# Interview Guide

Question bank for every phase. Small batches (3 to 5 per turn), reflect answers back, never a form dump. Many executives will not know exact numbers. Always offer an estimate path and mark anything estimated so the deliverables can label it.

## The "I don't know" rule

For every quantitative question, give three ways to answer: (a) the exact figure, (b) a range or estimate, (c) "pull it from public data." If they choose (c) and the company is public, go find it. If nothing is available, ask for a rough estimate and tag the value `estimated: true` in the data. Never block progress on a number the executive doesn't have.

## Phase 1: Company intake

Ask, in small batches:

1. A short background on the company in their own words. This narrative primes every later answer; capture it verbatim.
2. Website URL.
3. Total employees (headcount).
4. Employee categories and rough counts: by function (engineering, sales, ops, G&A), by type (full-time, part-time), by level (IC, manager, exec). Whatever split they think in.
5. Number of contractors / contingent workers, and roughly what they do.
6. Locations and headquarters (needed later for salary benchmarking).
7. Anything unusual about how the company is structured that a generic org chart would miss.

## Phase 2: Web research and public-data enrichment

Using the website and the company name, with the session's web-search and web-fetch tools plus any connected data sources (Claude connectors such as a CRM, document drive, or news tools):

1. Determine public vs. private. Look for investor relations pages, ticker symbols, SEC/EDGAR filings, stock listings.
2. If public: locate the latest 10-K (or annual report) and pull revenue, gross margin, operating/net profit, and employee count. Pull the same metrics for as many of the past 20 years as are available (10-K filings, macrotrends-style history, annual reports).
3. If private: look for reported revenue estimates, employee counts on the company's own site and public profiles, funding announcements, press. Be honest about confidence.
4. Pre-fill everything you find and show the executive the source and the number for confirmation. Mark each value as `sourced` (with a citation) or `estimated`.
5. For anything you cannot find, say so explicitly: tell the executive this information could not be found online (common for young companies, private companies, or thin coverage) and that they need to supply it. Then ask directly, including how the metric has changed over time (even three or four remembered data points make the history chart meaningful).

Tag every data point with its provenance. The company-overview deliverable renders sourced and estimated values differently, and the executive must be able to see which is which.

## Phase 3: AI literacy, implementation, and velocity

The answers here set how fast the org is modeled to spin up agents. Ask:

1. Board and executive AI literacy: who personally uses AI tools, who can evaluate a vendor claim, who advocates. (Evidence, not self-report; surveys overreport.)
2. Workforce literacy: is there training, what percentage actively use AI weekly, where is the shadow usage.
3. Current AI implementation: what is deployed, piloted, announced. What data exists and whether it is clean, structured, and accessible.
4. Whether an AI strategy exists on paper and in budget, and who owns it.

Synthesize these into an **agent-spin-up velocity**: Slow, Moderate, or Fast, with a one-paragraph rationale. This velocity is the growth-rate input to the node-mix projection. If the executive has run the other skills in this collection (AIDE Opportunities Audit, AIDE Strategic Framework), pull those readiness scores and reuse them rather than re-asking.

## Phase 4: Regulatory and risk

1. Which regulatory regimes govern the business (sector rules, privacy, sector-specific AI rules).
2. Data sensitivity and customer-facing exposure.
3. Existing governance: is there an approval path for new technology, an AI use policy, a risk committee.

State a risk posture (Low / Moderate / High) with rationale. High-risk postures slow modeled agent adoption and raise the bar on the governance relationships in the org chart.

## Phase 5: The human org chart

Build the current, all-human org chart:

1. Interview for the structure top-down: CEO, their direct reports, each of those leaders' teams, down to the level of granularity the executive can sustain. Capture role title, name or placeholder, function, level, and reporting line (each node's manager).
2. Cross-check with public sources: leadership pages, LinkedIn-style public org info, proxy statements for named executives. Fill and confirm.
3. Introduce the language shift explicitly, in conversation and in the deliverable: from here forward, employees are **human nodes**, and they will sit alongside **AI agent nodes**. This is not cosmetic. It reframes the org as a network of capacity, human and machine.

Render the human org chart in the draggable canvas. Ask the executive: does this match your expectations? Loop on revisions until they confirm.

## Phase 6: Deploying a single agent

Pick one concrete agent to deploy (draw from their strategy work or propose one grounded in their profile). Ask, and capture as relationships to specific human nodes:

1. **Reports to:** which human node does the agent report to operationally.
2. **Accountable:** who is accountable for the agent's outcomes (the single owner).
3. **Consulted:** who is consulted on changes to the agent.
4. **Informed:** who is informed about what the agent does.
5. **Governance:** who holds governance for the agent within the org chart.
6. **Spends whose budget:** when the agent spends money, whose budget line does it draw from.
7. **Funds the agent (tokens/inference):** who is responsible for the agent's own running cost, separate from what it spends. This is often a different owner than the budget it spends.
8. **Launch approval:** who approves promoting this agent into production.

Render the second org chart: the same human structure with the agent node added, given a bright neon-blue background, and a labeled reporting line for each of the eight relationships. Ask the executive to review and revise.

## Phase 7: Enriched multi-agent org chart

Using everything known about the organization (this interview plus any prior skills), auto-populate where agents would plausibly exist across the whole org chart from Phase 5. For each proposed agent: which human node it attaches to, its function, and its primary relationship lines. Render as a third org chart showing all agents and their reporting relationships. Present it as a proposal, not a prescription. Ask the executive to review and revise; take their edits.

## Phase 8: Reimagine from the ground up (greenfield)

Ask the executive to set aside how the company is organized today. Given what they now know about agents, agent orchestration, and scaffolding, how would they rebuild the company as an AI-native organization? Ask about:

1. The shape: is it still a pyramid, or flatter, or a mesh, or a small human core orchestrating many agents.
2. Reporting lines: what reports to what, and how many layers.
3. Number of nodes: how many human nodes, how many agent nodes, and the ratio.
4. What the humans do that no agent does.

Render the greenfield org chart. This is the aspirational endpoint the projection curve bends toward.

## Data provenance discipline (applies throughout)

Every figure carries a provenance tag: `sourced` (with citation text), `estimated` (by the executive), or `modeled` (computed by the skill). The deliverables surface these tags. The point is intellectual honesty: an executive taking this to their board must be able to say exactly which numbers are real and which are projections.