Rowan

Engineer · AI usage, harness practices and where spend could have gone further.

Demonstration report. This engineer and their usage are synthetic.

This page is for the engineer and their manager to find cheaper defaults and harness fixes. It is never used to rank people.

Period

Summary

FY2027 to date

AI spend

$2,360

159 sessions · $15 each

Savings opportunity

$515

22% of spend, at list price

Practices on target

4 of 7

Focus: rtk coverage, Knowledge MCP use

Cache hit rate

85%

Team 84% · org 82%

FY2027 to date · estimated at list price

$515 of $2,360 spend

Largest saving first

Improve prompt caching$270
Cache hit rate 85% against an 85% target.
Keep stable context at the front of prompts and avoid switching models mid-session, so 85% of input reads from cache.
Right-size models$164
40% of routine spend ran on Opus: code research $136 of $355, documentation $78 of $187, ops $146 of $357, presentation $33 of $75.
Run docs, decks, ops and read-only research on Sonnet instead of Opus.
Keep context lean$64
8% of input tokens were past 100K of context.
Compact or /clear between tasks instead of carrying more than 100K tokens of context.
Route commands through rtk$17
66% of 1,788 Bash calls ran through rtk; 2.6M tokens of command output went into context.
rtk compacts test, build, git and search output before it reaches the model (about 70% smaller).

Compared with Agent Platform and the organization

Practice	Rowan	Team	Org	Target	Recommendation
rtk coverage	66%Needs attention	74%	56%	≥ 80% of Bash calls	Install the rtk hook so test, build, git and search commands return compact output.
Knowledge MCP use 6 findings saved back	31%Needs attention	48%	30%	≥ 40% of sessions	Search the knowledge MCP before re-researching a codebase, and save findings back with remember.
Cache hit rate	85%On target	84%	82%	≥ 85% of input	On target
Context past 100K	8%On target	6%	10%	≤ 10% of input	On target
Opus on routine work	40%Needs attention	29%	44%	≤ 25% of routine spend	Default routine sessions to Sonnet and keep Opus for harness, design and hard debugging work.
Sessions using subagents	61%On target	59%	50%	≥ 30% of sessions	On target
Tool error rate	2.2%On target	2.5%	3.0%	≤ 4% of tool calls	On target

Is the engineer moving toward the targets?

Month	rtk	Knowledge	Cache	Context	Model fit	Subagents	Savings opportunity
Feb 2026	–	–	61%Off target	9%On target	62%Off target	39%On target	$35
Mar 2026	–	–	62%Off target	8%On target	62%Off target	47%On target	$34
Apr 2026	–	–	65%Off target	9%On target	62%Off target	34%On target	$57
May 2026	26%Off target	–	69%Off target	8%On target	55%Off target	42%On target	$73
Jun 2026	47%Needs attention	15%Needs attention	77%Needs attention	8%On target	53%Off target	69%On target	$82
Jul 2026	72%Needs attention	27%Needs attention	77%Needs attention	8%On target	40%Needs attention	41%On target	$71
Aug 2026	73%Needs attention	32%Needs attention	82%Needs attention	10%On target	38%Needs attention	53%On target	$83
Sep 2026	74%Needs attention	37%Needs attention	82%Needs attention	7%On target	30%Needs attention	60%On target	$34
Oct 2026	69%Needs attention	35%Needs attention	91%On target	7%On target	27%Needs attention	77%On target	$46

FY2027 to date

By model tier, by month

Routine types (documentation, presentation, ops, code research) rarely need Opus

Cache reads are billed at a fraction of uncached input; a high share is what good caching looks like

Model	Requests	Uncached input	Cache reads	Cache writes	Output	Spend
Opus 5.5	5,120	10.4M	246.3M	16.4M	7.8M	$328
Opus 5	4,115	23.5M	172M	12.5M	4.7M	$398
Opus 4.8	2,656	24.4M	112.8M	8.8M	3.7M	$326
Opus 4.6	2,050	32.5M	82.6M	7.4M	3.3M	$332
Opus 4.5	80	1.4M	2.7M	260K	116K	$13
Sonnet 5.5	7,597	15.6M	408.9M	27.1M	11.3M	$253
Sonnet 5	5,757	39.2M	277.4M	20.2M	8.2M	$399
Sonnet 4.6	2,196	27.5M	91.1M	7.6M	3M	$183
Sonnet 4.5	107	2.1M	3.9M	379K	150K	$11
Haiku 5.5	11,521	22.6M	642.4M	42.4M	14.6M	$21
Haiku 4.5	3,552	28.1M	174.4M	12.9M	5.3M	$88
GPT Realtime (meetings)	99	693K	4.5M	334K	128K	$8
Total	44,850	227.7M	2219M	156.2M	62.3M	$2,360

Spend and tokens come from LiteLLM gateway logs and imported agent session history, priced at API list rates.
Savings are estimates: 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 the session's remaining turns; and 50% of input past 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).
Generated Oct 20, 2026, 7:46 AM.