Levan Cheishvili

Done-for-you automation for small teams

Bring one manual workflow you want off your plate. I’ll turn it into a reviewed automation workspace.

I help small businesses turn repeated manual steps into simple working automations: reports, follow-ups, data checks, customer updates, and tool-to-tool handoffs.

Availability

Limited availability — currently accepting a small number of early requests while the service is being developed.

Also open to joining an early-stage or product-led team in a hands-on technical delivery role.

I start with one high-value manual process, build a small working version, and then decide with you what is worth automating next.

small first step · fixed scope · human review · full workspace retained

Client workspace preview
Workflow control room Status, approvals, tools, and activity for one request
human review
live workflows3reporting, follow-up, intake
needs approval2reply draft and data export
connected tools5email, docs, CRM, forms, API

Operator review

pending: approve customer reply
checked: no credentials in request

Activity

09:42 form request saved
09:44 AI draft prepared
now waiting for human decision

Levan Cheishvilibuilds and operates the workspace

contextual catalog search

Find relevant examples

Describe a workflow, then choose Find relevant workflows.

what you receive

What you get after the first step

The first step is intentionally narrow.

After the first step, you receive:

  • a plain workflow summary,
  • a boundary list,
  • a first demo plan,
  • a sample-data approach,
  • a list of what not to connect yet,
  • and a recommendation for a small isolated version or future implementation path.

first commercial step

Start with the core of a real system

I begin with one workflow and build a useful foundation you own — not a throw-away experiment. It can grow with your needs after it proves its value.

  • 30–45 minute scoping call.
  • One workflow, chosen for business value.
  • Sample or approved data only at first.
  • Written plan, boundaries, and next step.
  • A clear list of what waits until later.
  • A small working version you can build on.

Your first consultation is free. Before paid work begins, you receive a clear scope and a quote.

The NIS price list is available after secure sign-in when pricing access is enabled for the account. This limited launch is for businesses established or operating in Israel.

No automatic checkout. You pay only after you approve the written scope and quote.

Pay only for the agreed work

Your quote names the work, price, and acceptance terms. One deal gets one customer-specific invoice or payment link — never a generic public checkout.

Do not send card, bank, PayPal login, or payment details through the public form.

grow with control

A useful core first. More only when it earns its place.

Tier Plain meaning Cost control
First working version One clear workflow, safely tested with sample or approved data. No unapproved ongoing infrastructure cost.
Core system Strengthen the working version with monitoring, access controls, and more real work. Scope and monthly cost are agreed in writing.
More capacity Larger or always-on resources only when the proven workflow needs them. Explicit approval before higher recurring cost.

I choose technical tools only after the workflow, value, and cost are clear. You approve any move that adds a recurring cost.

Ready to test one workflow?

Send a high-level description; detailed data comes later only after scope.

Send one workflow for review

open a request

Start with one workflow

Send a short description of the process you want to make easier. I will review it and reply with the safest first boundary, what I would connect, and what I would leave out for now.

Limited launch: requests are accepted only from adults acting for a business established or operating in Israel.

Privacy note: a high-level overview is enough. Please leave out patient data, credentials, confidential files, regulated records, and proprietary datasets.

For new requests, this form is the communication path.

Already opened a request?

Verify your email with a secure link or six-digit PIN to reopen your request.

Verify your email first

Before verification, only your email is collected. Name, company and workflow details come afterwards.

manual work

Manual work I can help remove

The workspace, dashboard, and technical plan stay intact. The starting point is simpler: one repeated business task that should stop depending on copying, chasing, or checking by hand.

Reports prepared automatically from spreadsheets, CRM exports, forms, or approved databases.

Form and message intake routed into the right request, CRM field, task list, or owner queue.

Leads, invoices, and customer replies kept from disappearing in WhatsApp, email, or spreadsheets.

Data checks and cleanup handled before bad rows enter a report, dashboard, or customer record.

A usable dashboard and trail remain so you still see what ran, what needs review, and what should happen next.

what the client sees

Each approved request can become a personal dashboard and infrastructure plan

The visible product is a client workspace: dashboard, connected workflow map, approvals, logs, and an implementation package that can later become Terraform, AWS CDK, or another reviewed deployment spec.

Illustrative AICloudSolo workflow dashboard showing example tasks, review status, logs, and metrics.

Personal client dashboard

A real client dashboard would show live automations, approvals, connected tools, secure notes, activity, and operator review.

Illustrative dashboard example using fictional or sample data. Actual metrics depend on the scoped workflow.

Open dashboard image
Example client workspace architecture showing isolated workflow components, review points, logs, and optional security controls.

Example isolated workspace architecture for approved client work

The generated diagram maps the request path, safety checks, client data stores, guarded AI workspace, human approval console, and client dashboard.

Example architecture only. Specific controls depend on signed scope, hosting model, and approved budget.

Open diagram image
Approved-client flow: after identity, consent, scope, and operator approval, the workspace can generate a personal dashboard, a workflow diagram, and a reviewed deployment package such as Terraform, AWS CDK, or a GCP deployment spec.

how it works

The work moves through a simple reviewed path

A request enters the workspace, AI prepares the next useful step, the operator approves what matters, and the result remains visible to the customer.

inbox Request inbox

New messages, consent, contact details, and session status stay attached to the client dashboard.

prepare AI prepares work

AI drafts, classifies, summarizes, checks context, or prepares a tool action inside the approved workspace.

approval Approval queue

AI drafts, exports, and sensitive actions wait for a human decision before they go out.

dashboard Client dashboard updates

Decisions, tool events, replies, and operating notes remain visible in the same client cloud.

daily_report.log

Work log — enough detail to see what ran and where it ended.

intake_trace.txt

Connector trace — tool updates are retried, checked, and kept narrow.

runbook_report.md

Runbook fragment — the plain notes needed to operate or review the setup.

Representative examples only. Client names, credentials, records, and proprietary data are not shown.

lightweight demo

A small real workflow can run in its own cloud lane

For an approved demo, I can run a simple sample workflow on serverless cloud infrastructure and show the result in the client workspace. The visitor sees the run, status, log, and result; the underlying workflow engine stays behind the scenes.

request

Choose a harmless sample workflowFor example: validate a small CSV, prepare a tiny report, or check a folder inventory using sample data.

run

Run it in serverless cloudThe job starts in a separated cloud lane with bounded inputs and no access to private customer systems.

review

Review the resultThe operator checks the log and output before it becomes part of the visible workspace.

show

Show it on the dashboardThe client sees status, time, result, and the next suggested step without needing to understand the infrastructure.

Demo runcompleted

The demo uses sample or explicitly approved data only. It is meant to make the service concrete: a small workflow runs, leaves evidence, and appears in the same customer workspace.

For technical teams For larger workflows, the same approach can be expanded into a technical plan with stages, review points, logs, and deployment choices.

eligible users · pipeline planning

Describe a data flow, receive a high-level pipeline map

For selected eligible users, the workspace can turn a plain-language data processing flow into a high-level pipeline map: inputs, checks, intermediate steps, fan-out work, reduce/merge points, controlled feedback paths, outputs, and review gates.

describe

Tell the flow in normal words

What arrives, where it comes from, what should happen to it, what output you expect, and what must be reviewed.

diagram

Use a directed pipeline map

The best fit is a high-level directed pipeline map, optionally grouped into lanes for input, processing, review, output, and operation.

patterns

Show map-reduce and cycles clearly

Parallel work is shown as fan-out branches, reduce is shown as a merge step, and repeated processing is shown as a controlled cycle with an exit condition.

cost control

Generate only when it is justified

Public examples stay deterministic and cheap. Real diagrams are queued, cached, limited per customer, and generated only for eligible requests.

approval

Make live only after review

The user can request “make this live,” but implementation waits for operator review, chat confirmation, ETA, and instructions.

High-level pipeline mapDirected flow with review gates
High-level data pipeline map Input moves through validation, fan-out processing, merge, review, cycle, dashboard, and report output. input map cycle output file / form / API validate split work parallel step A parallel step B merge review gate new attempt controlled cycle dashboard report / export
Flow map: shows the business data path, branching work, merge points, review gates, and revised attempts.

Example: a shop uploads a supplier CSV, the workspace validates it, splits checks by product group, merges exceptions, asks for review, then updates a dashboard and export.

The diagram is high-level first. It becomes an implementation plan only after operator review, user confirmation, and agreement on data, access, runtime, cost, and responsibility.

From approved diagram to build package planning first · implementation after review
approveAgree on the diagram

The user sees the proposed flow, asks for corrections, and the operator confirms that the scope is realistic.

briefGenerate a production brief

The approved diagram becomes a precise build prompt with inputs, outputs, runtime, permissions, checks, costs, and open questions.

planPlan before building

The next step can stay in planning mode until data access, payment, responsibility, and acceptance criteria are clear.

buildCreate the deployment package

After approval, the build package can include infrastructure-as-code, a workflow definition, run notes, and a simple dashboard.

ownStop at a checkpoint

At any stage the user may stop, pay for the completed checkpoint, and keep the useful intermediate artifact.

No production action is automatic from the public form. The generated build prompt is used only after user confirmation and operator approval.

behind the scenes

What gets set up so the workflow can run

Behind the simple dashboard is the technical work: where the workflow runs, how messages arrive, which tools it may touch, what gets logged, what retries, and what waits for approval.

workspace

A bounded runtime for this request

The conversation, status, contact details, scope notes, and next actions stay tied to the same customer request.

tool links

The few connections that matter

I connect only the tools needed for the workflow, using controlled APIs, queues, or handoff points.

AI help

AI where it is useful

The assistant can draft, classify, summarize, or prepare work, but it does not quietly take over the customer relationship.

operator view

A view for decisions

Messages, approvals, retries, and exceptions are visible so the work can be handled without digging through inboxes.

work trail

A trail of what happened

Important events are logged so the path from request to reply or automation run can be reviewed.

next step

A practical handoff

You get the working setup, the limits, the operating notes, and a clear next step.

what you keep

The result should be useful after the call ends

The goal is not a slide deck. You should leave with a small working environment, the important decisions written down, and a clear idea of what can safely happen next.

first setup

A private working area

A small cloud workspace for this request, with the conversation, status, notes, and customer channel in one place.

  • Scope for this customer request
  • Persistent request channel
  • Access and dependency notes
  • Safe next-step recommendation
  • Visible status
  • Clear boundary statement
connected tools

A workflow that can run

The needed tools are connected into a monitored path with retries, logs, and review points.

  • Working automation path
  • Tool connector or data flow
  • Monitoring and logs
  • Human approval point
  • Recovery or fallback note
operating notes

Notes someone can actually use

Enough detail to understand what runs, who approves it, where the limits are, and what should be improved next.

  • Runbook or operating notes
  • Log and status evidence
  • Known limits and access model
  • Next improvement path

On scope: the first paid step is deliberately small. It defines the boundary, tool access, approval points, and useful path before any sensitive integration is built.

controls

Useful automation needs limits

Important steps stay under human review. The first version uses sample or approved data and avoids broad access. For real integrations, work can be set up in your cloud account or in a separated account where you keep ownership and grant limited delegated access.

Where the work is allowed to happen

input

  • One business process
  • High-level context only
  • Consent and contact channel

control

  • Customer-specific request ID
  • Separated session and records
  • Access requirements documented
  • No credentials in public form

output

  • Clear workspace boundary
  • Safer integration path
  • Traceable follow-up

Where AI is allowed to help

agent role

  • Draft, summarize, classify, prepare
  • Stay inside the workspace context

control

  • Guardrails before agent use
  • Human approval for customer replies
  • Logs for important events
  • Fallback to manual operation

result

  • Controlled automation
  • Lower operational ambiguity
  • Human-operated service channel

Formal compliance claims are made only when they are part of a signed scope. The default promise is practical isolation, narrow access, and a reviewable work trail.

private workspace · by approval

Deeper access comes later, after the shape is clear

Prototype agents, connectors, and data paths are tested with sample or approved data first. Production access, secrets, and sensitive files stay outside the public request flow.

boundaries

What the public form is not for

Sensitive and regulated data

Use the form for high-level summaries only. Work involving patient data, credentials, confidential files, regulated records, or proprietary datasets requires a signed scope of work and appropriate access, data handling, and confidentiality terms.

Access model

Tool access, production responsibility, monitoring, and rollback expectations are defined per workspace. No broad access is implied by the public form.

AI systems

AI-assisted steps are designed with guardrails, human review, logging, and clear data boundaries. The goal is controlled automation with appropriate oversight.

Formal compliance

AICloudSolo can work carefully with sensitive operations, but formal certification or regulated hosting is only promised when explicitly scoped and contracted.

how a request starts

Start small, then build only what is justified

intake

Tell me one workflow

Describe the process at a high level and confirm the communication/session terms.

boundary

Decide what belongs inside

I clarify the tools, access model, approval points, and practical risks before connecting anything sensitive.

build

Build the narrow path

Connectors, AI assistance, logs, and operator controls are added one safe path at a time.

operate

Use it and improve it

You receive the working path, notes, limits, and the next improvement option.

feedback

Help make the site clearer

Use this form after reading the page, opening a request, or seeing a demo. It is separate from the request form and has no incentive attached.

Refer a business owner

Know someone with a workflow they want to make easier? Send them your referral link or code yourself. Do not enter someone else’s email or private details here.

Later, after a referred business becomes a paid client, both sides may receive ₪250 credit toward a Workflow Review or Demo Plan. Credit is manual, not cash, one per new client, and only after paid conversion.