JK
Human lead · accountable for every agent below
Thirty years of IBM i, RPG and enterprise systems. Scopes every engagement, reviews every change, approves anything risky, and is the person you call.
How we work
A senior human lead. A bench of working agents.
Solanko is deliberately small. One experienced engineer leads every engagement, and a bench of specialized AI agents does the heavy lifting: operations, code, documentation, follow-up. It is the same way we run our own company.
Every agent works under a named human who is accountable for the outcome.
The bench
Who does the work
Our agents have names because we work with them every day. What matters to you is their job, and who answers for it.
Lola
Operations & engineering
Runs our infrastructure: deployments, monitoring, security-key rotation, research and code. Works alongside JK every day.
Cola
Customers & revenue
Runs our lead pipeline: intake, CRM follow-up, inbox triage and campaigns. Thousands of records, every action traceable.
Argus
Watch & alerts
Watches everything we run: certificates, backups, form health and servers. When something drifts, he sends the alert, so problems get caught early.
Rola
Defense programs
Cola's counterpart for defense contractors. Same discipline, stricter rules: she works only inside the client's environment, under the client's controls.
An engagement
Assess. Pilot. Scale.
Small, measurable steps. Nothing moves to the next stage until the last one has proven itself on a real metric.
Assess
One to two weeks. We map your systems, your data boundaries and the one process where an agent would pay for itself first.
Pilot
Four to eight weeks. One agent, one job, one metric, running inside your environment with a human approving anything risky.
Scale
If the pilot earned it, we extend it and keep running it with you. If it didn't, you'll hear that from us first.
Ground rules
What our agents never do.
- Act on anything high-risk alone.Payments, deletions, customer messages and production changes wait for a person.
- See more than their job needs.Least-privilege access, scoped per agent and per task.
- Hold the keys.Credentials stay encrypted, and the keys that open them are a person's to hold, not an agent's. An agent is handed only the one secret its job needs, nothing more. At a client the same rule applies with the client's own keys, so control stays where it belongs.
- Pretend to be human.When an agent writes to you, it says so.