Avakata's org chart has nine boxes. Two contain humans. The other seven are software — named roles with job descriptions, defined deliverables, and a weekly review, implemented as agents and pipelines. Drawing that chart before picking any tools is the single decision that kept our stack coherent while everyone around us collected subscriptions.
Most solopreneurs assemble tools and hope a business emerges. Draw the chart first and the stack designs itself.
Here is ours, box by box, with what each role costs and how the reporting lines actually work.
Draw the chart before the stack
The exercise takes 45 minutes. Draw the org chart your business would need at ten times its current size — every role, nine to twelve boxes, as if you were staffing it with people. Then walk the boxes one at a time and mark each: human, software, or not yet. Every tool decision afterward becomes a hiring decision with a job description attached, instead of an impulse buy with a login.
The chart kills the shiny-object problem at the root. A tool that does not fill a box on the chart does not get bought, however good the demo.
It also exposes the opposite failure: boxes nothing fills. Ours had no bookkeeper for a year. The chart made the gap embarrassing enough to fix.
The not-yet label matters as much as the other two. It stops you from building ahead of need, which is the software version of hiring ahead of revenue.
Paper first. Stack second. The order is the method.
The nine boxes on ours
Ours reads like any small agency's chart. A CEO and a client lead, both human, both me — the boxes stay separate because the jobs are different even when the person is not. Then seven software roles: a research analyst, a content writer, a critic that gates everything before publish, a publisher that handles deploys and distribution, a GEO monitor running citation audits, a sales-ops role keeping the pipeline warm, and a bookkeeper reconciling the money weekly.
Each software box has a one-page job description: inputs, outputs, tools it may touch, and the numbers it is judged on.
Writing those pages felt like paperwork theater the first week. It is now the most-read document in the company, because it is the spec.
Nine boxes, two salaries, one of them mine twice.
The two boxes that stay human
The rule that decides human versus software: any role where being wrong costs a relationship stays human. That leaves two boxes. The CEO box owns strategy, pricing, and what to say no to. The client lead box owns trust — the calls, the hard conversations, the judgment about what a specific client actually needs this month. Everything else produces artifacts, and artifacts can be reviewed before they cost anything.
Notice what is not on the human list: writing, research, monitoring, bookkeeping. Prestige does not earn a box human status. Consequence does.
The two human boxes consume about 12 hours a week between them. That is the entire human payroll of the agency.
Guard those hours. They are the only thing on the chart that does not scale.
Reporting lines are contracts
In a human org chart, a reporting line means meetings. In a software org chart, a reporting line is a typed contract: every role receives a brief in a defined shape and must emit two things — a deliverable and a log of what it did and why. The writer emits a draft plus its sources. The critic emits pass or fail plus reasons. The monitor emits a diff of citations gained and lost.
The contract test is the automation test. If you cannot write down what goes in and what must come out, you do not have an automatable role. You have a vibe.
Logs are non-negotiable. A role that cannot explain its work gets fired from the chart, software or not.
Contracts also make roles testable. Because the critic emits pass or fail plus reasons, we can replay a month of drafts through a new critic implementation and diff the verdicts before trusting it with anything live.
We validate every handoff against its contract automatically. A broken contract fails loudly at the seam, not quietly in the client deliverable.
The staff meeting takes 40 minutes
Management overhead for seven software roles is one meeting: Monday, 8:40 to 9:20 in the calendar, me and seven logs. Read each role's weekly log, approve or reject the two or three items queued for human sign-off, adjust one brief if a number looks soft, done. Forty minutes to manage the equivalent of seven junior functions is the leverage that makes a one-person agency viable at all.
The meeting has an agenda like any staff meeting: wins, exceptions, escalations, changes. The difference is that nobody performs attendance.
Roughly one Monday in four, something in a log earns a deep dive — that is where the real management happens.
Escalations carry a service-level rule like anything else: whatever a role flags as blocked gets a human decision within one business day, or the role routes around it by a rule it already knows.
Skip the meeting twice and you will feel it. Software drifts politely, then all at once.
Hiring, firing, and promotion for software
The chart gets managed like any org. Hiring: a new box opens only when a recurring task shows up in my own hours three weeks running — recurrence is the requisition form. Firing: each role's implementation gets challenged quarterly, and if a new model or tool beats the incumbent on the golden set by 20% on quality or cost, we swap it. Promotion: a role that runs eight clean weeks gets a wider allowlist and fewer approval gates.
The job description survives every swap. We have replaced the writer role's underlying model three times without touching the contract around it.
That separation — stable role, disposable implementation — is what makes the stack durable while the tools churn.
Software does not get feelings about performance reviews. It is the easiest team you will ever manage honestly.
What the chart costs
The seven software roles cost about $410 a month all-in — model usage around $140, subscriptions and hosting the rest. The first junior hire in any of those boxes would cost roughly $4,800 a month before overhead, plus the management hours a human deserves. The chart is not equivalent to a team of seven people, and pretending otherwise is how the pitch gets oversold. It is equivalent to leverage: nine functions running for less than a car payment.
The honest comparison: seven part-time junior functions with perfect attendance and no growth ambitions, supervised by two human roles that do all the judgment.
Draw yours this week. Forty-five minutes, twelve boxes, three labels. The chart will tell you what to build next — and if you want ours as a starting template, it is the first artifact of every Avakata strategy engagement.