/ About
We run the work nobody wanted to keep doing
Agentiq is an automation agency. We map how your operations actually run, build agents on the tools your team already uses, and stay on to operate them once they are live.
/ Why we exist
We started Agentiq after watching the same scene in company after company: three people chasing invoices by hand, a support queue triaged by whoever got in first, a Monday report rebuilt from scratch every single week. None of it was hard. All of it was expensive.
The tooling caught up in 2023. The practice did not. Plenty of teams could demo an agent, far fewer could put one in production and keep it honest six months later. That gap is the whole business: we build agents that survive contact with real operations, and we stay on to run them.
So we measure everything in the only two units that matter to an operations lead. Hours returned to your team, and cost per run. If an agent cannot beat the person doing the job today on both, we do not ship it.
/ Inside the room
Every Thursday
The whole team reviews one live workflow together, run log open, before anything ships.
Human gate first
The handoff rule is written before the prompt. An agent earns autonomy by proving it on evals.
No eval, no launch
Nothing reaches a client account without a failing test written first and a costed baseline.
Three time zones
San Francisco, Amsterdam, and Sydney, so a broken run has someone awake inside the hour.
/ Principles
Humans approve the edge cases. Autonomy is earned, not assumed.
Every workflow ships with a handoff rule written before the prompt. The agent runs the ninety percent it has proven it can handle, and your ops lead sees the rest in Slack with one click to approve.
Every run is logged and costed. If we cannot measure it, we do not ship it.
You get the run log, the eval scores, and the cost per run from day one. That is also how we price the work: no agent enters managed operations until the numbers beat the manual baseline.
You own the stack. Leaving us takes an afternoon, not a migration.
Prompts, evals, and integrations live in your repositories and your accounts. We are useful because staying is worth it, not because leaving is painful.
005 / By the numbers
/ The team
Daniel O
Ran support operations for a 400-seat helpdesk before deciding a machine should do most of it.
Ines V
Builds the eval suites. Holds the rule that no agent ships without a failing test first.
Rahul M
Owns the thirty-eight connectors. Has opinions about webhook retries you will hear about.
Theo B
Writes the audit that starts every engagement, and the monthly ledger that ends each one.
/ How we got here
Two people and a spreadsheet
Ines and Daniel take on one client, chasing invoices with scripts and a rules engine. It works well enough that the client asks for a second workflow before the first is finished.
The first agent in production
Support triage goes live behind a human gate for a 40-seat helpdesk. Eighty-four percent of tickets stop reaching a person, and the eval suite becomes the thing we sell.
Managed operations
Clients ask us to keep running what we built. We stop shipping projects and start operating them, with a monthly ledger of hours returned.
247 agents, three time zones
Fourteen people across San Francisco, Amsterdam, and Sydney, running 1,406 workflow runs on an average day at four cents apiece.