# KYOKOlabs

> KYOKOlabs is an AI agency for agencies and service teams. Tagline: "Your team, running on AI." We help agencies accelerate AI adoption to unlock their full potential, installing AI as the way a team actually works. We work beside a client's people, on live client work, until more goes out the door and better goes out with it. Get your team on AI this year; be hard to compete with next year.

KYOKOlabs was founded in 2023 by Calvin Gerus. We work remote, in the client's channels and on their calendar, against their real work rather than a sandbox.

## The maturity curve

There's a path from asking AI a question to handing AI a job. Three stages; nobody skips one.

1. Chat: AI as a better search bar. You ask, it answers, you paste the answer into work you were already doing. Where most of the industry has sat for two years. The tool never sees your actual work, so it never gets better at it.
2. Working alongside you: AI in the room with you. Claude Cowork with your tools connected, working on the real thing: your files, client channels, project management, codebase. It builds the asset instead of describing one, runs the task instead of explaining it, tells you what needs attention before you go looking. Where teams are arriving now, and where Kyoko spends most of its time with clients.
3. Agents in roles: AI holding a job. One agent that owns a function start to finish and comes back to a person where judgment is required. Almost nobody is here yet; the teams that get stage two right are the ones who will get here first.

## Services

Two services, one practice: we set your team up on Claude, we build the workflows they run every day, then we build the agents we only find by watching how you work.

### AI OS Enablement (stage one to stage two)

An AI OS is where your team's day starts. We build it on Claude, connected to every tool the team already uses, so it knows their clients, their projects, and the way they work. Three tiers; each carries everything before it, all built on the client's real work.

- Foundation (3 weeks, whole team): discovery call and a map of how work really moves; one-to-one coaching for the AI leads the client names from their own team; Claude 101 for everyone; a first shared Project loaded with the client's brand, clients, and live work; accounts, permissions, and core tools connected.
- Workflow Install (6 weeks, most teams start here): everything in Foundation, plus a prioritization workshop; two workflows built end to end and handed over as skills the team owns; Projects per client and per service; the whole tool stack connected with plain notes on limits; Claude 201.
- Full AI OS (8 to 10 weeks, the whole company): everything in Workflow Install, plus the AI OS itself (a shared memory for the company that team and AI both keep current); developers set up on Claude Code inside their repos; designers on Claude Design and Figma with prototypes drawn from the client's component library; an optional monthly retainer; a handover built so the client's AI leads run everything without Kyoko.

### Custom AI Agent Development (stage two to stage three)

An agent is software that runs one job start to finish and comes back to a person where it matters. The best ones are found, not specified, which is why this usually comes after enablement. Process: Scope (map the real task and where a human stays in the loop) → Build (wired into the client's own tools, living in their systems) → Prove (run on live work beside the manual version until the output holds up) → Hand over (deployed, documented, owned by the client; maintained by Kyoko only if wanted).

In build now: a project manager agent for small client engagements (the work too small to staff a PM against), found inside a live engagement rather than on a roadmap.

## How we work

- Time from the client: about five hours a week from their AI leads while a workflow is in build.
- Who they work with: Calvin Gerus, in their channels and on their calendar for the length of the engagement.
- The team: nobody is trying to make the client's team smaller; the goal is a team that's harder to beat. The person most nervous about AI is usually the one who ends up running it.
- What the client keeps: every skill, project, and connection built is theirs. Packaged, documented, and still running after Kyoko leaves.
- Pricing: fixed per engagement, never hourly; the whole plan and price are visible before anything is signed.
- What Kyoko doesn't do: clean up data, migrate old systems, or rebuild infrastructure. We build on what already runs.

## Tools

Anthropic Claude: Claude Cowork, Claude Code, Claude Design. Skills, Projects, connectors, and the Model Context Protocol (MCP).

## Contact

Book a free 30-minute discovery call: https://kyokolabs.xyz/#contact
We'll dig into how your team really runs, work out which stage you're on, and map the path out loud. If it's not a fit, you still leave with the map.
Email: hello@kyokolabs.xyz
Website: https://kyokolabs.xyz
Client portal: https://portal.kyokolabs.xyz/portal