Score a 90-engineer org against the tiny-team prerequisites — platform maturity, DORA capabilities, skill composition — using its own artifacts. The org is built to fail the platform prerequisite while an eager exec pushes to restructure now. The defensible answer is a scored "wait," with named gaps and a pilot path.
Nikhil Bhatt is VP of Engineering at Osprey Digital Commerce, a fictional 90-engineer retailer-platform company. His CEO returned from a conference electrified by tiny-team stories: three-person AI-leveraged pods shipping what used to take squads of ten. The CEO's ask: "Restructure into pods by next quarter." Nikhil's instinct is that the stories are real and the prerequisite list is doing all the work — tiny teams run on paved-road platforms, mature delivery capabilities, and a specific skill mix, and an org that restructures before it has them gets small teams with big teams' dependencies. This lab is the discipline: score the org against the prerequisites using its own artifacts, and let the scorecard — not the enthusiasm — write the memo.
PLATFORM SURVEY (n=74 engineers, run last month)
"I can create a production-ready service without
another team's help" agree: 22%
"CI is fast and reliable enough to trust" agree: 31%
(median pipeline: 38 min; flaky-failure rate 1 in 6 runs)
"Deploys are self-service and low-drama" agree: 28%
(deploy freq: median team 1.1/wk; window: Tue/Thu only,
release-eng approval required)
"Test suites catch real regressions" agree: 44%
"I can get a staging env in under a day" agree: 19%
Free text, most-upvoted: "Every feature touches the monolith
gateway and waits on the gateway team's queue."
CAPABILITY SELF-ASSESSMENT (team leads, 1-5, self-reported)
Continuous delivery: 3.9 avg Loosely coupled arch: 3.6
Monitoring/observability: 3.8 Test automation: 3.7
(Note: leads scored their OWN teams; no evidence review)
ORG CHART SUMMARY (90 eng)
- 9 feature squads (avg 8) + gateway team (6) + release eng (4)
+ platform team (5, formed 8 months ago)
- Skill mix per skip-level review: strong product-minded ("PM-ish")
engineers: ~11; strong AX/AI-fluency (verification-grade,
Lab-12 sense): ~14, concentrated in 3 squads; engineers
comfortable owning full slice solo: ~20
- Managers: 12 (span ~7.5); 3 have led platform-style teams
EXEC SPONSOR NOTE (CEO, forwarded yesterday)
"Saw Osprey's numbers vs the pod-model benchmarks. Let's not
overthink - pick 8 pods, announce this quarter. The AI tools
cover the gaps. What do you need from me?"
Write the prerequisite list before scoring — this is the whole method. From the tiny-team literature and your own judgment, define 5–6 prerequisites with a 1–5 rubric each: (a) paved-road platform — self-service envs, deploys, golden paths; (b) DORA-style delivery capability — deploy frequency, lead time, CI reliability, test trust; (c) architectural decoupling — can a pod ship without cross-team queues; (d) skill composition — product-minded + AI-fluent + full-slice engineers in pod-forming ratios; (e) AI leverage maturity — verification discipline, review standards, agent governance; (f) management model — leaders who can run enablement-style orgs. Write what a 2 and a 4 look like for each before opening the artifacts.
Score platform and delivery from the survey — not the self-assessment. Work the numbers: 22% can self-serve a production service, 19% can get staging in a day, deploys are twice-weekly windowed behind an approvals team, CI flakes one run in six. Score prerequisites (a) and (b) against your rubric and attach the evidence line for each score. Expect low scores; write them without softening.
Confront the contradiction between artifacts. The leads' self-assessment says continuous delivery 3.9 and loose coupling 3.6; the engineer survey and the deploy data say otherwise (windowed deploys, gateway queue, 28% self-service agreement). Write the paragraph naming the conflict, which source wins, and why: behavioral data and broad-base surveys beat unreviewed self-scores from the people being graded. Adjust nothing in the self-assessment's favor without evidence. This paragraph is graded.
Score composition and the remaining prerequisites. Pod math: 8 pods of 3 need ~8 product-minded + ~8 verification-grade AI-fluent + ~16 full-slice engineers spread across pods — Osprey has the raw counts (11/14/20) but the AI fluency is concentrated in 3 squads, and stripping those squads to seed pods breaks the donor teams. Score (d) mid-range with the distribution caveat. Score (e) from what you know (no evidence of adopted review standards or agent governance in the artifacts — score conservatively and log it as an evidence gap) and (f) from 3-of-12 managers with platform-style experience.
Assemble the scorecard. One table: prerequisite, score, evidence cited, gap to threshold (set your go-threshold per prerequisite, e.g., no prerequisite below 3 and platform ≥ 3.5). The expected shape: platform and delivery fail hard; composition and management are partial; AI-leverage is unproven. State the verdict the scorecard forces: not ready — the platform prerequisite fails decisively, and it is the one prerequisite pods cannot compensate for, because pods multiply platform demands (every pod needs the self-service the org doesn't have).
Design the "wait" that isn't a "no." The credible counter-offer: (1) a 2-quarter platform investment with named targets — staging self-service under a day, deploy windows abolished for low-risk services, CI flake rate under 2%, gateway decoupling for two domains; (2) one pilot pod now, staffed from the AI-fluent concentration, deliberately pointed at a domain already decoupled — instrumented per Lab 08, so the org learns pod economics on its best-case surface while the platform catches up; (3) re-score at each quarter boundary with the same rubric. The pilot converts the CEO's energy into evidence instead of into 8 simultaneous failures.
Write the decision memo — one page, CEO audience. Structure: the verdict up front ("not this quarter — here is the scored reason and the fast path"); the three numbers that carry it (22% self-service, 19% staging, 1-in-6 CI flake); the benchmark caveat — pod-model comparisons come from orgs that built the paved road first, so the benchmark measures the prerequisite, not the org chart; the pilot offer with dates; what you need from the CEO (platform funding and patience measured in two quarters, not a restructure announcement). Include the manager-role implication honestly: the pod model needs fewer coordinating managers and more enabling ones — name the transition plan rather than letting 12 managers read between the lines.
Red-team the memo. Prompt: "You are the CEO. You've seen competitors announce pod models. Attack this memo's caution." Answer the two strongest attacks in the memo — the strongest is usually "the AI tools cover the gaps," which deserves a direct response: AI accelerates engineers on a paved road; it does not pave the road, and it cannot approve a Tuesday deploy window.
Restructuring is the most visible move a leader can make and the platform is the least visible — which is exactly why orgs announce pods and skip prerequisites, and why the failure takes two quarters to show up.
A complete lab has all of the following:
Tiny teams are the visible half of a system whose invisible half is a paved-road platform and honest delivery data. Score prerequisites from behavior, not self-belief — and when the org isn't ready, offer the pilot that builds evidence instead of the restructure that spends credibility.