AI automations
The repetitive work, handed to software, in about a week.
AI workflow automations that read documents, move data between your tools and route requests, with a person approving anything that matters. Tested, logged and under a monthly cost ceiling.
Prefer to write? Send your requirements
In short
PutThrough builds AI workflow automations (invoice and document processing, CRM and spreadsheet syncing, lead enrichment and routing, and scheduled reports) in about a week, at a fixed price agreed in writing after a free scoping call. Each automation runs one workflow end to end: a trigger, AI steps that read and extract, checks against your rules, and actions in your tools. Anything the AI isn’t sure of goes to a review queue, every run is logged, and running costs stay under a monthly ceiling you set.
01Who it’s for
For teams copying data between tools by hand.
-
Finance and operations teams
Invoices, receipts and forms read and entered for you, with anything unusual left for a person.
-
Sales teams
New leads researched, scored and routed to the right person, with the CRM updated as it happens.
-
Founders doing it all themselves
The weekly report, the onboarding emails and the spreadsheet updates, run on a schedule.
02Example uses
What a one-week automation usually is.
-
Invoice and document processing
Reads invoices, receipts or forms, extracts the fields you need and creates the record in your accounting or operations tool.
-
CRM and spreadsheet syncing
Keeps your CRM, spreadsheets and database in step, so nobody types the same data twice.
-
Lead enrichment and routing
Researches each new lead, scores it against your criteria and hands it to the right person with a summary.
-
Scheduled reports
Pulls the numbers you check every week and sends a written summary on a schedule.
03What fits in a week
One workflow, end to end, with checks.
Fits in a week
- One workflow from trigger to action, such as invoice to accounting
- Two or three tools connected, such as your inbox, CRM and a spreadsheet
- AI steps that read, extract and classify, with confidence checks
- A review queue for anything uncertain
- A log of every run, and an alert when one fails
- A monthly cost ceiling
Needs longer, or a call first
- Payments or anything irreversible without a person approving
- Many workflows at once
- Replacing a whole system such as your ERP
- Formal compliance programs such as HIPAA or SOC 2
04A sample week
Seven days, one workflow.
The same seven days as every build, with this product’s work in them. Your week is planned on the scoping call.
- Day 0
Scoping call and fixed quote
The one workflow, what triggers it, the tools it touches, and which steps need a person’s approval.
- Day 1
Access, rules and test cases
Access to the tools involved, your rules written down, and a set of real past cases with the right outcomes.
- Days 2–5
Build, in the open
The workflow, its AI steps and its checks, run against the test cases every evening, with a preview you can try.
- Day 6
Harden
Failures and retries tested, the review queue and alerts switched on, and the cost ceiling enforced.
- Day 7
Launch and handover
Running on your real data, then the handover: how to read the logs, change the rules and pause it.
05What you get
What you walk away with.
- The automation, running on your real data
- The test cases, and the results it passed at launch
- A review queue for anything it isn’t sure of
- A log of every run, and alerts when one fails
- A monthly cost ceiling, and a guide to changing its rules
Plus the standard handover every build ends with: source code, a written handover, a recorded walkthrough and every account in your name. The full Day 7 list
06Stack
Plain code you own, not a maze of zaps.
Built as code in your repository and run on your own accounts, so there are no per-task fees to a workflow platform. Where Zapier or Make already does the job well, we say so.
- Triggers
- Webhooks, email and schedules
- AI
- Claude or OpenAI, behind one interface
- Runtime
- TypeScript on serverless functions
- Records
- Postgres, for the run log and the review queue
07Pricing
A fixed price, and a fixed ceiling on running costs.
Your exact price is fixed in writing after the scoping call, before you commit to anything.
After launch, the care plan is from $499 per month.
Prefer to write? Send your requirements
08Questions
01 How is this different from Zapier or Make?
Zapier and Make are good for simple, standard connections, and if one fits your case we will tell you. We build when the workflow needs judgment, such as reading messy documents, applying your own rules or adding a review step, or when per-task fees would add up. You own the code and run it on your own accounts.
02 What happens when the AI gets something wrong?
Every AI step reports how confident it is. Anything below the threshold you set goes to a review queue for a person, and every run is logged, so a mistake can be traced and fixed.
03 How is an automation different from an AI agent?
An automation follows the same steps every time and uses AI for the parts that need reading or judgment. An agent decides its own next steps. Most repetitive back-office work is better served by an automation: it is cheaper to run and easier to check.
04 What will it cost to run?
That depends on how many runs it makes and which AI model it uses, which is why every automation has a monthly cost ceiling you set. We estimate running costs on the scoping call.