Thinksy EngineeringEngineering Effectiveness Report

FY2027 · Feb 2026 – Jan 2027Data through Oct 2026

Director of Engineering view

Engineering Effectiveness

AI spend and forecast, delivery, adoption and friction across the engineering organization.

Demonstration report. Thinksy's engineering organization, people, spend and results are synthetic.

Period

Organization

Executive summary

FY2027 to date

FY2027 AI spend forecast

$85,657

5% under budget · $90,000 budget · 61% spent to date

AI spend, FY2027 to date

$55,031

Run-rate $8,531/month

At-bats shipped

116

39 won (34%)

AI spend per at-bat

$474

Lower means cheaper experiments

Active AI users, Oct

24 of 24

100% adoption

Key findings

Spend is forecast to finish FY2027 at $85,657, 5% under the $90,000 budget.
Opus's share of AI spend fell from 68% in the first three months to 52% in the last three.
116 at-bats shipped FY2027 to date (34% won), at $474 of AI spend per at-bat.
Harness practice gaps cost an estimated $15,109 (27% of spend) FY2027 to date, led by missed prompt-cache hits ($8,461).
Biggest friction point: test suite failure loops (Content Pipeline), 41 occurrences.

Spend and forecast

FY2027, cumulative · forecast to fiscal year-end

Cumulative AI spend, FY2027

Actuals through Oct 2026; forecast through Jan 2027 with an 80% range

Actual
Forecast
80% range
FY budget

FY2027 forecast by team

Fiscal year to date, forecast for the remaining 3 months, and variance to budget

Team	FY to date	Forecast, rest of year	FY total	FY budget	Variance
Jamie Park · Content & Media
Content Pipeline	$6,065	$3,629	$9,694	$10,200	−$506
Video Rendering	$8,923	$5,241	$14,163	$15,600	−$1,437
Social Publishing	$5,128	$3,104	$8,233	$7,200	+$1,033
Subtotal	$20,116	$11,974	$32,090	$33,000	−$910
Sam Okafor · Growth & Web
Growth & SEO	$5,607	$3,394	$9,002	$9,600	−$598
Web Platform	$6,739	$5,056	$11,795	$11,400	+$395
Data & Analytics	$5,968	$2,706	$8,674	$10,800	−$2,126
Subtotal	$18,314	$11,156	$29,470	$31,800	−$2,330
Robin Silva · Platform
Agent Platform	$9,130	$3,409	$12,539	$12,000	+$539
Infrastructure	$7,471	$4,088	$11,559	$13,200	−$1,641
Subtotal	$16,601	$7,497	$24,098	$25,200	−$1,102
Total	$55,031	$30,627	$85,657	$90,000	−$4,343

Spend breakdown

FY2027 to date, by month

AI spend by manager group

API list-price equivalent, from the LiteLLM gateway.

Content & Media (Jamie Park)
Growth & Web (Sam Okafor)
Platform (Robin Silva)

AI spend by model

By model tier; teams move to each new version within weeks of release (table lists versions)

Opus
Sonnet
Haiku
Other

Delivery and adoption

FY2027 to date

Prompt-cache hit rate

78%

Higher is cheaper

Tool error rate

3.0%

Failed tool calls

Sessions using subagents

50%

Work delegated to specialist agents

Headless (automated) runs

28%

Scripted and pipeline agent runs

At-bats by outcome

Shipped experiments with a measurable hypothesis

Hypothesis won
Didn't win (learned)
Still running

How teams use agents

Share of sessions by type

Harness
Freeform
Code research
Documentation
Ideation
Ops
Presentation

Friction

Where agents and engineers get stuck, FY2027 to date

Top friction points

From failed tool calls and retries

Harness practices and savings

FY2027 to date · estimated at list price

Savings opportunity by lever

$15,109 of $55,031 spend (27%) if every session followed the practices below

Engineers on target

Per practice, of engineers with data this period

rtk coverage · ≥ 80% of Bash calls1 of 24
Knowledge MCP use · ≥ 40% of sessions3 of 24
Cache hit rate · ≥ 85% of input4 of 24
Context past 100K · ≤ 10% of input21 of 24
Opus on routine work · ≤ 25% of routine spend2 of 24
Sessions using subagents · ≥ 30% of sessions24 of 24
Tool error rate · ≤ 4% of tool calls22 of 24

Practices by engineer

Grouped by team. Each cell shows the engineer's value against the target; open an engineer for their usage and recommendations.

Engineer	rtk	Knowledge	Cache	Context	Model fit	Subagents	AI spend	Savings opportunity
Content Pipeline
Avery · lead	65%Needs attention	31%Needs attention	84%Needs attention	9%On target	41%Needs attention	57%On target	$2,277	$549 · 24%
Blake	60%Needs attention	29%Needs attention	77%Needs attention	8%On target	88%Off target	53%On target	$2,030	$654 · 32%
Casey	61%Needs attention	30%Needs attention	81%Needs attention	9%On target	42%Needs attention	55%On target	$1,758	$395 · 22%
Video Rendering
Devon · lead	1%Off target	28%Needs attention	81%Needs attention	31%Off target	43%Needs attention	48%On target	$3,203	$1,038 · 32%
Emery	1%Off target	28%Needs attention	83%Needs attention	9%On target	43%Needs attention	46%On target	$3,627	$841 · 23%
Finley	3%Off target	30%Needs attention	82%Needs attention	9%On target	40%Needs attention	53%On target	$2,093	$383 · 18%
Social Publishing
Harper · lead	65%Needs attention	3%Off target	83%Needs attention	9%On target	45%Needs attention	47%On target	$2,104	$587 · 28%
Indigo	61%Needs attention	3%Off target	82%Needs attention	10%On target	46%Needs attention	47%On target	$1,798	$456 · 25%
Jules	63%Needs attention	8%Off target	82%Needs attention	9%On target	41%Needs attention	44%On target	$1,226	$269 · 22%
Growth & SEO
Jordan · lead	61%Needs attention	30%Needs attention	79%Needs attention	25%Needs attention	81%Off target	44%On target	$2,407	$1,005 · 42%
Kai	62%Needs attention	31%Needs attention	72%Needs attention	23%Needs attention	41%Needs attention	49%On target	$1,791	$844 · 47%
Lane	66%Needs attention	29%Needs attention	86%On target	7%On target	38%Needs attention	49%On target	$1,410	$324 · 23%
Web Platform
Logan · lead	63%Needs attention	30%Needs attention	82%Needs attention	8%On target	41%Needs attention	51%On target	$2,651	$569 · 21%
Marlo	64%Needs attention	32%Needs attention	82%Needs attention	8%On target	44%Needs attention	49%On target	$2,284	$543 · 24%
Noel	3%Off target	28%Needs attention	67%Off target	8%On target	40%Needs attention	52%On target	$1,803	$787 · 44%
Data & Analytics
Oakley · lead	54%Needs attention	30%Needs attention	81%Needs attention	8%On target	48%Needs attention	43%On target	$2,199	$741 · 34%
Parker	61%Needs attention	30%Needs attention	80%Needs attention	9%On target	42%Needs attention	45%On target	$2,157	$656 · 30%
Quinn	65%Needs attention	31%Needs attention	66%Off target	9%On target	82%Off target	49%On target	$1,612	$965 · 60%
Agent Platform
Reese · lead	80%On target	64%On target	87%On target	2%On target	18%On target	57%On target	$3,515	$398 · 11%
Riley	77%Needs attention	50%On target	77%Needs attention	10%On target	31%Needs attention	59%On target	$3,254	$966 · 30%
Rowan	66%Needs attention	31%Needs attention	85%On target	8%On target	40%Needs attention	61%On target	$2,360	$515 · 22%
Infrastructure
Sage · lead	77%Needs attention	51%On target	88%On target	4%On target	21%On target	53%On target	$3,057	$357 · 12%
Tatum	63%Needs attention	28%Needs attention	79%Needs attention	8%On target	43%Needs attention	52%On target	$2,485	$770 · 31%
Wren	64%Needs attention	28%Needs attention	82%Needs attention	9%On target	41%Needs attention	47%On target	$1,929	$497 · 26%

Team and engineer detail

Grouped by reporting line, not ranked

Engineers see their own row first; managers use it to find friction to remove. It is never used to rank people.

Engineer	Role	Sessions	Tool calls / session	Subagents	Headless	Tool errors	Cache hits	AI spend
Content Pipeline · Jamie Park
Avery	Lead engineer	191	33.9	57%	31%	2.4%	79%	$2,277
Blake	Engineer	184	28.3	53%	28%	2.8%	77%	$2,030
Casey	Engineer	146	31.4	55%	31%	2.1%	80%	$1,758
Video Rendering · Jamie Park
Devon	Lead engineer	200	28.8	48%	27%	2.7%	77%	$3,203
Emery	Engineer	204	27.2	46%	25%	2.9%	79%	$3,627
Finley	Engineer	126	28.9	53%	30%	3.5%	80%	$2,093
Social Publishing · Jamie Park
Harper	Lead engineer	211	29.1	47%	26%	3.2%	78%	$2,104
Indigo	Engineer	187	29.7	47%	26%	4.1%	80%	$1,798
Jules	Engineer	130	29.4	44%	26%	4.0%	80%	$1,226
Growth & SEO · Sam Okafor
Jordan	Lead engineer	241	29.6	44%	26%	3.0%	78%	$2,407
Kai	Engineer	167	31.0	49%	28%	2.8%	68%	$1,791
Lane	Engineer	139	29.7	49%	27%	3.3%	80%	$1,410
Web Platform · Sam Okafor
Logan	Lead engineer	206	32.3	51%	29%	2.6%	79%	$2,651
Marlo	Engineer	169	33.0	49%	26%	3.0%	79%	$2,284
Noel	Engineer	136	30.0	52%	32%	2.5%	64%	$1,803
Data & Analytics · Sam Okafor
Oakley	Lead engineer	178	30.7	43%	22%	3.9%	77%	$2,199
Parker	Engineer	185	28.9	45%	26%	3.3%	77%	$2,157
Quinn	Engineer	141	30.7	49%	28%	2.9%	65%	$1,612
Agent Platform · Robin Silva
Reese	Lead engineer	189	38.1	57%	31%	2.5%	87%	$3,515
Riley	Engineer	184	32.0	59%	33%	2.8%	77%	$3,254
Rowan	Engineer	159	37.7	61%	33%	2.2%	82%	$2,360
Infrastructure · Robin Silva
Sage	Lead engineer	205	33.6	53%	28%	2.9%	86%	$3,057
Tatum	Engineer	159	33.7	52%	28%	3.2%	76%	$2,485
Wren	Engineer	130	32.3	47%	27%	3.4%	79%	$1,929

Methodology

Spend comes from LiteLLM gateway logs and imported agent session history, priced at API list rates.
Forecast: a linear trend fitted to each team's last six months of spend, summed across teams and projected to Jan 2027. The shaded range is an 80% prediction interval, widened cumulatively.
Fiscal year FY2027 runs Feb 2026 through Jan 2027. Budgets are set per team at the start of the year.
Savings are estimates at list price: routine Opus work repriced at the Sonnet available at the time; uncached input moved to cache reads up to an 85% hit rate; rtk removing 70% of unfiltered command output, which stays in context for about half of the session's remaining turns; and half of the input beyond 100K context avoided by compacting. Levers overlap, so the total is an upper bound. rtk and the knowledge MCP are scored from their rollout (rtk May 2026, knowledge MCP Jun 2026).
Sessions are classified by usage type from the tools they used and the files they changed. At-bats are shipped initiatives with a measurable hypothesis.
Generated Oct 20, 2026, 7:46 AM.