passcut

AI agents for small businesses

Automate any repeated process in your business. We build AI agents and workflows for your company. Show us how the work is done today, and we turn it into an agent that does it.

Already know the workflow you want built? Get a quote →

You own everything we build Every action logged Live US-hours overlap

Invoice entry

under 1 min

down from 5–10 min each

Inbox triage

minutes of review

down from ~1 hr/day/person

Support first response

minutes, any hour

down from hours, timezone-bound

Typical results for these workflows. We measure your baseline in the audit and commit to numbers per engagement.

How it works

Hand it off. Stay in control. Get the hours back.

1

Hand it off

Pick one repetitive workflow. The agent learns it from your real data, the way your team does it today.

2

Stay in control

The agent prepares the work. Your team decides. Anything that touches money, customers, or your books waits for a human sign-off.

3

Get the hours back

The repetitive work stops landing on your team. We monitor, fix, and improve the agents every month under the Care retainer.

What changes

A workday with agents running

Your team keeps the same inbox and the same accounting tool. There is no new dashboard to learn. The difference is how much is already done when they sit down.

5:58 PM

You log off

The inbox holds invoices, receipts, pending approvals, 214 unread. Normally your team works through it tomorrow morning.

Inbox

208 unread
S

SupplierCo

Invoice #0447, $2,340.00

5:41 PM
S

Sarah M.

Re: re: re: approval??

5:32 PM
N

no-reply

urgent

URGENT: pls respond

5:20 PM
S

scan@hp

Receipt (blurry photo).jpg

4:58 PM
A

A. Chen

Where is my order?

4:30 PM
A

AR Dept

urgent

Overdue notice, again

3:12 PM

+ 202 more waiting for tomorrow

2:00 AM

Agents at work

The agent reads, codes, matches, and drafts every document overnight. Anything ambiguous is set aside for a person.

agent-run.loglive

22:14:11 invoice_supplierco_0447.pdf 14 items extracted

22:14:12 GL coding: 6110 Software (11), 6220 Freight (3)

22:14:14 PO match: PO-1187 , budget line

22:14:15 queued for approval @sarah (threshold $500+)

22:31:08 receipt_amazon_8821.pdf 6 items extracted

22:31:09 matched to card ending 4417 , GL 6415 Office

22:47:52 statement_chase.csv 128 rows reconciled

23:02:17 exception: receipt_blurry_019.jpg human queue

processing queue

0 processed2 exceptions, 0 errors
9:00 AM

You approve

You open a prepared queue: drafts ready to send, invoices ready to post, and the few items that need your judgment.

Prepared for you

overnight, ready to approve

done

handled overnight

0 / 214

0 drafts ready to send

Review

0 need your judgment

0 invoices queued for approval

weekly digest: 3 recurring requests worth automating next

While you sleep

Your agents work around the clock

You keep your hours; the agents don't keep any. Work that arrives at 6 PM is processed, checked, and queued for approval by your morning. A monitoring agent watches the system overnight and fixes what it safely can; anything it can't gets escalated to our team.

Your team

business hours

Passcut agents

around the clock

6 PM

you hand off and log out

2 AM

agents processing, monitored

8 AM

queue ready for approval

Integrations

Plugs into what you already run

The agent connects to the tools you use today and works through their APIs: accounting, help desk, inbox, chat, CRM. There is no migration and nothing new for your team to learn.

Tool not shown? The audit maps your exact stack: official API first, exports second, browser automation as the last resort.

GmailIntercomZendeskHubSpotXeroStripeDriveQuickBooksPasscut agent
  • QuickBooks
  • Gmail
  • Stripe
  • Xero
  • Zendesk
  • Shopify
  • HubSpot
  • Intercom
  • Notion
  • Google Drive
  • Asana
  • Help Scout
  • SAP
  • Expensify
  • Brex
  • Airtable
  • Dropbox
  • Mailchimp
  • Trello
  • ClickUp
  • Zapier
  • Make

Human in the loop

Your team approves every sensitive step

Every agent starts in prepare-and-approve mode. It reads, codes, matches, and drafts; a person makes the final call. Nothing touches money or customers without human approval until you widen the agent's autonomy, category by category.

invoice_supplierco_0447.pdf

prepared by Passcut agent, overnight

ready
Amount$2,340.00
GL coding6110 Software, 6220 Freight
PO matchPO-1187, within budget
Supplierverified, 12 invoices, 0 disputes
Waiting for your approval, always human
Approve & post Flag for review

This week, invoice agent

241 items

232 auto-filed9 approved by you
0 acted without your approval, 14.2h returned

Why teams trust it

Ownership, logs, open pricing

You own the system

Code, config, API keys, data: all in your accounts. If we vanish, your agents keep running.

Every action logged

Each action records what was read, what was done, and who approved it. The log is exportable for your accountant or auditor.

Published pricing

Fixed-price pilot, published build ranges, cancel-monthly maintenance. See pricing

agent-run.loglive

22:14:03 watching [email protected], 3 new documents

22:14:11 invoice_supplierco_0447.pdf 14 items extracted

22:14:12 GL coding: 6110 Software (11), 6220 Freight (3)

22:14:14 PO match: PO-1187 , budget line

22:14:15 queued for approval @sarah (threshold $500+)

22:31:08 receipt_amazon_8821.pdf 6 items extracted

22:31:09 matched to card ending 4417 , GL 6415 Office

22:47:52 statement_chase.csv 128 rows reconciled

23:02:17 exception: receipt_blurry_019.jpg human queue

06:00:00 nightly report: 241 processed, 2 exceptions, 0 errors

processing queue

0 processed2 exceptions, 0 errors

FAQ

Common questions

Longer answers live on how we work and pricing.

What does Passcut do?

We build and maintain AI agents inside the systems you already run: accounting tools, help desks, shared inboxes. The agent does the repetitive work. Your team approves anything sensitive. You own the system we deliver.

Do we need to replace our current software?

No. The agent connects to what you already run: QuickBooks, Xero, Zendesk, Gmail, Slack, your ERP.

Can the AI act without a human?

Only where you allow it, category by category, after you have watched it work. By default, every payment, refund, customer-facing send, and posting to your books needs one human click.

How do engagements start?

With a free 45-minute audit. We map three processes agents could take over, estimate what each would save, and quote the cost. Most clients then run one paid pilot on one workflow.

Where is the team located?

Jakarta, with live overlap with US business hours every day. The agents run around the clock, so work you hand off in your evening is usually done by your morning.

What does it cost?

Pilots start at $4,900 fixed price, production builds from $9,000, and maintenance is $1,500 per month per workflow. Full ranges and what's included are on the pricing page.

How long until an agent is live?

A working pilot on your real data typically runs within 30 days of starting. Production hardening adds a few weeks after that, depending on the systems involved.

Is our data used to train AI models?

No, not for training. The work runs through your own accounts with the model providers, under their no-training API terms, and we document that in the engagement.

Which AI models do you use?

The best fit per task: typically Claude or GPT-class models via API, smaller models where they suffice. Model choice is documented and swappable; you are never locked to one vendor.

What happens when the agent makes a mistake?

Mistakes surface in the run log and the human queue, not in your books. Sensitive actions wait for approval, so errors are caught before they post, and under the Care retainer fixing the cause is our job.

Start with a free workflow audit

A 45-minute call. We map three processes agents can take over, estimate what each would save, and quote what it would cost. No obligation.

Get your workflow audit45 minutes, written summary, no obligationAlready know what you want built? Get a quote