— 03
Active2026—
Signal.
Every product calls itself “AI.” Most of it is noise. Telling them apart is the work.
[ signal — the trace that survives the noise ]
— The practice
An operator's approach to AI — not an engineer's.
The villain isn't AI. It's the market noise — every product slapping “AI” on its label to raise a round. Telling the real from the rest, and putting the real into the work, is what clients are actually paying for.
Signal is a small consulting practice. Vendor-neutral, light-touch, done inside the software clients already pay for — the Microsoft or Google stack, Slack or Teams, the CRM. Less is more. Adoption is the bottleneck; learning one new tool is enough.
— How it runs
Three phases, in order.
Phase
What it is
Window
01 — Discovery
Time with the actual work — the spreadsheets, inboxes, slow Tuesdays. Ends in a plain-English memo: what's worth automating, what isn't, what to build first.
~2 weeks
02 — Build
Set up the LLM provider. Wire it into the tools the team already uses. Write the skills the work actually needs. On-site onboarding.
4–8 weeks
03 — Retainer
Monthly, ongoing. New skills as the business changes, onboarding for new hires, office hours. Most relationships last more than a year.
Ongoing
— Positioning
Quiet, considered, vendor-neutral.
- Provider-agnostic. “Your LLM provider,” not a specific brand. The right model for the job, swapped without ceremony.
- Inside the existing stack. No new platforms to roll out, no parallel system to maintain. Work where the team already is.
- Editorial restraint. Real implementations, plainly explained. No demo theatre, no decorative dashboards.
- Any size. The shape of the engagement scales; the principles don't change.
Wondering what's real for you?
Discovery is the cheapest part — two weeks, a memo, no commitment to build. The easiest place to start is a short conversation about what the work actually looks like.