Instill
Vol IV · MMXXVI · The method

Two stories,
one thread.

Your team hires the work. An agent runs it. A named expert stands behind it.

№ iThe wedge

Why an agent alone
isn't enough.

Generalist AI

  • Scale, no craft.
  • Trains on the crowd.

A consultant

  • Craft, no scale.
  • Trains on a career.

Instill

  • Agent + expert, on call.
  • Trains on one practice.
№ iiThe cast · three voices, one trip

Three voices,
one trip.

№ i.

The buyer

  • Arrives with a problem.
  • Asks the expert's agent.
  • Leaves with a signed note.
№ ii.

The agent

  • Runs the long arc of the work.
  • Knows its own edge.
  • Defers when review is needed.
№ iii.

The expert

  • On call, not on stage.
  • Verifies in minutes, not days.
  • Their voice. Their bar.
AI stack budgeting · session 4f2a
Live
The team
$40K a month across a dozen AI tools — the board wants it justified before renewal.
◎ The agent
Scoring your stack against Kanchi's rubric — overlap, usage, and attribution.
◌Kanchi flagged for verification · pre-briefed by her agent
The expert · verified
Cut the two redundant tools and consolidate analytics — ~30% of spend, no pipeline impact. Budget memo attached.
■ Budget memo · delivered · verified K.G.Ledger ✓
№ iiiThree axioms of the practice

How the practice
runs.

i.

The agent is always first. The human is always optional — until they're not.

ii.

The agent knows its own edge. It pauses, names the uncertainty, and pings the expert.

iii.

Every artifact carries provenance. Agent-authored, expert-verified, expert-revised — always labeled.

№ ivThe record · the trail you keep

Built to
show your work.

Named

Every agent carries one expert's methodology — signed, not anonymous.

Verified

Published only after passing its author's own simulated engagements.

Audited

Every question, consent, and escalation in an append-only ledger you can export.

Engagement ledger · append-only
export · csv
14:02consent.grantedTool list & spend · this engagement only
14:04consent.declinedCustomer data — never accessed
14:05finding.createdRedundant tools (3 overlapping)
14:11flagged.for_review→ Kanchi · pre-briefed
14:16verification.resolvedVerified by the expert, in thread
14:18deliverable.verifiedBudget memo v1 · K.G.
6 entries · nothing edited, nothing deletedVerified ✓
Where your data goes, what's consented, what's exportable
№ vFor experts · Cohort I, by review

Your craft,
compounded.

I.

Author it once.

A 90-minute Studio conversation. Your methodology — the questions you ask, the calls you make — drafted as an agent in your voice.

II.

Verified before it ships.

You run it against simulated clients until it meets your bar. Nothing publishes until it does — your name is on it.

III.

It earns across accounts.

Sessions, deliverables, escalations — your agent works many engagements at once. You're pinged only when your review is the deliverable.

Begin where you are

Vol IV is open
by invitation.

Talk to us about a pilotApply for Cohort I