We build agents and run them
Umbra Studio.
An applied AI studio that rebuilds the workflows your company runs on as agents — and keeps running them for you.
Not consulting. Not prototypes. We embed in one high-friction workflow, rebuild it with intelligence at the core and humans above the loop, ship it to production, then operate it on our own rails. We don't sell you an agent — we sell you a system that keeps getting better while we run it. Governance stays non-negotiable.
Not a deck. A working system, in production.
Umbra Studio is the applied AI arm of Umbra Group. We work with operators on high-friction, judgment-dense workflows — the ones AI hasn't touched because the stakes are too high — we rebuild them as agents end-to-end, and we run them for you. Four rules govern every engagement.
AI is everywhere — except on the bottom line.
The gap isn't model capability. It's the workflow around it. Most organizations bolt AI onto a legacy process and wonder why nothing moves. Studio rebuilds the workflow so the value actually lands.
The technology can increasingly be found everywhere — except on the bottom line.
McKinsey Growth, Marketing & Sales Practice, Reinventing marketing workflows with agentic AI, April 2026.
Start small, ship the build, then go continuous.
Three steps, one discipline. A low-commitment diagnostic proves the fit, the Sprint ships the system to production, and — by default — Watch keeps running and improving it for you, month after month. Watch is where the value compounds, not an afterthought. You only scale once it's earned. The Pulso credits 100% toward the Sprint.
Seven platforms. One cockpit, in orbit.
These aren't pitches. They're the agents Studio builds, operating in production today on a live reference deployment. One cockpit — Atlas, a meta-agent with memory — sits at the center and orbits six executors that do the work. Every Sprint deploys the subset a client's workflow needs.
A living system — the world, and the agents that orbit it.
Umbra Studio isn't a logo. It's an eclipse with agents in orbit — a world that contains the intelligences that do the work. The prime is the world itself; Scout and Atlas are two of the faces inside it.
The eclipse the agents orbit. Not an agent — the system itself, serene and steady, holding the whole stack.
The scanning-eye probe. Sweeps the field, surfaces what matters, and drafts the first read before anyone asks.
The command-eye with a crown of agent-nodes. Watches the stack, routes the work, and keeps judgment above the loop.
Fixed-fee, outcome-tied, two tracks.
You start with a low-commitment diagnostic and only scale once it's proven a fit. Two tracks — US and LATAM — adjusted for purchasing power, not discounted as a favor. The engagement quality is identical.
Pulso
A two-week instrumented diagnostic of one workflow — the spec, baseline metrics, and a real estimate of what a Sprint would do. No slide deck. If you decide to build, it was free.
Lighthouse Sprint
One workflow taken end-to-end — observation through production — then operated by the Studio on Watch by default. Complex or regulated workflows are scoped per engagement; running it yourself is the exception, priced separately.
Watch · Essentials
We keep it running: foundation maintenance + governance + monthly health report. No +1 improvement.
Watch · Standard
The default: Studio keeps operating + improving it — Essentials + the eval-gated +1 improvement + quarterly architecture review.
Watch · Pro
Standard + named-engineer commitment + 4 hrs/mo advisory + 4-hr P0 SLA.
We tested it on our own newsroom first — Indietheka.
Digital media · editorial pipeline · 8 weeks · Q4 2025. Applied internally to Umbra Group's own outlet — one editor, manual end-to-end, human-bottlenecked at every step, founded June 2012 with no outside capital. Six months later: 26 workflows running 24/7, Atlas orchestrating, 8 patterns externalized as @umbra/* packages, zero missed publish days across the cutover.
Before — manual, solo editor
- 210 min end-to-end cycle time per article
- 2–3 articles/week, capped by one person's hours
- ~8% error rate · platforms frequently skipped
- 1–2 of 5 distribution channels actually reached
- 4.5 hrs/week on playlist curation alone
After — 16-agent system, 24/7
- Cadence daily, no human bottleneck
- RSS → discovery → write → SEO → publish → distribute → analytics, automated
- Every post born SEO + JSON-LD governed (119 enriched day-1, 0 failures)
- IG / FB / Lnk.Bio every cycle, never skipped
- Atlas orchestrating; weekly performance dashboard self-fixing
"We didn't automate the editing. We automated everything that wasn't editing. Then the editing got better, because the editors had time to edit."
See the Pulso deliverable built from this engagement.
The Indietheka case, instrumented as a sample Pulso — the AI-opportunity audit a client receives, reverse-engineered from this exact Sprint.
One workflow. A system we run for you. In production — or it doesn't count.
Start with a Pulso: a two-week instrumented diagnostic of one workflow, credited 100% toward the Sprint. If it's a fit, the Sprint ships it and Watch keeps running it by default; if it isn't, you keep the map.