Clerkwell builds AI systems that read the inbox, pull data out of documents, update your CRM and draft the weekly report. A person on your team approves anything that matters before it goes out.
Built in Canada for teams of 5 to 100 people across Canada, the US, the UK and Australia.
If someone on your team copies information from one place to another for hours each week, that work can probably be automated. These are the jobs we see most.
We map how work moves through your team and pick 3 to 5 tasks where automation will save the most time. You get a written plan with estimated hours saved.
About 1 weekWe build on the tools you already use, test with your real (approved) data, and measure time saved and accuracy against how things ran before.
2 to 4 weeksWe watch the automations, fix anything that breaks, tune them as your business changes, and send a monthly report on what they handled.
Monthly, cancel anytimePick a sample invoice, or change the total and run it again. These are the same checks we set up for clients, running here in your browser on example data.
Move the sliders to match one repetitive task on your team. This is simple arithmetic on your own numbers, not a promise.
Each one is adapted to your tools and rules. Yellow steps are where a person on your team stays in control.
Typical result: replies go out in minutes instead of the next day.
Typical result: data entry shrinks to reviewing exceptions.
Typical result: Monday's report is ready before Monday.
Most agencies make you book a call to hear a number. Ours are below. Your exact quote depends on how many systems are connected, and you get it in writing before any work begins.
Half up front. The other half only when it passes. We agree the success measures in writing first, such as hours saved and accuracy on your own data. The second payment is due when the build meets them.
The fee is credited to your build if you go ahead.
Best first project for most teams.
For teams automating a whole department.
Support hours and response times are set in writing.
Our first three clients get a tightly scoped pilot for $1,500–2,000 CAD. In return we ask for structured feedback and, if you're comfortable, permission to publish a short case study.
Apply for a pilotPrices exclude applicable taxes. US clients can be invoiced in USD at the equivalent rate. AI model and software usage is billed to your own accounts where possible, so you see exactly what it costs.
Ask us, and ask everyone else you talk to. These are our answers.
Published price ranges, and a fixed quote in writing before work starts. No hourly billing surprises.
We measure the task before and after on your own data, against targets you agreed to. The final payment depends on it.
Fixed rules check the AI's work. Anything uncertain, and anything going to a customer, waits for a person on your team.
You do. It runs in your accounts with your keys. If we part ways, it keeps running.
Defects are fixed free for 30 days. After that, an optional monthly plan covers monitoring and fixes with set response times.
Payments, outgoing customer messages and important decisions wait for a person on your team to approve them.
We build inside accounts you own wherever possible. If we part ways, everything stays with you.
Each client's credentials and data are kept separate, and secrets are stored encrypted.
Each automated step is recorded, so you can see what happened, when and why.
When something fails, you get an alert and a person picks it up. Nothing fails silently.
We design to Canadian privacy requirements (PIPEDA and provincial law) and the rules where you operate.
Usually not. We connect to the tools you already have, such as Gmail or Outlook, Excel or Google Sheets, HubSpot and other CRMs, QuickBooks, and most systems that offer an API or email export.
No. Most of what we build runs in the background: reading documents, moving data between systems and preparing drafts. A chatbot is only one option, and often not the most useful one.
We design for it. Uncertain results are flagged and sent to a person, required fields are checked by fixed rules, and important actions need approval. During the pilot we measure accuracy against your own data so you can see the error rate before relying on it.
You own the workflows we build for you and your data. We keep the right to reuse our general-purpose building blocks across clients, which is part of why our prices are lower than fully custom development. The agreement spells this out.
An audit can usually start within a week. A starter project typically takes 2 to 4 weeks from kickoff to handover.
Yes. We work remotely with clients in the US, the UK, Australia and elsewhere, and schedule calls in your time zone.
We'll reply within one business day with a time to talk. On the call we'll look at one workflow and tell you honestly whether automation is worth it.
Or email [email protected]