Stop sorting work by hand.
We build the automations that sort, route, and answer the repeat work landing in your inbox, CRM, and Slack. Fixed price, built in the tools you already use, usually live in about a week.
Tell us the bottleneck. We reply within 1 business day with fit, the likely first build, and a realistic range.
Harrison Vance, founder, reads every inquiry.
- $2,500
- Bottleneck Audit, paid after delivery
- ~1 week
- Typical first build
- 1 day
- Reply to every inquiry
- Prices published
- Pass/fail checklist
- One free fix round
- Runs in your accounts
- Reply in 1 business day
Every request finds its owner.
Nobody sorts the inbox by hand.
Three places teams usually start.
Each is a fixed-price first build, run by a Rivet worker in the tools you already use. Not the one you need? See all 15 workers.
Fix lead follow-up
Every inquiry gets sorted, prioritized, and handed to the right person with the right context.
- In
“We have 3 clinics and new patient inquiries sit in a shared inbox for days. Can you quote a fix? Hoping to start next month, budget around $5k.”
- DecidedOwner Sales · Maya R. New project with a timeline and a budget. High fit.
- Did
- slack_post_message → #sales-leads Done Who, what, timeline, budget, for Maya to act on
- crm_upsert_contact → HubSpot Held for approval CRM writes wait for a person at Level 2
{"id":"run_7f3a","worker":"lead-intake","autonomy":2,"status":"awaiting_approval","route":{"lane":"sales"},"actions":[{"tool":"slack_post_message","outcome":"executed"},{"tool":"crm_upsert_contact","outcome":"held","approval_id":"apr_2c91"}]} Example run. People and companies are illustrative.
Sort the inbox
Incoming requests are sorted by urgency, category, and owner before the team touches them.
- InInvoice 4471 charged twice
“Our card was charged twice for invoice 4471 this morning. Please refund the duplicate today, our finance close is tomorrow.”
- DecidedOwner Billing · Sam P. Duplicate charge with a same-day deadline. Urgent.
- Did
- slack_post_message → #billing Done Flagged urgent, with the invoice number and deadline
- webhook_send → label_email Done Labelled Billing / Urgent in the inbox
{"id":"run_91bd","worker":"inbox-triage","autonomy":2,"status":"handled","route":{"lane":"billing"},"actions":[{"tool":"slack_post_message","outcome":"executed"},{"tool":"webhook_send","outcome":"executed"}]} Example run. People and companies are illustrative.
Answer repeat team questions
We turn docs, SOPs, and scattered know-how into a simple assistant that shows the source behind each answer.
- In
“Can I approve a $350 refund myself or do I need finance?”
- DecidedAnswer Yes, up to $500. Found in Refund policy › Who can approve a refund.
- Did
- kb_search → workers/knowledge Done 1 matching section, cited
- slack_post_message → #account-managers Done Answer posted with the source linked
{"id":"run_c04e","worker":"team-answers","autonomy":2,"status":"handled","actions":[{"tool":"kb_search","outcome":"executed"},{"tool":"slack_post_message","outcome":"executed"}]} Example run. People and companies are illustrative.
Done means it passes the checklist you signed.
Before we build, we agree a short list of pass/fail tests with you. The build is finished when every row passes. If a row fails, we fix it at no cost.
- A-01A test form fill shows up in the CRM with an owner within 2 minutes.Pass
- A-02A lead with no company name goes to the review queue, not to sales.Pass
- A-03The owner gets a Slack message with source, fit, and urgency.Fix round
Every row has test steps and an expected result. Vague goals like “faster” become a number or are left out.
You check each row in your own tools, with your own sample data.
If a row fails, we fix it at no extra cost. New ideas go on the list for the next build.
You decide how much the workflow does on its own.
Every first build launches at Level 2: the workflow does the work, and a person approves anything risky. It moves up only when the monthly numbers support it.
- L4Handles exceptionsSelected edge cases are handled automatically, under rules you set.
- L3Acts on its ownApproved work runs without a person. Unclear items still go to review.
- L2Acts with approvalThe workflow does the work. Risky steps wait for a person.Every launch
- L1SuggestsThe workflow recommends. A person acts.
- L0ManualA person does everything.
Run Rivet yourself, or have us run it.
Every build runs on Rivet, in your accounts and with your keys. Start small; each step stands on its own. See every price →
- Run it yourself
Rivet
$0Open source, Apache-2.0. Your Claude key.
Run Rivet free - 01 Have us run it
Bottleneck Audit
$2,500Paid after delivery. Credited toward the build.
Book the $2,500 audit - 02 Have us run it
First worker
$3k–$7kFixed price, about a week, pass/fail checklist.
See prices by worker - 03 Have us run it
Care plan
$300–$500/moPer worker. Monitoring, fixes, monthly scorecard.
How it works
Three slots for founding clients.
Our proof pages are modeled builds, not client case studies. We would rather say that plainly and offer a fair trade.
20% off your first build
Applies to the first fixed-price build, including builds that start from a Bottleneck Audit.
Direct access to the builder
No account-manager layer. You work with the person building your workflow.
In return: a named case study
Once the workflow is live, you agree to a short public case study or a reference call.
What does it cost, all in?
A first build has a fixed price agreed in writing, usually $3k–$7k. The $2,500 audit is optional and credited toward it. Software and Claude API usage are billed to your own accounts at cost. The care plan ($300–$500/mo per worker) is optional.
How much of our time does it take?
About 45 minutes to show us how the work runs today, and a short session at the end to run the acceptance checklist in your own tools. Everything in between is async.
Are we locked in?
No. Workers run on Rivet, which is open source, in your accounts and with your API keys. The care plan is month to month with 30 days’ notice, and the workers keep running if you cancel.
What happens to our data?
It stays in your tools. Workers call Claude with your key, nothing is copied to our systems, and every action lands in an audit log you can read.
What if it gets something wrong?
Every worker launches at Level 2: routine steps run, risky ones wait for a person’s approval, and unclear items go to review. If a checklist row fails at handover, we fix it at no cost.
Tell us what needs fixing first.
One bottleneck, the current tools, and where work gets dropped is enough to start.