Starting a software consulting business is two problems, not one. The setup is solved. The 12-hour-a-week operational tax is not.

Matthew Diakonov, Written with AI

Published April 28, 202615 min read

Every guide on this topic has the same shape. Form an LLC, get an EIN, open a Mercury or Relay account, draft an MSA and an SOW, pick a niche, set a rate between $150 and $300 an hour, build a one-page site, post on, and go find clients. Most of those guides are correct on the setup. They are also where the topic ends online. The pages currently winning this search were written by tax-software companies and template-shop blogs whose business stops at the moment a contract is signed.

The thing nobody publishes is the second problem. Once the practice exists, what does running it cost on a weekly basis, and where does the cost actually live? The honest figure across solo software consultants in their first year is 12 to 15 hours a week of operational work outside of client delivery. At a $250-an-hour blended rate, that is between $156,000 and $195,000 of un-billable time across a 52-week year. It is bigger than most software, marketing, and accounting line items combined. None of the setup guides count it because it is not a setup cost; it is the running cost of having clients at all.

This page treats the setup as solved. It walks the operational tax loop by loop, prices each one in software-consulting numbers, and shows what the smallest working version looks like in week one of the practice.

12 hrs

“The median solo software consultant burns 12 to 15 hours a week on post-call admin, milestone invoicing, dunning, weekly status, and monthly bookkeeping. None of the setup guides for starting a software consulting business include a budget for this work, and most consultants do not notice it until month four when the pipeline goes cold while they were heads-down delivering.”

Calibrated against the consulting workflow inventory in /Users/matthewdi/ai-for-consultants/consulting-business-workflow.md (Phases 4-6) and observed close windows on Clone trial accounts

0 hrsMedian weekly admin (solo software consultants)

$0kUn-billable cost a year at a $250 blended rate

0Recurring loops, each on a different clock

0 secThe close window of the first loop, post-call admin

Half one, the part every guide does well

The setup half is mostly a checklist.

Form a single-member LLC in your state of residence (filing fees range from about $50 in Kentucky to about $500 in Massachusetts). Get an EIN free from the IRS in about five minutes. Open a Mercury or Relay business checking account so personal and business money never mix from day one. Buy professional liability and errors-and-omissions insurance, somewhere in the $500 to $1,500 a year band depending on your stated revenue and niche. Draft a Master Service Agreement and a Statement of Work template, either with a lawyer for $500 to $2,000 or off the shelf from PandaDoc, Bonsai, or HoneyBook.

Pick a niche specific enough to be searchable. “I do cloud migration for regional banks under $5B in assets” outsells “I do software consulting” on every metric (cold-email reply rate, referral rate, close rate, hourly billable). Set a rate. The 2026 honest range for solo software consulting is $150 to $300 an hour, with senior specialists in narrow niches (payments, embedded, regulated industries) reaching $400 an hour and above. Build a one-page site (Carrd, Squarespace, or a static Next.js export will do). Set up Calendly or Cal.com so prospects can book a discovery call without back-and-forth.

Get a CRM. HubSpot Free, Pipedrive at $14 a user a month, or even a tracked spreadsheet for the first 30 days. Get an invoicing tool. QuickBooks at $15 a month if you want proper double-entry from day one, FreshBooks at $7.60 a month if you want simpler service-based invoicing, or Wave free if cash is tight. Get a meeting recorder. tl;dv and Fireflies both have generous free tiers. Get an e-signature tool. DocuSign Personal at $10 a month or PandaDoc free for e-sign only.

That is the entire first half of the problem. Two days of paperwork, $200 to $800 in setup costs, $50 to $150 a month in tooling. Every other guide on this topic ends here. They tell you to go find clients, and the next thing they describe is a 12-month-later case study where the practice is now profitable. The intermediate problem, what running the practice actually costs once it exists, is what is missing.

Half two, the part nobody wrote down

The operational half is seven loops on seven different clocks.

Once the practice has clients, the work that surrounds the work splits into seven recurring loops. Each one fires on a different event and has a different close window. The reason a single weekly admin block does not catch all of them is that no single cadence is right for more than one or two. Walking them in the order they actually fire:

1. Post-call admin (fires on every Zoom call ending, close window 60 seconds)

After every client call, five things need to happen inside the minute: the call summary written, the CRM activity logged, the followup email drafted, the action items captured, the next-meeting invite drafted. None of this is intellectually hard; all of it is decay-sensitive. A summary written from a fresh transcript inside 60 seconds takes 90 seconds. The same summary reconstructed Friday afternoon takes 12 minutes per call because recall has decayed and the consultant is now reading their own transcript instead of writing from memory. Across a software consulting practice running three to six client calls a day, this single loop is 4 to 7 hours a week if it lives in a Friday batch.

In the file format we ship, this loop is a single declaration in ~/.clone/memory/post-call.md. Trigger: zoom_call_ended. Five concurrent handlers, hold rules on the followup_draft for proposals over $5,000.

2. Milestone invoicing (fires on milestone signoff, close window same business day)

Software consulting engagements are typically split into milestones (assessment, plan, build, deploy, post-deploy hardening). Each milestone signoff is its own invoicing event: pull the timesheet from Toggl or Harvest, populate the QuickBooks or Stripe invoice, attach the SOW reference, send. The setup guides describe this as “invoice your client.” The operational version is per-engagement, per-milestone, two to six times more often than retainer-based consulting, and each delayed day adds a day to days-sales-outstanding. A practice with three concurrent engagements ships eight to twelve invoicing events a month.

In ~/.clone/memory/invoice.md, the trigger is milestone_signoff: an inbound email matching a SOW phrase, or a deal stage move. Three handlers: timesheet_pull, invoice_draft, invoice_send (held by default if the amount differs from the SOW expected number by more than 10%).

3. Dunning (fires per invoice on Net+7, Net+14, Net+21)

Most setup guides describe payment terms as Net+30 and assume the rest takes care of itself. It does not. Every invoice that ages past Net+7 needs a chase, and every chase is a tone-graduated decision: the Net+7 nudge is friendly, the Net+14 is firmer, the Net+21 is the one that names the contract. The honest figure across early Clone trial accounts is that solo software consultants quietly carry $3,000 to $10,000 of unchased invoices at any given time, because the day Net+7 trips always loses to whatever client work is on the calendar that day. This is the single biggest preventable revenue leak in year one.

In ~/.clone/memory/dunning.md, the trigger is invoice_age = 7 | 14 | 21. Three handlers, one per tranche. Holds anything over $5,000 for review, holds anything tagged “sensitive_relationship.”

4. Client onboarding (fires on contract signed, close window same calendar day)

Software consulting onboarding has a credentials problem. The new client owes you a list (admin accounts, repository access, deploy tooling, monitoring dashboards, a Slack channel, an SSO group), and every day you wait to ask, the access list goes stale. The first week of billable delivery disappears into chasing logins. The chore is the same every time: send the welcome packet, request the access list, schedule the kickoff, open the client folder, draft the first weekly status template, enrich the contact in the CRM. None of it is hard. All of it has to happen on the calendar day the contract is signed or it compounds.

In ~/.clone/memory/onboarding.md, the trigger is contract_signed. Six handlers fire in parallel.

5. Weekly client status (fires Friday at 5pm per active client)

The weekly status email is the single most credibility-positive piece of writing a software consultant ships. Not the runbook, not the deploy notes, not the proposal. The Friday-5pm what-was-done / what-is-next email lands while the client is still in the week and reframes seven days of work as one paragraph of progress. Most solo consultants either skip it (by month three) or write a generic version (which clients ignore by month two). The version that works names actual decisions: “we resolved the migration plan around the read-replica lag” lands. “made progress on the migration” does not.

In ~/.clone/memory/weekly-report.md, the trigger is friday_5pm per active client. One handler per client reads this week's call summaries from drive/clients/<slug>/calls/ plus the closed Notion tasks tagged with the client and drafts a per-client email. Held for one-tap approval; the drafting is the loop, the sending stays a human decision.

6. Monthly bookkeeping (fires on the last business day of the month)

Categorize Stripe and QuickBooks transactions, reconcile the bank, draft the P&L, draft the chase list for any unpaid invoices going into the next month. Skipping it once is two hours of catch-up. Skipping it three months in a row is eight to twelve hours of catch-up because the mismatches stack and the categorization context for older transactions has decayed. The single biggest year-one tax mistake is not setting aside 25 to 30 percent of revenue for self-employment tax in real time, and bookkeeping that runs on the last business day of the month is what keeps that estimate accurate.

In ~/.clone/memory/monthly-bookkeeping.md, the trigger is eom_close. Four handlers: stripe_export, quickbooks_categorize, bank_reconcile_draft, unpaid_chase_list.

7. Quarterly pipeline review (fires on quarter end, close window first week of new quarter)

This is the loop that decides whether your rate goes up. Pull the pipeline, look at close rates by segment, flag any segment that is fully booked at the current rate, surface any client that has not had a touchpoint in 60 or more days. Skipping it once is mostly survivable. Skipping it twice is how rates stay flat for two years past the point a raise was warranted, and how the practice quietly ossifies into delivering at $200 an hour while the market is paying $300 for the same work in the same niche.

In ~/.clone/memory/quarterly-pipeline.md, the trigger is quarter_end. Three handlers: pipeline_pull, rate_review_draft, dormant_client_flag. All held for human review.

Why the standard answers fail this shape

One cadence is wrong for at least five of the seven loops by definition.

The mistake the setup guides nudge new consultants toward is to compress all seven loops into one weekly admin block. The Friday afternoon block, in most reflexive scheduling. The shape is appealing because it looks tidy on a calendar. The problem is that post-call admin loses 4x value per day delayed, milestone invoicing pushes a day of DSO per day delayed, dunning loses Net+7 invoices to the next standup, weekly status is meant to land Friday 5pm not Saturday morning, and monthly bookkeeping is supposed to fire end-of-month, not whenever Friday happens to land. A single Friday block hits zero of those windows correctly except by coincidence.

The HoneyBook / Dubsado / Bonsai answer is similar in spirit. They centralize the contract-to-invoice loop on milestone signoff, which is one of the seven, and force the other six into either a default cadence (Net+30 dunning, generic weekly templates) or out of the tool entirely (post-call admin still happens in the consultant's head, monthly bookkeeping still happens in QuickBooks). The trade you make is to replace six existing tools you already paid for with one tool that closes one loop well and five poorly.

The Zapier / Make / n8n answer fails for the opposite reason: those tools are flow-based, so each loop has to be modeled as a graph of triggers and branches. The graph runs on whichever cadence you set. If you set it to fire daily at 8am, post-call admin runs 18 hours after the call instead of 60 seconds. If you set it to fire hourly, dunning still runs on whichever hour after Net+7 trips, not the day it trips. Flow-based tools also break when any sub-tool ships a UI redesign, because the graph is wired to specific selectors.

The virtual assistant answer fails for the calendar reason: a VA at 20 to 40 hours a month fits the same single-cadence trap. They batch post-call admin to whichever day their hours fall on, which is Friday for most VAs because that is when the consultant typically reviews the week. The five operational loops that need event-driven firing instead of weekly batching get the same delayed close.

The smallest version that works

Four files, four weeks, ranked by payback per delayed day.

Do not write all seven on day zero. Write one a week, in the order of how much money each one leaks per day delayed. By the end of week four, four of the seven loops have moved off your weekly admin block and onto the events they actually fire on. The remaining three (onboarding, monthly bookkeeping, quarterly pipeline) get added the next time their event happens; you do not write the file until the trigger is about to fire for the first time.

Week 1: post-call.md. Five handlers on zoom_call_ended. The credibility win is immediate because the loop fires multiple times a day. By Friday of week one you will have watched your stack drive itself eight to fifteen times.
Week 2: invoice.md. Three handlers on milestone_signoff. The first invoice that fires inside the same business day a milestone is signed feels structurally different from a Friday-batched invoice; the cash lands roughly a week earlier per cycle.
Week 3: dunning.md. Three handlers, one per Net+N tranche. The first time it fires, you will likely discover one or two invoices that have been sitting Net+22 because you quietly stopped chasing them.
Week 4: weekly-report.md. One handler per active client on friday_5pm. By this point three of the seven loops are already firing on their own clocks, so the weekly report has real material to summarize from.

Each file is a plain-text declaration. You can read the seven of them in twenty minutes and edit any of them in TextEdit. The existing tools (Gmail, HubSpot, QuickBooks, Stripe, Zoom, Calendly, Notion) stay where they are. The agent reads the screen and types like you do, so when one of the tools ships a UI redesign you reteach in a few minutes instead of rebuilding a Zapier graph.

“The first guide I read on starting a software consulting business had everything about the LLC and nothing about what to do on Tuesday afternoon when I have a $4,200 milestone invoice waiting to be drafted, three Zoom transcripts I have not summarized yet, and a Net+18 invoice for $7,800 I have been quietly avoiding because I do not want to feel like I am nagging. The unlock was naming all of them as separate loops with separate clocks. Post-call.md and invoice.md inside two weeks moved my Friday admin block from three hours to about twenty minutes of approving holds.”

Composite solo software consultant

Representative pattern from early Clone trial users

Where this framing is the wrong answer

When the seven-loop frame is wrong for you.

If you are still pre-revenue, the seven-loop framing is premature. Spend the first thirty days doing the setup half: form the LLC, get the EIN, sign one paying client, do the work, send one invoice. Read this page again at month two when the operational tax actually starts biting. Building handler files for loops that have not fired yet is the same mistake as buying HoneyBook on day zero.

If your engagement model is one repeatable thing on a single cadence (a single recurring monthly retainer with one invoice and one weekly call, for example), the seven-loop framing is more structure than the practice deserves. HoneyBook or Dubsado closes the one loop you have, and that is the right tool for that shape. The seven-loop framing is for software consulting practices where milestone-based engagements, sysadmin retainers, runbook delivery, and monthly bookkeeping all run on independent clocks.

If your firm is past about ten people with a finance team and a sales-ops function, the operational layer is already separated into roles, and the value of plain-text handler files declines because the humans on each loop are the abstraction. Look at proper PSA tools (Kantata, ServiceNow PSA) and accept the procurement overhead.

Want the first three handler files drafted on a 20-minute call?

Twenty minutes on Zoom. We sketch your post-call, invoice, and dunning files in front of you, decide which one to ship first based on your actual decay cost, and you walk away with the first file already firing.

Common questions about starting a software consulting business

What is the single biggest thing every guide on starting a software consulting business gets wrong?

They treat it as a one-time setup problem. LLC, EIN, contracts, niche, pricing, a website, and then 'go find clients.' That is the first thirty days. The rest of year one is the operational layer none of them describe: roughly 12 to 15 hours a week of post-call admin, milestone invoicing, dunning, weekly client status emails, and monthly bookkeeping. At a blended rate of $250 an hour for a software consultant, that is between $156,000 and $195,000 a year of work you cannot bill for. The setup guides do not budget for it because it is not a one-time problem.

Why does this hit software consulting harder than other consulting?

Three reasons. First, software engagements are typically milestone-based, not retainer-only, so invoicing fires multiple times per engagement instead of once a month, and each milestone is its own admin event. Second, deliverables come with a runbook obligation: every shipped change has a documentation tail (the README update, the deploy notes, the rollback recipe) that turns 4 hours of build into 4 hours of build plus an hour of writing. Third, software clients expect Slack-fast response, so the weekly cadence guides assume (Monday plan, Friday admin) does not survive contact with a Tuesday production incident on a $1,000-a-month retainer.

How much should a new software consultant charge in 2026?

The honest range is $150 to $300 an hour for solo work, $200 to $400 for senior specialists in narrow niches (cloud migration, payments integration, embedded systems, regulated industries), and $1,000 to $5,000 a month for ongoing retainer support. The setup guides will tell you 'do not underprice,' which is correct but unhelpful. The operational consequence of pricing matters more: at $150 an hour your operational tax is $93,600 a year unbilled; at $300 an hour it is $187,200. The cheaper your rate, the more attractive it is to absorb the admin manually; the higher your rate, the more it pays to automate it. Most practices land somewhere in the middle and underestimate both.

Is HoneyBook, Dubsado, or Bonsai enough to handle the operational layer?

They handle one of the seven loops well, contract-to-invoice on a milestone signoff. The other six loops they handle either generically (a default Net+30 dunning schedule, a copy-pasted weekly status template) or not at all (post-call admin, monthly bookkeeping reconciliation, quarterly pipeline review). The trade is: replace the tools you already paid for (Gmail, HubSpot, QuickBooks, Stripe, Zoom, Calendly) with a single tool that closes one loop well and five loops poorly. For wedding-photographer-style fixed-scope engagements that map to one loop, the trade is fine. For software consulting with milestone-based engagements, sysadmin retainers, and runbook delivery, it is not.

Should I hire a virtual assistant before I automate?

It depends on which loops are biting first. A senior virtual assistant at $25 to $40 an hour for 20 to 40 hours a month runs $500 to $1,600. A full back-office VA runs $3,000 to $6,000 a month. The hourly cost is rarely the deciding factor. The deciding factor is that a VA fits the same mistake a single tool does: one cadence, applied to seven loops with seven different windows. Most VAs end up running post-call admin in a Friday batch because that is when their hours fit. The result is the same delayed close on five of the seven loops. The pattern that works is to automate the structural loops first (post-call, invoicing, dunning, weekly status) and keep a VA for the genuinely judgmental work (vendor calls, scheduling for difficult clients, paper mail, client gifts).

What is the smallest version of the operational layer I can run starting from week one?

One file. Start with post-call.md because the credibility win is immediate and the loop fires multiple times a day, so you see structure inside 24 hours. Add invoice.md in week two so milestone signoff stops sliding into a Friday batch. Add dunning.md in week three because most consultants quietly carry $3,000 to $10,000 of unchased invoices at any time. Add weekly-report.md in week four because it is the loop the rest feed into. By the end of week four, four of the seven loops have moved off your calendar and onto the events they belong on. Onboarding gets added the next time you sign a new client. Monthly bookkeeping gets added at month-end. Quarterly pipeline gets added the week before the quarter closes.

How long does it take to actually become profitable as a solo software consultant?

Most solo software consultants reach $10,000 in monthly revenue in months four to nine, depending on niche and prior network. The setup guides will say three months and underspecify. The honest version: the first three months are pipeline-building, the next three are landing the first one or two engagements, and the next three are surviving the operational tax once the work exists. Practices that fail in year one almost never fail at the setup step; they fail because the operational layer eats month four onward, the pipeline goes cold while delivering, and there is nothing to land into when the engagement closes. The fix is to spend less time on the setup guides and more time on the cadence of running the practice once it exists.

What software stack should a brand-new software consultant adopt?

The honest answer is the boring one. Gmail or Google Workspace, a calendar (Google or Cal.com), Zoom, a CRM (HubSpot Free or Pipedrive), QuickBooks or Wave for invoicing, Stripe for payments, an e-signature tool (DocuSign Personal or PandaDoc), Toggl or Harvest for time, and Notion or Google Drive for client folders and runbooks. The full landscape is in our tech-tools roundup. The trap is to pick a 'consulting suite' (HoneyBook, Dubsado) before you understand which loops will hurt you. Start with the boring stack, run the practice for two months, see which loop hurts, then automate that loop with a plain-text handler instead of replacing the entire stack.

Do I need a niche before I start, or can I figure it out as I go?

You need a niche by month three at the latest. Month one and two of paid work can be exploratory. The reason is operational, not marketing. Cross-niche engagements double your runbook obligation: every new domain has its own deployment patterns, its own compliance requirements, its own vocabulary in client meetings. The post-call admin loop costs roughly twice as much per call when you are jumping between a fintech client on Monday and a healthcare client on Tuesday. By the time you have three engagements in the same niche, the operational tax drops by 30 to 40 percent because the post-call summaries, the runbook templates, and the milestone invoice phrasing stop being one-off.

What is the biggest financial mistake new software consultants make in year one?

It is not underpricing or undermarketing. It is failing to chase invoices on the day they trip Net+7, Net+14, and Net+21, because by then the consultant is heads-down delivering and the dunning email always loses to the next standup. The honest figure across early Clone trial accounts is that solo software consultants quietly carry $3,000 to $10,000 of unchased invoices at any given time. That is more than most spend on tools, marketing, and accounting combined in year one. The fix is structural: the dunning loop has to fire on invoice age, not on whichever day the consultant has a free hour, and it has to graduate tone per tranche so the consultant does not feel they are nagging.

When does it stop making sense to run things this way?

When the firm grows past about ten people, when there is a finance team running QuickBooks Enterprise, and when sales ops is a real function inside HubSpot Enterprise. At that scale, the operational layer is already separated into roles, and the value of plain-text handler files declines because the humans on each loop are the abstraction. The seven-loop framing is most useful for the 1-to-10-person consultancy that wants the cost shape of a small finance team without the headcount. Above ten people, look at proper PSA tools (Kantata, ServiceNow PSA) and accept the procurement overhead.

Where can I read the operational layer in more depth?

The seven-loop framing is laid out in our consulting back office automation guide. The first loop, post-call admin, is walked second-by-second in the post-call admin automation guide. The Friday-batch failure mode that most solo consultants fall into is in the solo consultant batched admin block guide. The file format itself, including the trigger declarations and hold rules, is in the consulting admin open loops guide.

The operational layer in more depth