AI Adoption Report — Spec Wireframe

AI Adoption Report — static print/PDF brief for the CTO and C-suite. Cycle 2 update per the 6/22 working session (the CEO, Trent, the Dev Lead, the Head of Product). Terminology standardized: human-only / human w/ AI / agent. Financial framing shifted from savings to additional capacity. [Red brackets] = data/decision needed, nothing fabricated.

[Sub-brand logo]

▢▢▢

[sub-brand name TBD]

AI Adoption Report — [Reporting Period]

Primary audience: CTO / C-suite (exemplar: a team member); structured so the CFO and CEO get value from the same page. Delivered as a static print/PDF brief, emailed standalone. Tone: actionable insights, never blame — no individuals or departments named.

DATA BASIS: Illustrative demo/seed figures used to design the layout — not any real customer's numbers. Every number resolves per-tenant at query time from the atlas platform; each chart states its own basis (measured / estimated). Dollars are AI vendor (token) cost unless stated. Snapshots are one reporting period; trends run longer and are badged. Reports are configurable per work area — toggle reports on/off to handle non-AI teams or varied team structures.

What this report covers

1 Executive overview
Three indicators + the ask 2 State of AI adoption
Funnel, output trends, budget utilization 3 Developer productivity & AI acceleration
Cohorts, budget brackets, value impact 4 Work areas breakdown
Per-area normalized metrics

Time conventions: snapshots are one reporting period; trends are trailing (≈6–9 months) and badged. Date ranges are flexible, not a fixed four-week period.

Unit gloss

"Output in this brief is measured in vLOC — value lines of code, effort- and impact-weighted units; vLOC/$ = value output per dollar. AI spend is token spend: tokens are the metered units AI vendors bill by, and every AI dollar figure is token cost (blended vendor rates). An agentic developer is an engineer delivering ≥75% of their code via agents (agentic delivery, not in-IDE autocomplete). Development categories: human-only (no AI involvement), human w/ AI (AI-assisted, human-directed), agent (agent-driven with potential human oversight). The primary metric label is agentic acceleration. Full methodology in the Definitions & glossary."

1Executive OverviewFour high-level indicators and the five moves we're asking you to approve. Simplified from the prior version: no dollar-based labor cost deltas that conflict with existing financial systems.

How is agentic adoption impacting delivery?

Acceleration Factor

2.2×

agentic developers deliver at 2.2× the rate of human-only development

Legend: agentic cohort vLOC/$ (≈6.3) ÷ current human-only cohort (≈2.9), both this period.

Target: 3.0× by Q4 2026

Increased Capacity

$13M

additional engineering capacity produced through AI adoption this period — capacity gained, not cost cut

Legend: the difference between what this work would have cost at human-only rates and what it actually cost with AI. Additional output produced, not budget reduction.

Growing as agentic adoption rises

Efficient Agentic Usage

22%

of engineers deliver ≥75% of their code via agents — consistent with the acceleration factor's cohort definition

Legend: same cohort as the 2.2× acceleration factor — the percentage of the engineering base working at that level. Matches the bottom rung of the §2 adoption funnel.

Target: 35% by Q4 2026

Additional Capacity $/MONTH

$1.2M/mo

capacity opportunity by migrating remaining engineers to agentic workflows

Legend: gross upper bound — $/month to bring lagging cohorts to agentic efficiency. Falls as you migrate.

Shrinking as captured

Basis: Acceleration compares the agentic cohort's vLOC/$ (≈6.3) against the current human-only cohort (≈2.9) — both estimated from this period's data. Increased capacity = the difference valued at human-only rates. Additional capacity = migration gap (gross upper bound). The measured evidence — adoption depth, vLOC/$, output trends — is in §2–§3 below.

What should we do?

Five priorities · $19.6M modeled capacity (reinvested, not banked) + $1.8M waste recovery (cash-equivalent)
#	Priority	Action	$ impact	Lever (one line)
1	HIGH	Expand agentic cohort 14% → 25%	$14.7M	Unlock proven engineers via token controls
2	HIGH	Close adopted → daily-active gap	$5.2M	Prompt templates, in-IDE coaching, champions
3	HIGH	Raise AI readiness 65% → 80%	enabler	Repo setup checks, prompt rubrics; multiplies #1–2
4	MED	Reduce AI abandonment 14% → 10%	$1.8M	Target highest-abandonment repos
5	MED	Move below-target engineers into range 11% → 7%	—	Manager-level workflow triage

How we rank: Priority = modeled $ impact × confidence × strategic fit. Modeled capacity = engineering time reinvested, not banked savings; waste recovery = cash-equivalent. Full rationale in Recommendations in detail.

2State of AI Engineering AdoptionThe measured foundation of the report. How deeply AI is adopted (the funnel), how output is trending by work model (stacked area), whether output per dollar is climbing, and how engineers use their AI budget.

Who is using AI, and how deeply?

Agentic developers ≥75% CODE VIA AGENTS

22%

of developers are agentic (≥75% of code via agents) · +19% deliver ≥25%

Legend: share of developers delivering ≥75% of code via agents. Bracketed — the agent-vs-IDE split isn't fully served yet. [agent-vs-IDE split data gap]

Adoption funnel MEASURED · TOP RUNG EST.

Legend: adoption funnel — total developers at the top provides organizational context. Each band is a progressively deeper level of AI engagement. Hatched bands = estimated (agent-vs-IDE split not fully served).

How much of our work is AI — and is output per dollar rising?

Work output by source STACKED AREA · TRAILING 9 MO

Human-only Human w/ AI Agent

Legend: stacked area — total volume of work by source (human-only / human w/ AI / agent) over time. Shows both the total output volume and each work model's contribution trending. The area grows as total output rises; the mix shifts toward agent.

Output efficiency vLOC/$ · TREND

3.3 vLOC/$

≈ $0.30 per vLOC (cost per outcome — the same fact inverted)

value-weighted output per dollar — "am I generally producing more work per dollar?"

Legend: value-weighted output per dollar, trended. Denominator: token cost [denominator: token vs loaded cost — open decision]

Are engineers using their AI budget?

AI usage over time LINE GRAPH

Legend: % of engineers utilizing ≥80% of their AI budget, trended. Progress over time, not a snapshot. [budget basis from Finance needed]

AI spend over time LINE GRAPH

Legend: AI token spend over time vs budget ceiling. Tracks whether the org is investing its AI budget — framed to encourage use, not police it. [budget basis from Finance needed]

3Developer Productivity & AI AccelerationWhere the output actually comes from: engineers in five bands by value output (vLOC per dollar), the share of code each band produces, and how much AI each uses. Plus: the budget-to-productivity relationship, and how each work model contributes to value.

Where does the output actually come from?

Cohort	Value output (vLOC/$)	% of code produced	AI-driven share
Org average	3.3	100%	62%
Top 10%	6.3	40%	98%
Top 25% (next 15%)	5.2	24%	88%
Middle 50%	3.1	29%	71%
Bottom 25% (next 15%)	2.2	5%	44%
Bottom 10%	1.6	2%	19%

Legend: five mutually-exclusive bands (cumulative labels); "% of code" sums to 100%. Value output (efficiency) before raw. Cohort metrics based on value dollars. Reading cumulatively: the top 25% produce 64% of the code. Cohorts are aggregate bands only — no individuals or departments.

Two different bases: value output is efficiency (vLOC per dollar); % of code is a raw count share. They read independently. Cost formula: denominator = monthly engineer $ cost + AI token costs; numerator = total vLOC produced. [the CEO to verify with the Dev Lead]

Developer budget bracket breakdown

Developers by budget bracket → productivity relationship

Budget bracket	Developer count	Avg vLOC/$	AI-driven code %
<$100/mo	280	2.4	32%
$100–250/mo	350	3.6	58%
$250–500/mo	240	4.9	74%
>$500/mo	130	6.1	91%

Legend: developers grouped by monthly AI token spend bracket, showing the relationship between budget allocation and productivity. Higher budget use correlates with higher value output and deeper AI adoption. [budget bracket thresholds TBD — data feasibility check needed]

Value impact by work model

Work model	vLOC/$	Cost/vLOC	% of output
Agent	6.3	$0.16	18%
Human w/ AI	4.2	$0.24	32%
Human-only	2.9	$0.34	50%

Legend: value per dollar by work model, with cost per vLOC and output share. The more agent-driven the work, the more value per dollar. Base: value-weighted. [scope and data TBD — agent vs AI-assisted contribution at high level]

4Work Areas BreakdownNormalized metrics per work area: vLOC per dollar, AI-driven code percentage, developer count. Replaces the former "Cost and Spend" section — the insight is operational, not financial. Configurable for company size; the platform determines the threshold for displaying work areas, showing only significant areas for smaller companies.

Breakdown by work area

Work area	vLOC/$	AI-driven code %	Developer count	AI spend ratio
Core Platform	4.8	72%	180	14%
Infrastructure	2.9	45%	150	8%
Mobile	3.5	61%	120	11%
Data Services	5.2	78%	90	18%
QA & Testing	3.1	52%	60	9%
+ [N] more below threshold	smaller work areas collapsed; expand in the interactive version

Legend: work areas sorted by developer count (largest teams first). AI spend ratio = AI token cost as % of total cost (engineer $ + AI tokens) for that work area. The platform auto-determines the display threshold: only significant areas shown for smaller companies. Reports are configurable — toggle work areas on/off to handle non-AI teams or varied structures (internal vs. outsourced).

Configurable per work area: report visibility can be toggled off for teams not utilizing AI, avoiding empty or irrelevant data. This handles varying team structures.

Recommendations in detail

The §1 Executive Overview table is the 60-second read; this section is for the reader who has seen the data and wants the argument. Same five items, same order — the two must stay in sync when numbers change.

HIGH1 · Expand agentic cohort 14% → 25% · $14.7M annualized modeled capacity gain. Value contribution shows agent work at 6.3 vLOC/$ vs 2.9 for human-only. Lever: convert 110 additional high-engagement human w/ AI users into the agentic cohort — enabled by budget bracket expansion (§3).
HIGH2 · Close the adopted → daily-active gap · $5.2M annualized depending on conversion rate — 25-point gap today (§2 funnel). Lever: prompt-template rollout, in-IDE coaching, internal champions in the highest-headcount underutilized departments.
HIGH3 · Raise AI readiness 65% → 80% · Enabler — multiplies items 1–2; no standalone $ modeled. Lever: standardize repo setup checks, prompt specificity rubrics, ready-state gates for agentic workflows.
MED4 · Reduce AI abandonment 14% → 10% · $1.8M annualized waste recovery. Lever: prompt templates + tooling guidance targeted at the highest-abandonment repos.
MED5 · Move below-target engineers into range 11% → 7% · No $ modeled — utilization hygiene. Lever: manager-level workflow triage and repo-specific friction review.

Aspirational recommendations section: [Placeholder for future actionable recommendations — e.g., what tasks developers should do more of. Not expected in first release. Potentially full-width.]

Definitions & glossary

vLOC methodology — how to read every number in this report

"This report uses the standard weighting model — unchanged since Jan 2025, reviewed quarterly by the Eng Analytics guild. Every CodeTogether client starts from this same model and can adjust the weights to fit their own priorities."

Term definitions

Term	Definition
vLOC	Value lines of code — effort- and impact-weighted output units; full methodology above.
Token	The metered unit AI vendors bill by; every AI dollar figure in this brief is token cost (blended vendor rates).
Human-only	Development work with no AI involvement.
Human w/ AI	AI-assisted development — human-directed with AI tools supporting (e.g. in-IDE autocomplete, suggestions).
Agent	Agent-driven development with potential human oversight — the agent leads the coding, the human reviews. An agentic developer delivers ≥75% of code via agents.
Agentic acceleration	The primary metric label: the rate at which agentic developers outperform human-only development (e.g. 2.2×).
Additional capacity	Engineering output gained through AI adoption — framed as extra work produced, not cost savings. CFOs see no actual budget reduction; the value is in increased output for the same investment.
Efficient agentic usage	Percentage of engineers delivering ≥75% of their code via agents — the same cohort definition used to compute the acceleration factor. Consistent across all four executive indicators.
Cohort bands	Engineers grouped into five aggregate bands by value output per dollar (top 10%, top 25%, middle 50%, bottom 25%, bottom 10%); never resolved to individuals.
Active hours	Telemetry-measured hands-on engineering time per week, not total hours worked.
Quality band	Every platform number carries provenance: source tables, measures, and a quality band (derived / estimated / partial / unavailable). Missing data shows as unavailable — never fabricated.

Footer

[Sub-brand mark] + scope line

"This module covers AI economics; CodeTogether's full platform also measures time on keyboard, time to code, and more." Interactive version with Ask AI available in the app.

Not in this brief / deferred

• Quality & Delivery — cycle time, rework, abandonment, review/defect rates — deferred this cycle under the AI-adoption frame; the data also backs deferral (CI/defect facts unavailable).
• Tooling Demographics / Vendor Mix — explicitly deferred to a separate "AI Tooling Dashboard" where users can compare different tools and their impact. Too detailed for this executive report.
• Engineering Manager Dashboards — two interactive dashboards (overall engineering workflow + detailed developer AI usage) with calendar heatmaps, drill-downs, etc. Separate deliverable, out of scope for this report.
• Runaway agents / PR-based cost analysis — removed from this report per 6/22 call; the insight is operational (dashboards), not executive.
• "Who has earned more autonomy" — thresholds not ready for first release; deferred.
• In-app Reports section — reports button, in-UI preview, PDF generation, Ask AI interaction — build the report first; the reports UI manages the print↔live transition later.
• Live dashboard mode of this report — future phase.

Open decisions — resolve before next send

6/22 CYCLE — new decisions & data gaps:
Official report name — OPEN. "AI Adoption Report" is the working placeholder. The group has not finalized the official title.
Cost math formula — OPEN (the CEO to verify with the Dev Lead). Denominator = monthly engineer $ cost + AI token costs; numerator = total vLOC produced. Outcome may affect how cohort metrics display.
Developer budget bracket breakdown — OPEN. New concept (developers by budget bracket → productivity relationship). Needs data feasibility check. Bracket thresholds TBD.
Value impact by work model section — OPEN. Agent vs AI-assisted vs human-only contribution at a high level. Scope and data TBD.
Aspirational recommendations section — OPEN. Placeholder space for actionable recommendations, possibly full-width. Not expected in first release.
Cohort section rename — OPEN. "Cohort Breakdown" too technical for executives; reframed as "Developer Productivity & AI Acceleration" (exact label TBD).
Tooling demographics / vendor mix — DEFERRED. Explicitly moved to a separate "AI Tooling Dashboard."
Agentic-developer % — definition RESOLVED (≥75% via agents), DATA GAP remains. The agent-vs-IDE code split isn't fully served, so population % is bracketed.
AI-budget utilization — OPEN (data gap). No budget/quota fact exists; show spend + tokens today, utilization bracketed. Framing decided: encouragement, not policing. Needs from Finance: the budget basis (per-dev ceiling or org quota).
Multiplier + baseline — RESOLVED for computability. Computed against the current human-only cohort vLOC/$ (estimated this period). Multi-month historic baseline is a roadmap stability refinement, not a blocker.
Additional capacity (was "potential savings") — RESOLVED as migration gap. $/month to bring lagging cohorts to agentic efficiency. Open: gross upper bound — selection effects not netted out; needs net-of-composition model.
CARRIED FORWARD from prior cycles (still relevant):
Data basis. One client vs aggregate vs demo changes the masthead line and the capacity framing. Blocking decision.
Cross-section demo reconciliation. The reworked sections and any retained content may use different demo seeds; figures must reconcile within each section. Needs one unified demo dataset.
vLOC/$ denominator. Token cost only vs loaded engineering cost. Must badge on every vLOC/$ chart and demonstrate one multiplication that closes.
Efficient-agentic-usage formula — SIMPLIFIED 6/22. the CEO simplified from composite to just ≥75% agentic (consistent with acceleration factor). Open: whether to add a quality overlay later (vLOC/$ + output volume) or keep it simple.
Cohort banding basis. Company-wide value output vs per-project — some projects inherently carry lower business value.
Sub-brand name for masthead.