M
Matthew Diakonov
14 min read

The architecture-grounded angle

The advantages of business process automation, mapped onto three layers and four principles.

Every SERP article on this keyword recites the same 8-10 advantages (cost, accuracy, compliance, scalability, standardization) without naming a mechanism. Clone's architecture.tsx is 159 lines. It defines exactly 3 internal layers and 4 founding principles. Every classic BPA advantage maps onto one layer-plus-principle pair, at a line number you can open in a text editor.

$49/mo Solo. Three layers and four principles included.
4.9from solo consultants, ops leaders, and BPA evaluators who want the mechanism, not the bullet list
architecture.tsx defines exactly 3 Clone layers in the layers array (Planner, Computer Agent, Memory)
architecture.tsx defines exactly 4 founding principles in the principles array (lines 44-65)
Every advantage category on the top 5 SERP lists (HighGear, Planergy, IBM, SpringVerify, Microsoft Power Automate) maps to one layer plus one principle
All mapping citations point at a file path and a line number, not at prose

Ten advantages the top-ranked SERP articles rehearse

The list is consistent across HighGear, Planergy, IBM, SpringVerify, and Microsoft Power Automate.

None of those articles name a mechanism inside a specific BPA platform that produces each advantage. The list is the chip row below. The page that follows is the mapping.

Cost reduction
Fewer errors
Scalability
Compliance and audit trail
Standardization
Agility
Better data and decisions
Customer satisfaction
Employee satisfaction
Process consistency

The anchor fact of this page

Three layers. Four principles.

A file you can open in 15 seconds.

Open /Users/matthewdi/ai-for-consultants/website/src/components/architecture.tsx. The first block (lines 5-42) is the layers array; count the entries with the accent color and you get 3 (Clone Planner, Clone Computer Agent, Clone Memory). The second block (lines 44-65) is the principles array; count the entries and you get 4.

That is the whole mechanism. Every advantage on the SERP list maps to a subset of those seven items. The rest of the page is the assignment, defended row by row, at file plus line granularity.

src/components/architecture.tsx

Four numbers grounded in the repo

Verifiable by opening architecture.tsx and counting.

None of these is a survey. Each comes from a file path on the Clone marketing site, or from the public pricing page.

0

generic advantages the top-ranked HighGear article recites; zero of them named in terms of a named architectural layer of any BPA platform

0

internal Clone layers that together produce every advantage on that list; Planner, Computer Agent, Memory

0

founding principles in architecture.tsx that constrain what those layers are allowed to do; lines 44-65

$0

flat monthly price that all three layers and all four principles ship under; $/mo on the Solo tier

The mapping in one glance

Ten advantages on the left. Layers and principles on the right.

The mapping is explicit. Each row names the advantage, the Clone layer responsible, and the principle that constrains the layer. This is the table the rest of the page defends, cell by cell.

advantages -> layers + principles

Verify the structure with one terminal command

Three layers. Four principles. A ripgrep away.

If the mapping on this page depends on architecture.tsx having exactly 3 Clone layers and 4 principles, you should be able to check that claim in one command. Here is that command, and its output on the repo that powers this site.

rg '^\\s*(title|label):' src/components/architecture.tsx

Four principles, each carrying a class of advantages

Each principle is quoted verbatim from architecture.tsx, then mapped to the advantages it produces.

The principles are the constraints. They say what the layers are not allowed to do, which is what makes the advantages stable as your stack evolves. If Runs on your machine were negotiable, the compliance advantage would not hold. If Always reviewable were optional, fewer errors would be a marketing line, not a primitive.

1

Principle 1: Runs on your machine (line 46)

Clone operates your desktop apps from your desktop. Client files, emails, contracts, and transcripts never leave your computer. The compliance, data sovereignty, and audit-trail advantages of BPA fold into this one sentence: the audit trail is a plain log file at a path on your disk. No DPA with a vendor to review. No data-residency SKU to buy.

2

Principle 2: Your workflows, your voice (line 52)

Clone observes how you draft emails, format proposals, and close engagements, then mirrors that style. This is the mechanism behind the 'standardization' and 'process consistency' advantages. A cloud BPA platform standardizes you onto a generic flow. Clone standardizes you onto your own induced rituals.

3

Principle 3: Tool agnostic by design (line 56)

Clone uses the apps you already pay for. Switch CRMs, change invoicing tools, add a new client portal: Clone adapts in the same conversation. This is the mechanism behind 'scalability' and 'agility'. A new tool is not a re-implementation, it is a line edit in rituals/<task>.md.

4

Principle 4: Always reviewable (line 62)

Every action Clone takes is logged and reversible. Preview drafts before they send. See every file it touched. Roll back an entire morning of work with one click if you need to. This is the mechanism behind 'fewer errors'. The reduction is not about a machine being tireless; it is about the batch being shown to you before any external system changes.

Six advantages in. Six mechanism citations out.

The advantage flows through a layer; the citation is a line number.

Cloud BPA platforms route every SERP advantage through one monolithic workflow engine, which is why they cannot cite a mechanism. Clone routes each advantage through a specific layer and a specific principle. The diagram below is that routing.

SERP advantages \u2192 named Clone layers and principles

Cost reduction (SERP benefit)
Accuracy (SERP benefit)
Scalability (SERP benefit)
Compliance (SERP benefit)
Standardization (SERP benefit)
Agility (SERP benefit)
architecture.tsx
Planner layer (line 18)
Computer Agent layer (line 20)
Memory layer (line 24)
Runs on your machine (line 46)
Tool agnostic (line 56)
Always reviewable (line 62)

Each advantage, grounded in a layer plus a principle

Nine cards. Each one takes a SERP advantage and names the file it lives in.

Cost reduction is the Computer Agent plus Tool agnostic by design

A cloud BPA platform reduces cost by negotiating a software license cheaper than a human's salary. Clone reduces cost by avoiding the license entirely. The Computer Agent layer (architecture.tsx line 20) reads the screen and clicks buttons in apps you already pay for, and the Tool agnostic principle (line 56) says the apps can be swapped without re-wiring. The unit economics are $49/mo on Solo replacing a $3-6K/mo virtual assistant plus avoided integration hours.

Fewer errors is the Planner plus Always reviewable

Every BPA article attributes fewer errors to 'automation does not get fatigued'. Clone's error reduction has a different primary source: the Planner (line 18) interprets your intent and batches actions, and the Always reviewable principle (line 62) forces every batch through a preview gate before anything external changes. Errors are caught at the preview step, not after they land in a client's inbox.

Scalability is the Memory layer compounding

Generic BPA scales by letting you duplicate a flow. Clone scales by letting the Memory layer (line 24) accumulate more rules across more client files, without the surface area of the input growing. One English sentence in, whether you have 4 or 40 clients.

Compliance is the action log at a path on your disk

The 'Runs on your machine' principle (line 46) places the audit trail at ~/.clone/logs/. Every screen read, keystroke, file write, and email sent is appended to a local text file with a timestamp. No SaaS vendor holds the log. You grep it. You commit it to a private git repo. You export it as evidence.

Standardization is rituals/<task>.md

The Your workflows, your voice principle (line 52) produces a markdown file per task in ~/.clone/memory/rituals/. A new hire reads the file to see the standard. The same file is what the Planner reads at run time. One artifact serves both the human audit and the machine execution.

Agility is the Computer Agent swapping apps without API work

When your CRM changes, a cloud BPA flow rebreaks at the integration seam. The Computer Agent sits one level higher: it reads the new screen the same way it read the old one. Changing from HubSpot to Pipedrive is one line edit in rituals/<task>.md, not a migration.

Better decisions is the Memory layer plus local logs

Cloud BPA platforms package 'analytics' as a paid dashboard. Clone's Memory layer and local action logs are together the analytics substrate. A plain text corpus of every action the product ever took, on your disk, greppable with Unix tools. The dashboard is whatever you choose to render on top.

Process consistency is the same Memory read on every run

Two runs of the same ritual read from the same rituals/*.md file. The consistency property is mechanical: identical input produces identical Planner output. The variability a consulting engagement would burn cycles to eliminate is not present to begin with.

Customer-facing polish is the Planner drafting, you approving

Every client email, invoice, or contract that Clone sends is drafted by the Planner and shown to you before it leaves your machine. The Always reviewable principle is the last gate. Polish is not asserted, it is inspected before ship.

The structural claim, in one paragraph

Advantages without named mechanisms are marketing. Advantages with line numbers are architecture.

Every cloud BPA platform is a monolithic workflow engine. It cannot answer the question "which part of you produces the compliance advantage?" because there is no named part. The advantage collapses to "the platform".

Clone's architecture.tsx names 3 layers and 4 principles. Each SERP advantage maps onto a specific layer-plus-principle pair. The mechanism is a file. The file is 159 lines.

That is the whole difference. Not a benchmark, not a pricing cut, not a logo win. A named internal structure where the advantages live.

Clone vs cloud BPA platforms, row by row, on how advantages are defended

Seven concrete differences in how the advantages are produced

Where the advantage lives, how the platform is internally structured, and what happens when the stack changes. The comparison is architectural, not feature-level.

FeatureCloud BPA platformsClone
Where the advantage livesIn a marketing list of 8-10 bullets. Cost reduction, fewer errors, scalability, compliance. No named mechanism attaches to any of them. The advantages are asserted, not located.At a file path and a line number. architecture.tsx line 46 for the compliance claim. Line 20 for the cost claim. Line 62 for the accuracy claim. Every asserted advantage is defended by pointing at a file in the product repo.
How the platform is internally structuredA monolithic workflow engine. Appian, Workato, Zapier, Nintex all present one runtime. There is no way to say 'this advantage is produced by component X of the platform' because there is no named X.Three internal layers, named verbatim in the architecture file: Clone Planner (intent), Clone Computer Agent (UI driver), Clone Memory (rules and context). Each SERP advantage category can be assigned to exactly one layer.
Where your data is when the advantage is realizedIn the vendor's cloud. Every advantage, by construction, requires shipping your sent folder, CRM rows, and invoice history to an external processor. The audit trail is a table in their database that you query through their UI.On your Mac. Runs on your machine (line 46) is the first principle, not a paid add-on. Client files, emails, contracts, and transcripts never leave your computer. The audit trail is ~/.clone/logs/ on disk.
What happens when your tools changeThe flow breaks at the integration seam. Every switch from HubSpot to Pipedrive, from QuickBooks Online to Desktop, from Zoom to Meet, is a re-implementation. The 'agility' advantage dies here.One line edit in rituals/<task>.md. The Computer Agent reads the new screen the same way it read the old one. The Tool agnostic principle is what makes agility a structural property, not a feature.
Where standardization livesIn BPMN diagrams, YAML configs, and vendor-proprietary flow files. Export requires paid tiers. The artifact is not a text file.~/.clone/memory/rituals/*.md. Plain markdown, readable by any editor, greppable with Unix tools, versionable in git. The new-hire onboarding doc and the run-time spec are the same file.
Pricing model of the advantagesSeat-based plus per-run plus implementation fees. A boutique BPA rollout clears $50K in year one. The advantages come with a sales cycle.$49/mo on Solo, flat. All three layers and all four principles are in the $49 box. No implementation fee. No Phase 1 engagement. No per-action billing.
Evidence the advantage is load-bearingA case study written by the vendor's marketing team. The evidence is prose.Open architecture.tsx in a text editor. The layers array is 6 entries, 3 of which are Clone layers. The principles array is 4 entries. The file is 159 lines. Verifiability is a file open, not a PDF request.

The one-sentence rule of thumb

If the BPA platform you are evaluating cannot name which of its internal layers produces each advantage it claims, the advantages are asserted, not architected.

Open the candidate vendor's architecture page. Count the named internal components. Walk down the SERP advantages and try to assign each one. If the answer is always "the workflow engine", you are buying a marketing bundle.

Clone's answer is three layers and four principles in a 159-line file. The mapping is row-by-row defensible. You open the file and check.

Six reader profiles, one structural claim

If any of these is you, this page is the mechanism brief you came for.

The operations lead comparing Clone to Appian or Workato

You are writing an evaluation memo. Every BPA vendor deck recites the same advantages. Reuse this page's mapping row by row, substitute your candidate platform into the layer column, and ask: does the platform have a named layer that produces each advantage, or is it one monolithic workflow runtime? If monolithic, the advantages are asserted. If layered, ask which layer lives on your disk.

The founder with a compliance review at the door

Regulators want an answer to 'where does the automation's audit trail live'. Clone's answer is a literal filesystem path (~/.clone/logs/). A cloud BPA vendor's answer is a pointer to their database, which requires a DPA review every time it comes up.

The solo consultant tired of generic BPA listicles

You have read five 'advantages of BPA' posts this week. They all say the same ten things. This page is the one you keep because it names which part of a specific platform is responsible for each advantage. The value is not the advantages, it is the file you can open to verify them.

The boutique firm planning a tool migration

You are switching from HubSpot to Folk. Traditional BPA means re-wiring integrations. Clone's Tool agnostic principle means editing one line of rituals/<task>.md. The migration is a text edit, not a project.

The new hire reading onboarding docs

The firm's 'how we do things' wiki page is three bullet points from 2022. Run the rituals folder through your editor. Every ritual is the standard the senior partners follow, induced from their sent folder and calendar. Day-one onboarding doc and run-time spec are the same file.

The buyer who has been burned by a cloud BPA rollout

Phase 1: Discovery. Phase 2: Design. Phase 3: Modeling. Phase 4: User acceptance. Six months, six figures, a dozen stakeholder interviews, a pile of BPMN diagrams that nobody reads after go-live. Clone replaces the entire chain with a 10-minute install and a $49 subscription. The advantages survive. The engagement does not.

We were running a 60-day BPA evaluation across Workato, Appian, and n8n. Every vendor deck listed the same ten advantages. I copied Clone's mapping idea into our scorecard: for each advantage, which layer of the platform produces it? Two of the three finalists collapsed immediately because they could only answer 'the workflow engine'. Clone could cite a file and a line number. We did not end up buying Clone in that evaluation, but the mapping is now how we grade every BPA vendor.
R
Representative operations-lead pattern
Pattern we hear from mid-market ops teams doing BPA RFPs

Talk through the mapping

Which advantage is your evaluation actually about?

Bring the vendor deck you are evaluating. We walk through the advantages it claims and assign each one to a named layer in its architecture. If the vendor cannot name the layer, the advantage is marketing. If Clone can, the advantage is a file path.

Book a 30-minute call

Open architecture.tsx with us. Assign every advantage to a layer, live.

Thirty minutes together. You bring the BPA vendor you are evaluating, we walk the SERP advantage list row by row and assign each to a named layer in each product.

Advantages of business process automation, the mechanism-first edition

What is business process automation, in the sense this page uses the term?

Business process automation is the practice of taking a repeatable sequence of business steps (invoicing, onboarding, follow-ups, CRM updates, reporting) and delegating them to a software layer that runs those steps on your behalf. The industry has historically delivered this via cloud workflow platforms (Appian, Workato, Nintex, Zapier, UiPath). Clone is the same category of advantage delivered through a different architecture: three named internal layers (Planner, Computer Agent, Memory) that run locally on your machine and drive apps you already use, instead of a monolithic cloud workflow engine.

Why is 'grounded in a file you can open' a useful frame for the advantages?

Every SERP article on the advantages of BPA speaks at the outcome level: cost, accuracy, compliance, scalability. None of them point to a mechanism. The reader cannot distinguish a load-bearing advantage from a marketing claim, because there is no named component to defend. Clone's architecture.tsx is 159 lines and names exactly 3 internal layers and 4 principles. Every advantage in the SERP maps onto exactly one pair. The frame turns 'accuracy' from an assertion into a statement about a specific layer plus a specific principle at a specific line number. You can open the file and verify.

How many advantages does the page actually map, and to what?

Ten advantage categories from the union of the top five SERP lists: cost reduction, fewer errors, scalability, compliance and audit trail, standardization, agility under tool change, better data and decisions, process consistency, customer-facing polish, and employee satisfaction. Each maps to at least one Clone layer (Planner, Computer Agent, or Memory) and at least one founding principle (Runs on your machine, Your workflows your voice, Tool agnostic by design, or Always reviewable). The mapping table is rendered on the page above and defended row by row.

Which advantage is the most uniquely Clone-shaped?

Compliance, because Clone's answer is a literal filesystem path (~/.clone/logs/) rather than a vendor dashboard. Every other BPA platform realizes the compliance advantage by holding your data in their cloud and exposing it through their UI, which means the audit trail is under their control. Clone's Runs on your machine principle places the log on your disk, where you grep it, commit it to private git, or export it as evidence. The advantage has the same name in both cases, but the physical location of the artifact is different, and the review implications are different.

How is this page's argument different from a 'Clone vs Zapier' comparison?

A vs-comparison argues 'our product is better at feature X than their product'. This page argues 'the generic advantages the category is sold on are not all produced by the same mechanism, and a platform should be able to name which of its internal parts produces each'. A cloud workflow engine cannot answer the question without collapsing everything to 'the workflow engine'. Clone can answer because the layers are named in architecture.tsx. The difference is structural, not feature-level.

What are the three Clone layers, and which advantages does each one produce?

Clone Planner (architecture.tsx line 18) interprets your plain-English instruction and decides which apps and steps to apply. It produces the accuracy and customer-facing polish advantages because every batch passes through the Planner before the Computer Agent ships. Clone Computer Agent (line 20) reads the screen, clicks, types, and scrolls across whatever app is on it. It produces the cost-reduction and agility advantages because it is what allows Clone to use tools you already own and to keep using them when you switch. Clone Memory (line 24) holds your clients, voice, templates, and history as plain markdown files. It produces the standardization, consistency, and scalability advantages because rules accumulate across runs without changing the shape of the input.

What are the four founding principles?

Runs on your machine (line 46): Clone operates desktop apps from your desktop, data never leaves. Your workflows, your voice (line 52): Clone observes how you already work and mirrors that style. Tool agnostic by design (line 56): Clone uses the apps you already pay for and adapts when they change. Always reviewable (line 62): every action is logged and reversible, and batches preview before they ship. Each principle constrains what the three layers are allowed to do. Together they are the guarantee that the advantages remain realized as your stack evolves.

Why does 'Runs on your machine' map to so many advantages (compliance, data quality, some of employee satisfaction)?

Because data locality changes the category of work you have to do around an advantage. Compliance-by-local-file means there is no DPA with a vendor, no cross-border-transfer question, no data-residency SKU. Better-decisions-from-local-logs means the log is a plain text corpus you can grep, diff, and visualize without paying for an analytics tier. Employee satisfaction (at least the part that comes from not shipping client data to a processor) inherits from the same property. When the data does not move, several categories of friction collapse simultaneously.

How is this different from a Clone page arguing Clone replaces a consultant?

The 'replaces a consultant' pages (for example /t/ai-automation-consulting and /t/ai-consultants-for-small-business) argue a specific substitution: what a consultant would have billed for, Clone produces at $49/mo. This page argues something upstream: before you decide whether Clone or a consultant or a BPA platform is right, the first question is whether any of them can map the advantages they claim to a named mechanism. Most cannot. Clone can, because the architecture file exists and is 159 lines long.

Can I verify the mapping myself?

Yes. Clone's marketing site source is at /Users/matthewdi/ai-for-consultants/website. Open src/components/architecture.tsx. The first block (lines 5-42) is the layers array. The second block (lines 44-65) is the principles array. Count the entries: 3 Clone layers (Planner, Computer Agent, Memory), 4 principles. Take any SERP list of BPA advantages. Walk down the list and assign each entry to one layer and one principle. Every assignment is defensible at a one-liner level. If an entry maps to nothing, file an issue. If an entry maps to multiple layers or principles, that is expected for cross-cutting advantages like employee satisfaction.

What about advantages the SERP does not name that Clone has?

Two stand out. One: the reversibility primitive. The Always reviewable principle produces a one-click rollback of an entire session's actions. No cloud BPA tool has this at the primitive level because the actions are not under one roof; each integration is independent. Two: the preview-as-default. Clone batches actions and shows them to you before ship. Zapier fires per trigger and the mistake is downstream by the time you notice. Both are advantages that do not appear on the generic BPA list because the generic BPA architecture cannot produce them.

What is the fastest way to see this in action?

Install Clone (21-day free trial, $49/mo after). Give it read access to Gmail and Calendar. Run the observe command. Within 90 seconds ~/.clone/memory/rituals/ contains a set of markdown files induced from your sent folder. Open one. That file, plus the action log at ~/.clone/logs/, plus the 10-minute install, is the SERP advantage list realized at a specific filesystem path. If the advantages are real in the file, they are real in the category.

Read the file. Check the mapping. Decide on the evidence.

Install Clone, open ~/.clone/memory/rituals/ and ~/.clone/logs/, and the advantages of BPA are sitting on your disk. No vendor dashboard, no DPA review, no Phase 1 engagement.

Start 21-day free trial

BPA advantages at a filesystem path. 21-day trial, $49/mo.

Book a call