We don't sell "an AI agent." We field crews.
A single agent in production is a fragile thing. A small crew — narrow agents, a human editor, a real eval harness — is a system you can ship and audit. Below: the three crews running inside our studio right now. Each runs against the same studio rule of thumb.
1 crew : 1 outcome
1 human : 1 final approval
1 eval : every commit, no exceptions.
Reads inbound, enriches it, drafts a tailored response, hands it to a human within the hour.
- · scout.agent — inbound triage + ACRA / LinkedIn enrichment
- · drafter.agent — proposals from past-win library
- · scheduler.agent — calendar + follow-up loop
- · human-in-the-loop — every send is approved
Most of this site's articles are produced by this crew with a senior engineer as editor.
- · researcher.agent — sources + Singapore angle
- · drafter.agent — long-form, voice-controlled
- · fact-checker.agent — citations and numbers
- · editor (human) — every piece, before publish
Sensor data → adaptive control loops → reports the regulator can read. Embedded with WaterDoctor (NUS spinoff, wGrow co-investor). The hard kind of agentic.
- · ingestor.agent — sensor + lab telemetry
- · controller.agent — adaptive treatment loop
- · reporter.agent — auditable run summaries
- · two senior wGrow engineers, on the ground
Want a crew embedded with you?
We typically run a one-week diagnostic, then propose a crew shape, eval plan and weekly cadence. Most engagements start in 2–3 weeks.
Brief a crew →