Cos At Seed

Chief of Staff at Seed

Twenty people, no systems, and a founder who is personally the routing table for every decision in the company. Nothing is written down, so nothing survives a person being on holiday, and every process is reinvented each time it happens. The founder answers twenty small questions a day, which is precisely why they are not answering the three that would change the year.

The cost is not disorganisation, which nobody at this size minds. It is that the company stalls whenever the founder's attention moves, and at seed it moves constantly: to a raise that takes half their week for four months, to a customer emergency, to a co-founder disagreement. A company that only moves when the founder is watching cannot survive its own fundraise, and that is the failure this role exists to prevent.

Speed beats elegance here by a wide margin. A rough tracker people update beats a good system that arrives in six weeks. Build everything as though it will be thrown away, because it will be, probably within a year.

When to use this, and when not to

Use it when the company is under about thirty people with no functional leaders, when a founder is hiring their first operations or business generalist, when someone has just started in the role at a very small company, or when the diagnosis in chief-of-staff-by-stage returned the builder archetype. Use it also when a large round has been raised but the company is still run out of the founder's head, since that is a seed company with more money.

Do not use it once functional leaders own their numbers and headcount is past thirty; the job has become handover and rebuild, which is cos-at-series-a. Do not use it to design a permanent operating model, because nothing built at this size should be permanent. Do not use it as a route to becoming a co-founder by accumulation; the transfer register exists to prevent that.

An executive assistant, a first finance hire and a first recruiter are adjacent and different jobs. Where the real need is calendar and inbox, say so rather than hiring a Chief of Staff to do it, because the person will leave.

What you need before starting

The founder's last three weeks of calendar. The most useful input and the one people skip: it shows where the time actually goes, which is never where the founder says it goes. Missing: ask them to list the last fifteen things they were pulled into and who else could have handled each. Where a connected calendar tool exists, categorise three weeks from it; where it does not, do it on paper in an hour.

Runway and burn, to the month. They set how much of the year is available and whether a raise is about to consume the founder. Missing: build this first, before anything else. A company that cannot state its runway in one number is making every other decision blind.

Headcount and who actually does what. Not titles, which are decorative at this size. Missing: write it out in one sitting with the founder and watch for the roles two people both think they hold.

The list of what is broken, unowned, or argued about twice. Collected from every person in the company, in their words. Missing: that is weeks one to three of the method. Do not skip ahead to fixing.

What the founder has already tried and abandoned. Every seed company has a graveyard of tools bought and never used. Reintroducing one without knowing why it failed destroys credibility in week four. Missing: ask which tools were bought and stopped, and why.

Commitments already made to investors or customers. An update promised and not sent, a feature promised for a date. Missing: read the last three investor updates and the two largest contracts before building anything.

The method

1. Weeks one to three: learn and list, and fix nothing that takes longer than an hour. Sit in every recurring meeting, read the last three months of the shared drive and the investor updates, and talk to every person about what they are working on, what is unclear, and what they are waiting on. The urge to demonstrate value in week one by installing a tool is the most common way this role starts badly.

2. Build the founder time ledger. Categorise three weeks of their calendar into four buckets: decisions only they can make, decisions they make because there is no route around them, information gathering, and external work such as fundraising and customers. The second bucket is your entire job. Size it in hours a week and write the number down; it is the baseline you will be judged against at ninety days.

3. Weeks four to six: pick exactly three things. Score each candidate on founder hours returned per week against days to build, and take the top three. Three is a limit, not a suggestion. A seed company absorbs three changes a quarter, and a fourth means none of them are adopted.

4. Build the crude version and use it once yourself before showing anyone. Run the meeting with your own agenda before proposing the format; fill the tracker with real work before asking anyone else to. Systems introduced without a worked instance get politely ignored.

5. Give every artefact a named future owner on the day it is built. Into the transfer register, with the trigger that will move it: a role hired, a headcount crossed, a date. Two minutes at the time, impossible to reconstruct at eighty people, which is exactly when accumulated ownership becomes the reason the role fails.

6. Weeks seven to twelve: take one area off the founder's desk completely. Not helped with, owned. Choose on two criteria: expensive in founder hours, light in founder judgement. Investor updates, recruiting coordination, and tooling decisions are the usual candidates. Report on it rather than asking about it, which is the behaviour that earns the second area.

7. Install the writing habit, one habit at a time. Start with the decision log, because it is cheapest and the returns are immediate: half the chaos at this size is relitigating things nobody wrote down. Add a second written habit only once the first has held for a month.

8. Refuse work with a rule rather than a judgement. Take anything that returns founder hours or prevents a repeated failure. Decline anything that is a function's job once that function exists. Time-box everything else to a stated number of weeks, with the end date agreed at the start.

9. Review at ninety days against the ledger. Recount the calendar in the same four buckets. The comparison is the deliverable, and it makes the next ninety days easy to scope.

The six artefacts, in build order

1. The weekly leadership rhythm. One meeting, thirty to sixty minutes, written agenda, decisions recorded, actions with an owner and a date. This fixes more than everything below it combined.
2. One visible list of what is happening. Whatever the team will genuinely update. A shared document that gets updated beats a tool that does not.
3. The decision log. Date, decision, decider, one line of reasoning, kept where people already look.
4. The investor update. Monthly, fixed date, same shape every time, sent in a bad month as well as a good one. investor-update has the format.
5. A hiring process that exists. A scorecard, a consistent question set, a decision meeting. Companies at this size hire on impressions and pay for it for years. hiring-scorecard-and-interview-kit has the method.
6. The numbers, in one place. Five to eight figures the founder actually looks at, refreshed on a schedule, every figure traceable to a source.

Nothing else. A seed company does not need an objectives framework, a competency matrix, a values workshop, or a strategy document with three horizons. Building one is a reliable sign the role is filling its own time, and everyone can see it.

Decision rights, and the meetings to refuse

Decide alone: process, scheduling, the shape of documents, who is in a meeting, and tooling or vendor selection within a stated threshold, usually a few thousand a month at this size.

Escalate: anything about people, anything that changes what the company is building, anything a customer will see, and anything the founder has an unstated opinion about, which at this stage is most things worth asking about once.

Run: the weekly leadership meeting, a short all-hands on a predictable cadence, and the hiring decision meeting. Refuse, politely: standing coordination meetings a document would handle, project meetings with no decision in them, and anything recurring whose only owner is you. At twenty people an hour-long weekly meeting with six people in it costs six hours against a company week of roughly eight hundred, which is three quarters of a percent of the company's total working time and is worth saying out loud.

The habit that earns everything else: never surprise the founder in front of someone else. Disagree in private, in advance, as often as necessary.

Worked example

Situation. Tessellate, a company selling inventory software to independent pharmacies. Twenty-two people, fourteen months after a 3.4 million seed round. Burn 210,000 a month, eleven months of runway, recurring revenue around 640,000. The founder intended to start raising a Series A in five months. There were no managers and no recurring meetings except a Monday all-hands that ran ninety minutes and decided nothing.

Task. Get the company to a state where it kept moving through a four-month raise, and produce enough operating history for the Series A conversations. Good meant the founder recovering at least eight hours a week, and no critical decision waiting on them for more than two days.

Action. Three weeks of listening produced a list of forty-one items. The time ledger showed the founder in 51 hours a week of meetings and messages, of which 17 were decisions nobody else was allowed to make: pricing exceptions, hiring approvals, refunds, tooling purchases, and every customer escalation.

The wrong turn happened in week three. An objectives framework was drafted, with quarterly objectives, key results and a scoring scheme, because it was the thing a Chief of Staff was supposed to build. It was presented in week four, agreed politely, and never updated by anyone including its author. It was abandoned in week six. The diagnosis: a twenty-two person company with one product and eleven months of runway already knew what it was doing this quarter, and the framework added ceremony to an understanding that already existed. It cost eleven days.

The three chosen instead: a Wednesday leadership meeting with five people, agenda closed at noon the day before, decisions logged in the same document; a pricing exception rule letting the sales lead approve discounts up to 15 percent alone, with anything larger batched to the founder once a week; and a monthly investor update owned entirely by the role from month two. The transfer register was written the day each was built, marking the hiring process for the first people hire at forty employees and the numbers workbook for the first finance hire. By week nine the role owned recruiting coordination outright, with the founder attending only final conversations.

Result. At the ninety-day recount the founder's meeting load was 39 hours a week and the decisions-only-they-can-make bucket had fallen from 17 hours to 6. The pricing rule alone removed about four hours a week and shortened the sales cycle, because discounts stopped waiting for the founder to be free. The raise started in month five and took nineteen weeks, during which the founder was substantially absent; revenue kept growing and two hires were made without them in the room until the offer. The decision log was used twice in diligence to explain why particular commitments had been made, which nobody had anticipated.

What did not work: the visible list of work in progress was rebuilt three times and never held. The conclusion recorded at ninety days was that the weekly meeting was doing that job adequately and the tracker solved a problem the company did not yet have.

A second scenario, where it goes differently

A nine-person materials science company, two founders, 5 million seed, twenty-six months of runway, no customers and no revenue expected for two years.

The founder time ledger came back nearly empty: there were almost no routable decisions because the company was running experiments. The bottleneck was recruiting, since the plan depended on hiring four specialists in eighteen months from a worldwide pool of perhaps two hundred people.

Two artefacts were built rather than six: the decision log, because the scientific decisions and their reasoning were the company's actual asset, and a monthly investor update, because the investors had no revenue to read. The weekly leadership meeting was refused as premature at nine people who ate lunch together daily. Instead the role ran recruiting as a tracked programme with weekly outreach targets and became the company's first recruiter for a year.

What changed the plan was not size but where the constraint sat. Where the constraint is talent rather than founder attention, the six artefacts are the wrong list. Run the ledger anyway: its emptiness is the finding.

Output

Three artefacts, delivered together at the end of week six and reviewed at ninety days.

NINETY DAY PLAN
Founder hours returned, target: [number a week, from the ledger]
Weeks 1-3   Learn and list. Nothing fixed that takes over an hour.
Weeks 4-6   Three builds: 1. [item]  2. [item]  3. [item]
Weeks 7-12  One area owned outright: [area, from which date]
Not doing this quarter: [3 to 5 items, with why]

| Founder time ledger | Hours a week, week 2 | Hours a week, week 13 | What moved it | | Decisions only they can make | | | | | Decisions with no route around them | | | | | Information gathering | | | | | External: fundraising, customers | | | |

| Transfer register | Built | Future owner | Trigger to transfer | Done | | Weekly leadership meeting | | Chief executive | Stays | | | Hiring process | | First people hire | 40 employees | | | Numbers workbook | | First finance hire | Finance hire starts | |

Failure modes

Building the framework instead of the fix. Recognise it when an artefact has a scoring scheme, or when nobody has updated it a fortnight after launch. Fix by asking what founder hours it returns this month, and cutting anything that cannot answer.

Fixing in week one. Recognise it as a tool bought before anyone was interviewed. It buys three weeks of goodwill and spends a year of credibility, because it signals the role is here to install its previous company's systems.

The founder routes around the system. They approve a discount in a corridor or hire someone outside the process. Recognise it as the system being followed by everyone except them. Fix by naming it privately, once, with the specific instance, and asking what about the process is too slow, because usually something is.

Becoming the permanent owner of everything. Recognise it when the transfer register is empty or carries no dates. This is the failure that ends the role at eighty people.

Building the same artefact three times. Two failed attempts at one thing means the company does not have that problem yet. Record the conclusion and stop.

Nothing written down. Recognise it when the answer to why we do it this way is a person's name rather than a document. Install one writing habit at a time, starting with the decision log.

Edge cases

Two founders who disagree. The role cannot fix this and will be damaged trying. Do not carry messages between them and never become the channel through which they communicate. Name it once, in private, as the constraint it is, and otherwise route decisions to whichever of them owns the area.

Under six months of runway. The method suspends. There is one job: cash, the raise, and the shortest possible list of what keeps customers. Read fundraise-readiness and come back after the round.

The founder does not actually want the decisions taken away. Some do not, and they will say yes and behave otherwise. Test with one small area for a month and read the behaviour rather than the answer. Where the pattern holds, the honest options are narrowing the role to execution or leaving, and both beat two years of building systems that get routed around.

The company is pre-product. With no customers and no revenue the six artefacts drop to two, the decision log and the investor update, and the value of the role is usually recruiting or the research operating rhythm. Below about eight people the role is premature altogether.

The first manager appears without anyone noticing. At around eight people someone acquires reports informally, with no training and often without wanting them. Get it made explicit and make sure they have somewhere to take questions, because unmanaged first-time managers are the largest source of early attrition.

Quality bar

• The founder time ledger exists at week two and again at week thirteen, with hours in both, and the difference is stated.
• Exactly one area has come off the founder's desk completely, with the date it transferred.
• Every artefact has a named future owner and a transfer trigger, recorded on the day it was built.
• Nothing built is more elaborate than the company can maintain without you.
• Decisions are written down on the day they are made, in one place people already look.
• What is deliberately not being done this quarter is written and visible.
• No recurring meeting was added without one being removed.

Adapting this to your context

The defaults here come from venture-funded companies of five to thirty people with a single founder and a runway measured in months. Read the signals, not the round.

• "Seed" as a label. It stands for a company under about thirty people with no functional leaders, where the founder is the routing table and the board is one or two people. A family business, a two-partner agency, an NGO before its first programme director or a bootstrapped firm at that size is a seed company whatever it calls itself, and the funding language here can be ignored.
• The founder time ledger. Built from a calendar. Where the principal's day happens on a shop floor, a ward or a site rather than in meetings, reconstruct it from a fortnight of interruptions logged as they happen.
• Runway and burn. The pressure input, not the point. Substitute the seasonal cash trough, the grant period or the months of reserves where nobody raises money.
• The six artefacts. Ordered for a company selling software. A charity puts the funder report where the investor update sits; an owner-managed business needs the numbers page first.
• What not to change. Three changes a quarter and no more, and a named future owner with a transfer trigger recorded on the day each artefact is built.

Related skills

chief-of-staff-by-stage diagnoses whether this is the right playbook and produces the charter that scopes the role. cos-at-series-a is where this hands over, once functional leaders own their numbers and headcount passes thirty; the transfer register built here makes that handover a week's work rather than a quarter's.

operating-cadence-design has the full method for the weekly rhythm sketched here. investor-update and hiring-scorecard-and-interview-kit are the two artefacts worth building properly rather than crudely. decision-memo is the format for anything escalated to the founder. fundraise-readiness takes over when runway is short.