Workflow automation consulting bills for a canvas Clone refuses to ship.

Workflow automation consulting is a service where a specialist (a Zapier Expert, a UiPath partner, a boutique ops firm) maps your business processes into a vendor canvas and wires the triggers, branches, and selectors by hand. The engagement has four phases: discovery, design, build, and maintenance. The deliverable is the canvas itself: a Zapier workspace with 10 to 30 Zaps, a Make scenario library, a UiPath Studio project compiled to Orchestrator, a Power Automate solution in your tenant, or an n8n JSON graph running on a VPS.

Two reasons. First, the discovery phase: four to eight weeks mapping your processes into BPMN-style diagrams so the consultant can design the node graph, usually billed at $150 to $300 per hour. Second, the build phase: every trigger, every branch, every conditional path, every selector in the canvas has to be drawn, tested, and connected to a vendor tenant. A typical small-consultancy engagement hits $15,000 at the floor. Enterprise RPA engagements (UiPath, Blue Prism) regularly cross $50,000 and move into six figures with managed services. This is not gouging; it is a fair price for the labor of wiring a canvas. The question the rest of this page asks is whether the canvas should exist at all.

The canvas. Clone's architecture.tsx line 13 declares 'Clone Planner' as layer two of the shipped stack, with sublabel 'Understands intent, picks the right apps.' That layer is the artifact a consultant charges you to build by hand inside a vendor tool. It ships in the product, loaded when the app boots. how-it-works.tsx step one, line 20, states the stance verbatim: 'No workflow builder. No drag-and-drop automations. Just type what you want, the way you'd ask a junior employee. Clone figures out which apps to open, which files to touch, and which data to pull.' Read that sentence again: it is not a marketing claim. It is the architectural refusal to build the thing the consultant was going to bill you for.

The Planner layer (architecture.tsx line 13) reads your English instruction, the Computer Agent layer (line 19) reads the live screen of your actual apps, and the Memory layer (line 25) holds your voice, your templates, and your past runs. Between the three, the product figures out which buttons to click and which fields to fill for the ritual you described. You never draw a box, never name a trigger, never pick a connector. You type a sentence. The product resolves it against the apps you already have open.

No. Zapier and Make are canvases that call vendor-provided connectors over APIs. Clone's Computer Agent does not call a connector; it drives the same UI a human would use. The llms.txt file on the site, line 70, says the difference explicitly: 'Zapier requires you to configure every trigger, field, and branch by hand. Clone understands natural language and adapts to whatever tools are on your screen, including custom software without APIs.' Line 73 for Make and n8n: 'Flow-based automation tools require technical setup. Clone takes plain English instructions. No nodes, no branches, no debugging webhooks.' These are architectural boundaries, not feature comparisons.

They do not exist. On the traditional path, a consulting engagement ends with a 20 to 50 page PDF that documents every Zap, every scenario, every bot, every variable, every fallback. The PDF is the contract against which maintenance is billed. Clone replaces that artifact with ~/.clone/memory/rituals/, a folder of English-language markdown files on your disk. A ritual like 'every Tuesday 8am invoice the reconciled hours from last week' is one file. If you want to read it, you open it in any text editor. If you want to change it, you edit the sentence. The documentation is the specification is the code.

Three situations. First, you are on regulated infrastructure that requires a vendor canvas with an audit log in a specific compliance envelope (SOC 2 + HIPAA + FedRAMP combined, some financial services). Second, your processes involve human approval chains crossing multiple legal entities where a BPMN-governed canvas is required by your procurement department. Third, you already have 40 UiPath bots in production and the cost of migration exceeds the cost of maintenance. For the other 95 percent of workflow automation questions (solo consultancies, boutique firms, small ops teams), a product that ships the Planner layer eliminates the engagement.

Yes, but through a different mechanism. comparison.tsx line 15 says 'Uses the tools you already have,' including 'custom or legacy apps' (line 29). The Computer Agent drives them through the same UI you use, not through vendor connectors. Your Gmail, your QuickBooks, your HubSpot, your Notion, your Google Sheets, your Folk or Pipedrive CRM, your Zoom transcripts in tl;dv or Fireflies, your Stripe dashboard, your Drive folders. architecture.tsx line 31: 'Gmail, Docs, Sheets, Zoom, CRM, Custom.' The last word matters: custom software without an API is still automatable because the agent reads pixels, not connector catalogs.

Traditional engagement, week one: kickoff meeting, discovery questionnaire, Wednesday workshop, start of process mapping. Deliverable at end of week: zero. Clone, week one: Monday you install the .dmg, Monday afternoon you describe one ritual, Tuesday morning the first ritual fires against real apps, Wednesday you add two more, Friday you edit one that was not quite right. Deliverable at end of week: four rituals running, invoices sent, retros drafted, follow-ups queued. The engagement that was supposed to take eight weeks of meetings to produce a design document produces working automation on day two, at no consultant cost.

Because the pricing is for the product, not for labor. A consultancy bills labor: hours spent on discovery, design, build, and maintenance of a vendor canvas. Clone ships a product where the Planner layer is code (architecture.tsx line 13), not labor. The $49 (comparison.tsx line 179) covers the product license; you do the describing of rituals yourself, in English, with no wiring to do. If you want a human to sit with you on the first three rituals, the booking link at the bottom of this page gets you 30 minutes free. After that, you are on the product, not on a retainer.

Two options. Keep it running for the processes it already handles, and put Clone on the rituals it never got to (typically the long tail of 'small things every week that do not justify a change order'). Or, if the canvas is high-maintenance and selector rot is eating the retainer, redescribe each Zap or scenario as a one-sentence ritual in Clone and retire the vendor workspace. There is no migration step because there is nothing to migrate: the old canvas was a graph, the new ritual is a sentence. architecture.tsx line 56 on principle three: 'Clone adapts in the same conversation. No re-wiring required.'

The Solo plan (comparison.tsx line 179, $49 a month) has a 14 day free trial (llms.txt line 66), no credit card required. Install the .dmg on your Mac, describe three rituals that your current consultant has not got to yet, run them for two weeks against real apps. That is the A/B test. If the rituals run, the maintenance the consultant is billing for has already been commoditized inside the product. If they do not, you have learned something useful for 14 days of cost zero.