Skip to content
kancir AI

How it works

Software we run for you

kancir AI is software we run for you. You never log into it. It reads your leads, writes the follow-ups in your voice, and sends nothing until it is approved against wording you signed off on.

Step by step

What happens to a lead

We built the software and we are the ones who run it. Here is what it does with a lead, from the file it reads to the day it stops.

  1. It reads every lead

    Your inquiries, form fills and missed calls keep landing exactly where they land now, in the system you already pay for or the file it exports. Our software reads that file straight through, builds one record per person, and merges the duplicates, so the customer who called twice is one person and not two. Then it sorts everybody by how long it has been quiet: still in conversation, cooling off, or gone months ago. Anyone still mid conversation with you is your live business, and the software leaves them alone.

    Under the hood

    It cleans the file before it trusts it. Four ways of writing a phone number become one number. A name filed as Last, First gets turned back around. The same person entered twice becomes one person. Any column that reads as medical is dropped at the door, so it never loads at all.

  2. It writes in your voice

    The wording is yours. We sit down with you once and write the messages the way you actually talk to your customers, and you sign off on every one of them before anything runs. From there the software does the writing. It picks who is due, chooses the right message for where that person stands with you, fills in their first name and your business name, and hands back a finished sentence instead of a blank box. No sentence is being invented at the moment of sending.

    Under the hood

    Every message is checked three times before it can leave: when your wording file loads, when the names are filled in, and once more at the door on the way out. A first message may never carry a price, a date, a service, a second person's name, or anything medical, and it has to carry a one word way out. Wording that breaks a rule is refused by the software, and the software says which rule.

  3. Every message gets approved

    One card at a time on a phone: who it is going to, the exact words, and three buttons. Approve it, rewrite it first, or throw it out. Every card is read against the wording you signed off on, and approving does not send anything by itself. The send is a separate step and it only moves what was already approved.

    Under the hood

    There is no approve all button anywhere in the software. Not a screen, not a setting, not a flag in a file, and a config file is not even allowed to contain a key that mentions approval. Ten messages is ten taps, on purpose, because the day you can wave a batch through is the day something goes out under your name that nobody read. A rewrite goes back through the same rules and still waits for a tap.

  4. It sends, and it keeps going

    Approved messages go out under your name, from your number or your inbox, one at a time, with an uneven pause between them so a batch never leaves as one burst. They go only inside the hours you set, measured on your own clock, and never more than the daily count you agreed to. It keeps going after one message. We read the replies at least once a day, and an answer closes that person's conversation for good. No later wave picks them back up.

    Under the hood

    It keeps a record of every lead, every word sent, every reply, and every time it chose not to send. Held outside your hours, held by the daily count, passed over because somebody opted out, passed over because a message to them was already in flight: all of it is written down. Held is not lost. Those messages go out on the next run inside your window. That record is also why the numbers you get back are counted rather than estimated.

  5. It stops for good

    Anyone who asks to stop is blocked, permanently, and nothing we run reaches them again. There is no undo, on purpose. Everything else that comes back is handed to a person at your business to answer in their own words, in your real inbox or your real texting app.

    Under the hood

    Our software has no way to write a reply. That is not a policy we follow, it is a path that does not exist in the code, and a test goes red if anyone ever adds one. Replies come in on a channel that can only read.

Runs beside what you already have

We work from an export your system already makes. Nothing gets installed, nothing moves, and your front desk keeps working the way it already works. No migration, and no new dashboard for your team.

Every one of these works from an export you run. Open any of them to see what we ask for and what we leave alone.

  • GoHighLevel

    via export

    You keep the sub-account you already run. Nothing moves.

    Most agencies start with an export. Once a week you pull your contacts out of Smart Lists, which takes about two minutes. We read who went quiet, draft the follow-up in your voice, and hold every message until it is approved against wording you signed off on. Approved messages go out from your own email address, under your name.

    If you would rather we read the account directly, GoHighLevel has a token you create yourself in agency settings. It needs nobody's approval. That path also sees appointments and no-shows, which an export cannot show, and it needs a plan that carries API access. We have not run that path against a live account yet, so we would stand it up with you rather than promise it now.

  • Vagaro

    via export

    Your book stays your book. We write nothing into it.

    We start from one file: Reports, then Customers, then Export, then Excel. Only the owner can run that export, so that is the person we need for ten minutes. If you keep a cancellations and no-shows report, send that too.

    From that file we find the people who went quiet and write one message each in your voice. Nothing changes in your setup and your front desk learns nothing new.

    Vagaro already sends your reminders and its lost customer campaign. We work the people those miss.

  • Zenoti

    via export

    Zenoti stays the record. We do not write into it.

    Someone on staff pulls two exports every Zenoti account already has. Guests, from Manage Guests. Then the Appointments report. We join them on Guest Code and read who cancelled, who never rebooked, who came once and never came back.

    That is the whole job on your side. Once to start, then one refresh a week if you want it running.

    We draft a message per guest in your own words, hold every one for a person to approve on a phone, and send from your own email address. Nothing in the booking system changes.

  • Boulevard

    via export

    Your team keeps working in Boulevard. The booking flow does not change.

    Once a week an admin runs two exports every Boulevard account already has: the client list from the Clients tab, then one report from the Reports tab. It takes two files, not one, because the client list carries no dates.

    We find the clients who went quiet and write the follow-up in your voice. Nothing is written back into Boulevard, so your Boulevard number keeps doing what it does today. When someone rebooks, they rebook the way they always have and it lands on your calendar.

  • Jobber

    via export

    Jobber runs your quoting and your invoicing. We work the leads that stall in between.

    Once a week you email us two reports Jobber already builds for you: Requests and Visits. That is the whole job on your end.

    We read who asked for a quote and never booked, and who has not had a job closed in months. Then we write the follow-up in your words, and it goes out from your own email address once it is approved.

    Nothing gets installed in Jobber. Nobody on your crew gets a new login. Exporting is admin only and sits on select plans, so we check which plan you are on first.

  • A spreadsheet

    via export

    This is the path that works everywhere, and it is the one we have run the most.

    Send the file you already keep, as a spreadsheet or a CSV. Your column names do not have to match anything. The $750 audit reads the headers you already have, cleans the duplicates, and tells you what it dropped.

    From there it is the same as every other system on this row. We find who went quiet and write one message each in your voice. Nothing gets installed, and nothing moves out of wherever it lives today.

The approval gate

Nothing sends without you

Every message carrying your name is approved against wording you signed off on before it leaves. That rule is built into the software, and nothing in the software can route around it.

Draft

The software picks the person, picks the wording you approved, and writes the card. Nothing has moved and nobody has been contacted. It is a sentence sitting in a queue.

Under the hood

A draft is a state, not a message in flight. The only code that can move it forward is the code behind the tap.

Approval

It waits. Each card is read against the wording you signed off on, then approved as written or changed first and approved after. Nothing moves until that happens, and nothing ages into being sent.

Under the hood

An edited message is checked again the same way a fresh one is, so a rewrite cannot slip past a rule the original was held for. The page opens behind a token, and a tap without it is refused.

Send

What was approved goes out under your name, from your number and your inbox, inside your hours and at your pace. What was thrown out is closed and never comes back around.

Under the hood

Every send writes a receipt, and a send nobody can confirm is handed to a person rather than guessed at or quietly repeated. Your hours can be narrowed. The software refuses to start if anyone tries to widen them.

Anything that is not a plain stop reaches a person

Our software sorts a reply three ways: someone asking to stop, a wrong number, or a message for a human. That last one is the default, so anything complicated, anything upset, anything at all that is not a clean stop goes to a person at your business. Our software does not argue with your customer, because it was never given the ability to answer one.

The audit

Every number comes from a row in your own file

The $750 audit is plain arithmetic on the file you send us. Nothing in it is a guess.

The same file always gives the same answer

Run it twice and the numbers come out the same twice. No model is inventing figures about your business.

You can check every number yourself

Each figure comes from counted rows in the file you handed over. You can open it and check any of them.

It shows the work

Under every number, the report shows the count and the percentage it came from. You can add it up yourself.

We never open health or treatment records

Health and treatment columns get dropped on the way in. Our software never loads them, so they cannot reach a report.

What a run says, in English

We read 1,204 lines and found 1,156 real people after removing blanks and duplicates.

942 of them are sitting cold. 512 never replied, 174 booked and did not show, 256 were never contacted at all.

At $240 a visit, getting a normal share of them back is worth $15,286 to $25,586.

This is a fake business with made-up numbers. A run on your file prints your own, and the full report shows the math under every one of them.

Show the actual run
Sample run on made-up test data
$ python3 audit/run.py
    --client sample-studio

kancir AI - Pipeline Audit
Client: Sample Studio
made-up test data

[1/4] INGEST
  sample-90-day-export.csv
  1,204 lines (utf-8)
  = 1 header + 1,203 rows
  less 6 empty = 1,197
  less 41 merged = 1,156
  1,156 unique contacts
  2 medical columns
    excluded at ingest
  0.38s
[2/4] SEGMENT
  active 214 | cooling 268
  lapsed 331 | dormant 343
  sitting cold 942
    = 268 + 331 + 343
  ghosted 512
  no-show 174
  never contacted 256
  0.06s
[3/4] VALUE
  avg visit value: $240
    from config, not assumed
  active 214 excluded,
    still in conversation
  cooling 268 x 0.10-0.16
    x $240 = $6,432-$10,291
  lapsed 331 x 0.07-0.12
    x $240 = $5,561-$9,533
  dormant 343 x 0.04-0.07
    x $240 = $3,293-$5,762
  total across cold
    $15,286 - $25,586
  0.01s
[4/4] REPORT     outputs/
  brief.md, full-report.md
  0.05s

+--------------------------+
| TOTAL CONTACTS    1,156  |
| SITTING COLD        942  |
| RECOVERABLE VALUE        |
|   $15,286 - $25,586      |
+--------------------------+
What the audit includes

Sample report, made-up numbers

What you get every week

A report lands every week. Four of them add up to a month that looks like this one.

Seven quiet leads came back, worth about $2,450.

New inquiries
120
Follow-ups sent
48
Came back
7
What we changed
1
Moved the second follow-up from day 3 to day 2 after replies clustered there.

Every number here is counted, not estimated. We keep a record of every message that went out and every reply that came back, and the report is that record added up.

A sample report on made-up numbers, not a client result. It assumes 120 new inquiries a month, 60% of them booking, a $350 ticket, and that we bring back 15 out of every 100 that went quiet.

How this compares

Four ways a lead gets followed up

Doing nothing, or the front desk

  • Costs nothing extra
  • Nobody owns follow-up
  • Leads go quiet after one try

The automation built into your booking software

  • Sends messages
  • Nobody watches whether they work
  • Nobody checks whether it booked anyone

An agency

  • Generates leads
  • Rarely works the old ones
  • Priced for the campaign, not the follow-up

kancir AI

  • Works every lead you already paid for
  • A person approves every message
  • Software we run, with nothing for you to log into
  • One number every week: how many came back
  • One price, worked out from your own numbers

What came back means

A lead that went quiet answers again

Came back means someone who had gone quiet answers you again. They reply, or they book. That is the only thing we count, and it is the number in your report.

We do two things at once. The leads already sitting cold get worked back one at a time, and every new lead gets an answer so it never joins the pile.

We do not generate leads. Your marketing does that. Our job is making sure every lead you already paid for gets a real answer.

We cannot promise you a number before we have seen your leads. What we can promise is that you see every message that went out and every reply that came back, every week.

We are in Colorado Springs, so none of this has to happen over email. We can come in and watch how a lead reaches your front desk before anything gets set up.

See it run on your own leads

Fifteen minutes on the Pipeline Look, complimentary. Honest answer either way.

Get started