. For each one, say what would happen if you were away for 30 days. You'll see how much of the work still runs through you, which kind of dependency is strongest, and the fix that fits it.
The founder dependency score asks what would happen to 15 pieces of everyday work if the founder were away for 30 days. It groups the answers into five kinds of dependency (knowledge, decision, relationships, process and context), scores each out of 100, ranks them, and gives the fix that fits each kind. Work you choose to keep with you is left out of the score.
The question behind the score
If you were away for 30 days, which decisions would sit unanswered, which approvals would stop, and which supplier relationships would turn out to exist only in your phone? The answers show where your company still runs on you. They matter for growth, because every one of them is a queue with you at the front, and they matter more if you might sell, because a buyer asks whether the revenue continues after the owner leaves.
If pricing, customer exceptions and approvals all run through you, a buyer is acquiring a dependency along with the business.
Sonia Sarrami, founder of Noesis AI
The five kinds of founder dependency
Every time someone comes to the founder, the reason falls into one of five kinds. Knowledge: you know something nobody else does, such as a customer's special pricing. Decision: others have the information, but only you are trusted to decide. Relationships: key customers, suppliers and partners deal with you. Process: work moves because you push it. Context: the team could make the call, but lacks the history to feel confident.
Each kind needs a different fix, so the tool keeps them apart. If someone had all the information and still didn't know whether they were allowed to approve, the gap is authority, and better documentation won't close it. If they didn't know a customer had a special rule, the gap is knowledge, and the fix is to capture it.
What the five kinds look like in published case studies
Quoting is where knowledge dependency shows up most often. At Sweetwater Machine Shop, a job shop in Tennessee, every quote depended on the owner's knowledge and nothing recorded past quotes, so each one was rebuilt from scratch. Once requests came in through a form and estimates were built from stored data, the owner still did the quoting, in about 5 minutes per quote where it had taken over 30 1. The same pattern appears with senior staff. At Craftmasters Laser & Fabrication, a family-owned Texas fabricator of nearly 40 people, the VP of Operations owned every quote, so quoting could be neither handed off nor scaled 2.
Relationship dependency surfaces when someone leaves. Granite Properties had two senior leasing staff retire with 20 years of tenant relationships between them. The company moved to a system where engineers, brokers and leasing staff log each tenant interaction as it happens. Within six months, more than 30 people were updating records on over 80 tenants, and the time from lead to signed lease fell 14% 3.
Process dependency is the founder as the engine. Rolf Selvig, president of Selvig Development, a small California builder, was the go-between for field crews, subcontractors, billing and supplies. Shared schedules, mobile time clocks and change orders signed in one system meant he stopped chasing timecards and relaying schedules 4. In Alberta, the founder of Starburst Business Solutions, an accounting practice, was close to turning away new clients while recurring work ran on manual calendars and memos. With recurring tasks and reminders in place, the practice grew from about 30 to 50 clients, added two team members, and the founder took a holiday for the first time in years 5.
How the score is worked out
Each answer scores 2 when the work would wait for you, 1 when the team would go ahead only after checking with you, and 0 when it runs without you. The score is the points as a share of the most you could score on the questions you answered, out of 100. Each kind gets its own score the same way, and the kinds are ranked from strongest to weakest dependency.
Questions you skip and work you mark as should stay with me are left out, so the score measures only the dependency you would want to remove. There is no pass mark. The score is for comparing your business with itself: answer again after you change something, and see whether it falls.
An example result
Say the owner of a 40-person distributor answers all 15 questions. Six would wait for them: special pricing, where the money stands, non-standard quotes, a large customer with a problem, noticing what hasn't happened, and a situation the team hasn't seen. Four would go ahead after a check, four run without them, and they mark approving large purchases as work that stays with them. That is 16 points out of a possible 28 on the 14 questions scored, so the score is 57.
Knowledge comes out strongest at 83, because two of its three questions wait for the owner. Relationships and process come out at 50. The owner's first move is the special pricing, since every non-standard quote stops at their desk and a mistake there costs margin, and the fix for knowledge is to capture it from real cases. The six items that would wait for them are the first rows of the interruption log described below.
What should stay with you
Some work should stay with the owner: capital allocation, critical relationships, culture, high-risk exceptions and strategy. The exact list depends on the company. Marking these as should stay with me puts them in the keep human column, and the work around them, such as gathering the history and preparing the numbers, can still move to the team or a system.
The goal is to stop requiring founder judgment where it adds nothing.
Sonia Sarrami, founder of Noesis AI
From the score to a 30-day test
The score is a starting map. The method we use to reduce founder dependency runs over four weeks. In the first, change nothing: log every time someone comes to the founder, what they needed and what would have happened without an answer, and sort each one into the five kinds. In the second, take the dependency that carries the most risk and review the last ten real cases with the founder: why this quote was approved, why that one was rejected, what margin felt uncomfortable.
What comes out of those cases sorts into four bins: rules that can be written down, context someone needs at the moment of deciding, exceptions where the usual logic doesn't apply, and judgment that stays with the founder. Week three builds the smallest system that lets someone else carry that work. For non-standard quotes, the system assembles customer history, margin, payment behaviour, capacity and similar past deals, a manager approves inside set limits, and anything outside them reaches the founder as a complete brief.
Sometimes the most valuable thing AI does is make the company capable of preparing the decision.
Sonia Sarrami, founder of Noesis AI
Week four is the absence test. The founder steps out of that workflow for half a day, then a full day, then several days, and every time they get pulled back in, the team logs why. Each pull-back points to missing information, unclear authority, an undocumented exception, or a call that belongs with the founder. Fix the gaps and run the test again. The number to track is how many interventions remain that the founder's judgment did not need.
In an engagement, the first two weeks are the Workflow Diagnostic: the interruption log, the real cases and the four bins, with a baseline of how often the founder is pulled in. A Workflow Deployment builds the system in week three and runs the absence test with you.
What the score can't tell you
The score is a self-assessment built on the five kinds of dependency. It is a starting point for a conversation and has not been validated as a risk measure or a business valuation. Your team may see the same work differently, so ask a manager to answer the questions too and compare where your answers differ.
Your answers stay with you
The questionnaire runs in your browser. Nothing you choose is stored or sent anywhere.
Free tool Manual work audit: what your manual tasks cost Want the hours behind the work that waits for you? List the recurring tasks and see what each costs a year, with a starting label for each. Get my audit →Bring us the workflow behind your results
On a free workflow call we look at the workflow you choose, where the time goes, and whether a Workflow Diagnostic is worth doing. You leave with a clear next step, even when that step is to leave the workflow alone.
Workflow Diagnostic → Workflow Deployment → Managed AI Operations · How engagements work
Sources
Case studies cited above, from Noesis's review of 3,362 published automation and AI case studies. Results are as reported by the publishing organization. Methodology
- Sweetwater Machine Shop: RFQ Quoting · Manufacturing, United States · published by Paperless Parts · CS-2632
- Craftmasters Laser & Fabrication: Paperless Parts Quoting · Manufacturing, United States · published by Paperless Parts · CS-771
- Granite Properties: Tenant Relationship and Leasing (VTS) · Real estate, United States · published by VTS · CS-708
- Selvig Development: Construction Management Platform · Construction, United States · published by Buildertrend · CS-2169
- Starburst Business Solutions: Workload Management · Accounting & bookkeeping, Alberta, Canada · published by Teamwork · CS-945