Illustration: an inquiry from Dana Levi arrives from a website form, a CRM record is created and routed to Yoav, a meeting is scheduled for Thursday 17:30, and Yoav receives the context before he calls.
Website form09:14
Dana Levi · 4 rooms, Ramat Gan
4-room apartment, Ramat Gan
CRM recordNewMeeting set
Dana Levi
SourceWebsite form
OwnerYoav
MeetingThu 17:30
CalendarThursday
Free17:30
Viewing with Dana Levi17:30
Follow-up scheduled
Context ready09:16
Yoav
Inquiry4 rooms, Ramat Gan
Past contactNone
MeetingThu 17:30
Motion reduced by your system settingsIllustrative demonstration with fictional dataFictional data
The information exists. It just doesn't move.
One inquiry, written in six places, becomes one route that ends with a person.
Illustration: the same inquiry is written in six places — form, CRM, spreadsheet, WhatsApp, calendar and a document — then gathers onto one route that ends with Yoav and the context ready.
Website form09:14
CRM record09:14
Excel sheet09:15
WhatsApp on a phone09:15
Calendar09:16
Notes in a document09:16
Yoav
Context ready · 09:16
Copying between toolsForm, CRM, spreadsheet, calendar. Typed four times.
Follow-up from memorySome hear back today. Others fall through the cracks.
Context spread everywhereCalls here, WhatsApp there, notes in a document.
Admin that repeats itselfReminders, status updates and reports, by hand every week.
Illustration: a finished run arrives as raw Garmin data, code scores it (scoring, de-duplication, training load, patterns), the AI explains it in plain language, and Maya the runner decides what to do with the advice.
Rutz · Post-run analysisRun receivedExplained
Run receivedRaw Garmin data
ScoredCode: same run, same numbers
ExplainedAI: analysis in plain language
Maya decidesWhat to do with the advice
Code computeddeterministic
Run8.2 km · 42:10
Score7.4 / 10
Training load68 · +12%
DuplicateNone
Pattern3rd hard day in a row
AI explained06:58
Three hard days in a row and your weekly load is up 12%. Tomorrow, keep it easy: 30 minutes at a conversational pace.
A running coach built with a partner. Code does the math; AI explains it.
Illustration: an exercise submissions queue — pull requests come in, tests and drafted feedback go out, and a submission whose test timed out is flagged for the lecturer, who decides the grade.
Exercise 3 · Submissions today · 14 in the queue
Submission (pull request)Tests
#3141 · 09:1220/20
#2718 · 09:1518/20
#1618 · 09:2119/20
#4669 · 09:26Running tests19/20
#5772 · 09:31Running testsNeeds review
#5772 · Needs review09:31
EnvironmentUniform Docker runner
Tests17/20 · one timed out
pylint2 notes, with learning links
Feedback draftedTo the lecturer
Submission
Tests
Feedback drafted
Lecturer
Anonymised ids. Automation checks and drafts; the lecturer decides the grade.Fictional data
Pull requests in, tests and drafted feedback out, the lecturer decides the grade.
Illustration: a small job is posted, a verified helper takes it, payment is secured before work starts and released on completion, an attempt to share a phone number outside the app is hidden, and the job owner leaves a fair review.
QuestIL · Small jobPayment securedReleased · job done
Job postedNoa asks for help
IdentityA verified, accountable helper
PaymentSecured before, released after
ReputationNoa leaves a fair review
The jobThu 17:00
Carry a sofa to the 3rd floor
WhereRamat Gan
Price₪180
Payment secured · released on completionReleased to the helper
The helperIdentity verified
Amit Cohen
Jobs completed32
Rating4.9 · verified reviews
Fair review · Noa decides
Amit: call me on 05x-••• (hidden)Outside number hidden · stays in the app
Four guarantees: identity, payment, containment, reputation.Fictional data
Roll-out in stages: today's flow next to the new one.
Today
New flow
In stages
Monitoring: a tracking row with errors and the last run, and the person who reports what changed.
Errors this week0
Last run07:30
You tell me what changed
Fictional data
Frequently asked questions
Do I have to replace the tools I already use?
Usually not. Most of my work connects what you already have: CRM, spreadsheets, calendar, phone system, WhatsApp. Replacing a tool happens only when it genuinely blocks the process, and then in stages, with the old system running alongside.
Automation platform or custom development?
Both. Platforms like Make are quick to set up and easy to inspect. Custom code comes in when a flow needs logic a platform can’t express, or the team needs an interface. I recommend the simplest option that holds up.
When is AI actually useful, and when isn’t it?
AI helps when the input is messy: a voice message, a free-text inquiry, a thread to summarise. It doesn’t help where a clear rule exists; plain automation is cheaper and more predictable there. Decisions with weight stay with a person.
When does a WhatsApp flow hand over to a person?
Whenever your rules say so: the customer asks for a person, the question is outside the flow, the tone suggests frustration, or a commercial decision is needed. The handoff includes the conversation and a summary, so nobody starts from zero.
How are scope and maintenance agreed?
Scope is written down before building. Pricing is per project, by inquiry. Rollout is staged, with a walkthrough for your team; accounts and data stay in your name; and we agree in advance what I monitor after launch.
Tell me where the work gets stuck.
Maor Kertsman · Independent developer, Israel
Hi, I'm Maor, an independent developer in Israel. You work directly with me, from the first conversation to the last deployment.