Client Zero: How Animus Runs on Animus
Every automation company says their systems work. Few can show you the system running their own company. So here is ours, with real numbers: Animus finds its clients, researches them, writes to them, watches its own health, and reports to its founder every evening - using exactly the architecture we sell. We are client zero, and this is the case study we wrote before asking anyone else to be one.
The Job We Gave Ourselves
A new firm needs conversations, and getting them is exactly the kind of work we tell clients to automate: repetitive, research-heavy, and unforgiving of inconsistency. The brief we set mirrors a client engagement - connect to the systems involved, automate the repetitive middle, keep a human on anything that matters, and measure everything. The pipeline that resulted: a public register of 4,300 qualifying UK firms, refreshed weekly; research on each firm - its website read, its practice profile understood, a personalised three-email sequence drafted - at a pace of about thirty firms every working day; and sending, paced and human-supervised, from dedicated infrastructure.
per firm researched - after we moved the drafting from a premium AI model to a mid-tier one and taught the workflow to read websites directly instead of searching for them. The same cost discipline we apply to client builds, applied to our own.
The Approval Gate, Lived With
We preach human approval, so we ate our own cooking. In the first days, every drafted email landed in our Slack for review before it could go anywhere - the founder read them over morning coffee and clicked approve or reject, one by one. Rejections taught the system: a corporate trustee entity that slipped the filters, a firm whose website was down and whose draft was built on too little. When the drafts stopped needing edits, we dialled the review down - deliberately, on evidence - to spot-checks and a daily digest. That is precisely the trust-transfer process we describe to clients, experienced from the client's chair: gates first, evidence, then earned autonomy.
What Broke, and Who Noticed First
Honesty requires this section. Twice, our AI provider account ran out of credit and the research pipeline failed - silently, the first time, costing a day of work before anyone noticed. The second time taught us properly: our health monitoring now probes the AI service every thirty minutes and posts to Slack the moment anything fails, with a recovery notice when it clears. We also discovered our first emails were probably dying in law firms' spam filters - the security gateways that guard legal inboxes distrust links from young domains - so the workflows now write link-free first emails. Systems break. The product is not the absence of failure; it is how fast you notice and how permanently you learn.
The Cockpit
Everything reports to one dashboard: the pipeline funnel, every firm's status, why each rejected prospect was rejected, system health, and a green-line-red-line chart of the machine's daily output. Each evening at 17:55 a digest lands in Slack with the day's movement. This is the same shape of oversight every client build gets - because an automated system you cannot see into is not an asset, it is a liability with good manners.
What client zero proved:
- The five-day-shape build is real - our own pipeline went from nothing to operating in under a week of engineering
- Approval gates work as designed: early rejections caught real problems, and autonomy was earned, not assumed
- Costs are controllable: research per firm fell by roughly ninety percent through model choice and workflow design
- Monitoring earns its keep: both failures were caught by the system, not by luck
- The honest limit: automation got us conversations infrastructure - the conversations themselves still take humans
Ask any automation vendor: “show me the system running your own business.” We wrote this page so our answer is a URL.
When our founding clients have numbers of their own, their stories will sit beside this one - named, with permission, or not at all. Until then, client zero is on the record: the system works, we run our company on it, and everything above is checkable in a discovery call, live, on screen.
Ready to deploy your AI agent?
Our engineers connect your systems and build workflows like these for you - live in five working days, supported from then on.