The AI-native
board process
Same trust. Same people. A fraction of the production work.
A proposed operating model for how Operators delivers advisory boards.
One board today is roughly 30 manual tasks
Reconstructed from your delivery playbook. Every task is done by hand, in sequence. Each new board adds the same work again, so capacity is capped by hands, and quality depends on who is busy that week.
Setup & info gathering
- Agreement + invoicing
- Investment memo, FAQ, value prop, role description
- Website, LinkedIn, mailing
Operator scouting
- Candidate search on LinkedIn
- Manual outreach process
Selection & formalization
- Screening & interviews
- Founder alignment
- Contracts & rejections
Launch
- Scheduling & kick-off prep
- Chair briefing
- Dashboard & tracking setup
Task counts from the current playbook. Hours per task: to be validated together.
From documents by hand to one source of truth
The memo, the FAQ, the value proposition, the role description and the LinkedIn post are five renderings of the same facts. Today each is written and updated separately. The proposal: capture the facts once, generate everything from them.
Every document is hand-made
Each case produces a stack of documents written one by one. When something changes, someone has to remember to chase the other four. Knowledge about the case lives in the documents, and in people's heads.
One Case Profile, everything generated
One structured profile is captured at intake. Agents generate every artifact from it. Change the profile, and everything regenerates consistently.
30 tasks become 6 human moments
Humans keep every moment where trust is built. Agents take everything in between.
Eleven tasks, two human moments
- The intake conversation it builds the client relationship; the typing afterwards does not
- One approval of the full document pack one review covers memo, FAQ, value prop, role description
- Agreement & invoice triggered on signature
- Full doc pack generated from the Case Profile
- Website listing, LinkedIn post & mailing published after approval
Five documents are five renderings of the same facts. Generating them together removes the inconsistency that comes from updating one and forgetting the rest.
The pack is external-facing, so a human approves once before anything leaves the building. Derivative content (website, post, mailing) is covered by that same approval.
The network becomes searchable, not memorable
- Network pre-vet the moment a case exists, the competency database surfaces matches from the Collective
- Candidate sourcing scored longlist from public profiles for the competencies the network lacks
- Outreach drafts personalised per candidate, from the Case Profile
- Approving the shortlist
- Sending the outreach, under a real name first contact with a senior operator is personal and reputational
The network is the moat, and today it lives in people's heads. That does not scale, and it walks out the door with people. A competency database makes the strongest asset searchable in seconds, for every case.
Sourcing is high-volume pattern-matching over public data: exactly what people are slow at and agents are fast at. A miss costs nothing: a weak candidate simply never reaches the shortlist.
Agents read every candidate. People pick them.
- Screening every applicant read the same way, scored with a written rationale
- Candidate correspondence & rejections drafted, one-click approve before sending
- Interview guide generated fresh per case, never stale
- Contracts assembled from approved templates, e-signature, chased automatically
- Prep briefs before every call, recaps & CRM updates after
- Interviews
- Introductions to the founder
- Founder alignment on the final board
People skim when there are forty applicants; an agent reads all forty with the same care. Coverage and consistency go up. But the final pick is judgment about people, and that judgment is the product.
Rejections are sensitive and external. A rejection that feels robotic burns a person you may want on the next board. Agents handle the volume; a human owns the tone.
Coordination disappears. Ceremony stays.
- Scheduling & participant emails
- Board dashboard instantiated the moment contracts are signed no longer a setup task
- Kick-off pack & chair briefing doc synthesised from everything already in the system
- Recaps & reporting digest after launch founder and operators stay in cadence without anyone writing status emails
- The chair briefing call
- Running the kick-off
Launch is split in two: coordination overhead with zero judgment content (scheduling, emails, dashboards) and authority moments (the chair, the kick-off). The first should cost nothing. The second is the board's backbone and stays fully human.
Automation Triage: how we decide what agents get
A pure ROI score gives wrong answers in a trust business: it would fully automate rejections and outreach, and damage the brand. So two tests come before any scoring.
The Relationship Test
Is a human's presence here part of what the client pays for? If yes, agents prep and follow up, the human performs. No score overrides this.
The Exposure Test
Is the output external-facing and hard to un-send? If yes, agents produce, and a human approves before it leaves the building.
The score, for everything else
priority = hours per case × boards per year × automatability ÷ build effort · filled in with real numbers from a walkthrough of a past case, not estimates.
The build-order rule
Build in the order the data flows, not by score. The Case Profile and the competency database enable nearly everything else, so they come first.
Judgment work. Agents absent or ambient notes only.
Agents prep and recap. The human performs the moment.
Agents produce. A human approves in one click.
Agents execute. Humans see a log, not a task.
Named agents, each owning one job
Every agent maps to a step in the board pipeline. Highlighted agents are the proposed first wave, closest to the Case Profile.
Time saved is the smallest of the four wins
Consistency becomes structural
Every case gets the same depth of screening and the same quality of documents, independent of who is busy that week. Quality stops depending on workload.
Institutional memory compounds
Every case enriches the competency database. Board number twenty is cheaper and better-matched than board number one. The manual process learns nothing.
Parallel capacity without headcount
Production no longer scales with hands. Several boards can run at once, and the team's time concentrates on the six moments clients actually pay for.
Hours back, every case
The production work between the human moments largely disappears from the team's calendar. Exact hours: measured together, not asserted.
Build in the order the data flows
- Hours per task today, per phase
- Boards delivered per year, and the target
- Where the data lives today (CRM, Notion, mail, heads)
- What the Collective data currently captures
The method is process-agnostic. The same triage applies to any workflow, not just boards. · PLYO Lab · 2026