AI automation for small business, measured in reclaimed billable hours.

Because integrations and tasks are vendor metrics, not owner metrics. A small business owner cares about one number: did my billable utilization go up. A solo consultant who works 45 hours and bills 25 is at 56 percent utilization. The interesting question is which of the other 20 hours per week can be moved into the billable column or off the calendar entirely. Integrations are an input; reclaimed hours are an output. The workflow doc this guide is built on (consulting-business-workflow.md, Phase 4 weekly rhythm) names the admin slots explicitly: Monday for status updates and planning, Friday for invoicing and bookkeeping and pipeline review, ongoing for client comms with a sub-24h SLA. Those are the slots to grade any tool against.

Pure arithmetic. 21 admin hours per week, multiplied by a $150 hourly rate (the median rate from the same workflow doc), multiplied by 46 worked weeks per year (52 minus 6 weeks off, again per the workflow doc). 21 x 150 x 46 = $144,900. This is not the amount Clone promises to reclaim. It is the size of the surface that any AI automation should be measured against. Even reclaiming a third of that, around 7 hours per week, is roughly $48,300 of resellable capacity per year per consultant. The point is to give you a denominator the SERP listicles refuse to publish.

Those lists are organized around workflow tools that wire triggers to actions. Clone is in a different shape: it is a desktop agent that drives the apps you already use the way a human would, scheduled at named admin slots, with a per-client memory layer. The reason it does not show up on most listicles is that those listicles grade on the wrong rubric. If the rubric is integration count, Clone scores zero because it does not need any. If the rubric is hours reclaimed per Friday-admin block, Clone is the only tool that even publishes a target.

The pattern is illustrative, not prescriptive. Substitute your own slots. A bookkeeper might have Tue-payroll, Wed-AR, Thu-AP, Fri-month-end. A real-estate agent might have Mon-listings-update, Wed-showings-prep, Fri-CDA-review. The setup is identical: write one ritual file per slot in plain English, set a target_hours_reclaimed at the top, run it under your supervision once, then schedule it. The slots are yours to define; the technique of grading by reclaimed hours is the part that ports.

Both, at first. The first two weeks of any ritual you spend almost as long reviewing as you used to spend doing. By week three the per-client memory has captured the corrections you made in weeks one and two, and the drafts arrive correct enough that scanning takes a fraction of the original time. By week six the typical pattern is 1 minute of review per draft, where the original task took 8 to 15 minutes. The weekly delta report (clone weekly-delta) shows you the actual reclaimed-hour count against your baseline. If a ritual is not pulling its weight by week six, the report flags it and you edit the ritual file or retire it.

No. The Computer Agent drives whatever app is on screen the way a human would: clicks fields, types, scrolls, reads back what is displayed. The same instruction works whether you use QuickBooks Online or QuickBooks Desktop, whether your invoicing is in Wave or Clio, whether your CRM is HubSpot or Pipedrive or a custom Airtable base. There is no integration to provision, no OAuth flow, no admin permission to wait for. The architectural principle on cl0ne.ai is explicit: 'Clone uses the apps you already pay for. Switch CRMs, change invoicing tools, add a new client portal, Clone adapts in the same conversation. No re-wiring required.'

Three differences that matter for a solo consultant. First, cost: a competent VA who can run the Friday-admin block is $3,000 to $6,000 a month according to the tech-tools-landscape doc; Clone is $49 a month on Solo. Second, hours: a VA gives you 20 to 40 hours a week of human capacity, scheduled in their timezone; Clone runs every scheduled ritual on time and triages comms continuously without a timezone constraint. Third, ramp: a VA needs 4 to 8 weeks to learn your clients' tone, billing quirks, and escalation rules; Clone captures the same context faster because every correction you make goes back into the per-client memory file the same minute.

For most owners it is Friday invoicing. The reasons: the review step is the cleanest (drafts in your invoicing tool, you click send), the inputs are the most repeatable (last week's billable lines from your timer or notes), and the apps it touches are the messiest to integrate any other way (QuickBooks Desktop, Sheets, Clio, custom bases). The typical first-month outcome is reclaim 1.5 to 2 hours every Friday plus eliminate the late-Friday-night invoicing crunch entirely. That alone covers the subscription roughly 30 times over at a $150 rate.

The 45/25 numbers are a representative case from the workflow doc, not a constraint. Substitute your own. A consultant who works 60 hours and bills 30 has 30 admin hours, which puts the resellable surface at $207,000 a year at the same $150 rate. A part-time owner who works 25 hours and bills 12 has 13 admin hours, which is still $89,700 a year of recoverable capacity. The math compounds against any baseline. The sales pitch of every Zapier listicle ('save up to 5 hours a week') quietly assumes you work 40 hours and the saved hours have no replacement value; this guide assumes you have a rate and the saved hours convert to revenue.

There is a built-in weekly delta report. Run 'clone weekly-delta' from the menu bar or from your terminal and it prints a summary per ritual: scheduled time, run time, output count, hours reclaimed against the prior 4-week baseline, plus a weekly total in dollars at the rate you set in each ritual file. The report writes a markdown file under memory/reports/ so you can grep history later. This is the part the listicle vendors refuse to publish because their incentive is to inflate task counts, not to publish the only number you actually care about.

One ritual file, one full week of running it, plus a baseline you can subtract against. Most owners install Clone, write the Friday-admin ritual on a Tuesday, run it under supervision the next Friday, then let it run on its own for the following two Fridays. By the third Friday you have a credible reclaim number. Add a second ritual at week three (most pick Monday status), and by week six you have meaningful weekly delta numbers across two slots. There is no quarter-long onboarding curve. The ritual file is the configuration, and you can ship it the same day.