docs: [verified] answer the six-month AI Architect assignment
Build and publish Eldorado documentation / build (pull_request) Successful in 1m30s

This commit is contained in:
jarvis-at-skic committed 2026-10-08 16:59:36 +00:00
1 parent 88a4375776
commit 1a2d16bc91
53 files changed
+974 -390

No files matched your search

+58 -18
View File
@@ -1,8 +1,8 @@
---
id: target-operating-model
title: Target operating model
sidebar_label: Target
description: An autonomous agent executes and governs the SDLC; humans direct, guide and review the repeating cycle.
title: First six months as AI Architect
sidebar_label: Six-month strategy
description: Improve real engineering output through a solo, CTO-sponsored, team-partnered six-month learning plan.
hide_table_of_contents: true
---
@@ -10,28 +10,68 @@ import ArchifyDiagram from '@site/src/components/ArchifyDiagram';
<div className="strategy-content">
**TO-BE · Target operating model — the intended strategy, not a deployed system.**
**Proposed strategy · No company results or organizational commitments are claimed.**
## Autonomous SDLC, human direction
## Improve the work, not the AI usage
<ArchifyDiagram name="strategy-target" title="Autonomous agent executes and governs the SDLC; humans direct, guide and review" story />
<ArchifyDiagram name="strategy-six-months" title="Proposed six months: discover with teams, test one bottleneck, retain only supported value" story />
An autonomous agent executes and governs every phase. Humans direct, guide and review it; a separate human-controlled AI Agent contributes to each phase from outside the cycle.
The assignment describes close to **80 engineers, growing to around 100 by year end, across more than 10 teams**. Every developer already uses AI, unevenly. I start **solo, reporting to the CTO, with no formal authority over teams**. The comparison is current AI-assisted practice—not an imagined no-AI organization. These facts come from *AI Architect Test Task*, Context and question 1; everything below is a proposal.
**Solid arrows:** lifecycle · **Dotted spokes:** execution and governance · **Dashed arrows:** contributions.
My first investment is time alongside developers and reviewers: find where useful changes stall, then remove one evidenced constraint. Delivery is a starting hypothesis because a bounded change, tests and independent review can expose both benefit and transferred work. Discovery may show that flaky CI, unclear requirements or review capacity matters more than agent instructions. Fix that bottleneck instead.
**Scope → Architecture → UI/UX → Planning → Delivery → UAT → Security → Releases → Scope**
The [month-by-month plan](./roadmap.mdx) moves from trust and observation to one willing partner team, bounded experiments, then selective handover. I offer pairing and show verified examples rather than mandate tools. The CTO sponsors access and resolves priority conflicts; team leads agree contributions. Neither sponsorship nor this document conscripts a team.
The arrows show the intended SDLC cycle, not the order of skills rollout. Releases returns evidence to Scope; human direction informs the next iteration. The [Roadmap](./roadmap.mdx) starts with human + AI Agent work in Delivery before considering wider autonomous authority.
## Choose the smallest intervention that fits
**Illustrative Delivery thread:** an agent prepares a bounded code change and regression evidence, an independent human reviews the exact artifact, and existing release controls decide whether it can advance. Stage 01 makes this working practice dependable; Stage 02 separately considers what the central agent may execute. Skills supply knowledge and verification steps, never permission.
Missing repository knowledge suggests better context; inconsistent checking suggests coaching or a review checklist; repeatable mechanical work suggests deterministic automation; unreliable tests suggest test/CI repair. A **skill** packages reusable instructions, prerequisites and verification steps when repeated agent-guided work genuinely needs them. It is one option, not permission and not proof of correctness. Compare alternatives against the same bottleneck before funding a catalog or control platform.
Use the [worked experiment](./stage-01-skills-adoption.mdx) to connect a hypothesis, small initial measures, safety checks and a retain/change/stop decision. Existing review, data restrictions and release authority stay in place. The [16 metric contracts](./measurement/index.mdx) and skills references are optional depth, not startup bureaucracy.
## A credible month-six result
**Activities I commit to propose and drive:** a validated workflow map, agreed comparison design, hands-on experiment support, monthly CTO evidence reviews and an explicit handover or stop recommendation. Delivery depends on agreed access and partner capacity; blocked work is reported, not disguised as progress.
**Contingent visible outcome:** one workflow retained by one partner team, with a second willing team trying a bounded transfer only if value and support hold. Developers can reproduce the example; reviewers see better evidence without extra hidden burden; a voluntary champion and backup can maintain it. Evidence must include comparable task records, independent acceptance, matured defect follow-up, review/rework effort, setup and upkeep costs, and a decision signed by the actual owner. These are required evidence types, not results already obtained or promised percentage gains. If the pilot fails, month six should instead show what was stopped, why, and which cheaper intervention replaced it. Neither scenario means rollout to all 80–100 engineers.
## Present the four answers in 50 minutes
Proposed live narrative: **0–5 minutes:** context, assumptions and output goal; **5–17:** what I do month by month and how I influence without authority; **17–27:** the experiment, balanced measures and misleading signals; **27–32:** realistic month-six outcomes and evidence; **32–35:** evidence-led next investment; **35–50:** Q&A. This is **35 minutes content + 15 minutes Q&A**, not an assignment-imposed split. Content matters more than visual polish. Walk the main story, then open references only for questions.
Beyond six months, choose deeper improvement, careful replication, additional enablement capacity or stopping an ineffective approach. The [future options](./stage-02-autonomous-agent.mdx) include bounded autonomy only where a new business case and readiness justify it; autonomous SDLC is not the destination promised by this plan.
<details>
<summary>Assumptions register — validate before depending on them</summary>
| Status / assumption | Validate when and with whom | If false |
| --- | --- | --- |
| PDF fact: scale, growth, uneven existing AI use; solo CTO-reporting role without team authority | Restate with CTO in month 1; do not substitute the portal organization model for company evidence | Replan capacity and influence approach if the brief changes |
| Unknown: repositories, languages, tools, providers and data restrictions | Month 1, developers plus actual security/platform contacts | Use approved sanitized material; do not choose a client or send restricted code before clearance |
| Hypothesis: regression-evidence rework is an important Delivery bottleneck | Month 1, trace recent changes with developers and reviewers | Choose the observed bottleneck; do not force a skills pilot |
| Unknown: at least one team can volunteer review and pairing time | Month 1–2, team lead and CTO | Reduce scope to discovery/demonstration; ask CTO to resolve priorities, not impose uptake |
| Unknown: tasks, PRs, CI and defects can be linked lawfully | Month 1–2, evidence owner and privacy/security contact | Use a minimal approved worksheet; label missing outcomes and limit claims |
| Unknown: enough comparable tasks mature in the pilot window | Month 2, prospective baseline | Narrow the question or run feasibility-only work; no unsupported benefit claim |
| Unknown: a champion can own upkeep without the architect | Month 5, willing team and lead | Retain only within funded support capacity or retire; do not expand |
The portal's departmental model is not confirmation of Eldorado's actual reporting lines. Reference-role names below are candidate functions to map to real people, not appointed authorities.
</details>
<details>
<summary>Conditional long-term Target operating model — not the six-month promise</summary>
<ArchifyDiagram name="strategy-target" title="Conditional proposed target: autonomous SDLC under human direction" story />
Only if separately justified, an autonomous agent could execute and coordinate the SDLC under human direction, guidance and review, with separate human-controlled AI Agent contributions. **Solid arrows:** lifecycle · **Dotted spokes:** execution and governance · **Dashed arrows:** contributions. Scope → Architecture → UI/UX → Planning → Delivery → UAT → Security → Releases → Scope. This topology is retained as an option, not an inevitable next stage.
</details>
<details>
<summary>Supporting implementation options: trust boundaries and recovery</summary>
## Controls that support autonomous execution
The lifecycle above is the strategic direction. The earlier control design below is a supporting implementation option, not an agreed assignment of approval roles or a claim of installed controls. Detailed approval policy and responsibilities remain to be defined with humans.
The earlier control design below is a conditional implementation option, not the six-month strategy, an agreed assignment of approval roles or a claim of installed controls. Detailed approval policy and responsibilities remain to be defined with humans.
1. Humans direct, guide and review; the autonomous agent executes and governs work across the lifecycle.
2. The reviewed registry supplies pinned skills and qualified adapters to coordinator and local clients.
@@ -143,19 +183,19 @@ See the existing [organization structure](/organization/structure/), [department
<details>
<summary>Strategy rationale and reading map</summary>
### Govern skills first, expand authority only with evidence
### References support the chosen intervention
Eldorado's target is to make AI-assisted work dependable across the organization: useful guidance, clear human ownership, controlled execution and measurable outcomes. The first investment is a governed skills system, not an autonomous replacement for the organization. A skill packages reusable working knowledge, prerequisites and verification steps. It helps an agent perform a task; it does not grant permission or guarantee that the work is correct.
The first investment is learning where current AI-assisted work can improve. Use skills governance only if a skills intervention is selected; it is not a prerequisite for discovery, coaching or fixing deterministic tooling.
The target combines an autonomous agent executing and governing the SDLC with humans directing, guiding and reviewing it. Developer choice on local laptops remains a supporting option: Copilot, Claude Code and Cline can receive reviewed portable skill content through qualified adapters. Repository, CI and release controls support execution; detailed approval policies remain implementation decisions, not new roles established by this strategy.
The conditional target combines an autonomous agent with human direction and review. Actual client support requires qualification; names such as Copilot, Claude Code and Cline are examples, not known company tools. Repository, CI and release controls remain authoritative.
The strategy describes the intended operating model and the evidence required to introduce it. It does not establish that these controls are installed, that every client is supported, or that AI has improved Eldorado's performance. The current reference is the documented six-department organization. Rollout begins by validating actual clients, responsibilities, data boundaries and available evidence.
This strategy does not establish that controls are installed or that AI has improved Eldorado's performance. Validate actual clients, responsibilities, data boundaries and available evidence; the documented six-department model is not source evidence for this assignment.
Success means accepted outcomes with maintained quality, controlled risk and understood total cost. Faster activity without reliable outcomes is not success. More skill invocations, generated lines or tool subscriptions are not productivity targets. People remain accountable for product, customer, financial and personnel decisions.
### Navigate the strategy
- [Roadmap](./roadmap.mdx) — Skills adoption first; future autonomy separately justified, no promised dates or budgets.
- [Roadmap](./roadmap.mdx) — Proposed six-month priorities, capacity, dependencies and decisions.
- [Stage 01](./stage-01-skills-adoption.mdx) — Follow one workflow through evidence, decisions and feedback.
- [Baseline](./baseline.mdx) — Understand current work and record the starting position before changing it.
- [Skills governance](./skills-governance.mdx) — Contract, lifecycle, decision rights, adapters, locking, exceptions.
@@ -167,6 +207,6 @@ Success means accepted outcomes with maintained quality, controlled risk and und
</details>
**Next:** [Follow the Roadmap toward Target →](./roadmap.mdx)
**Next:** [Follow the six-month plan →](./roadmap.mdx)
</div>