Outcome, not activity
Token counts and seat licences measure effort. TIER measures results — quality-weighted engineering outcomes, tied to the spend that produced them.
An open measurement method, backed by Intenteon
We don’t just deploy AI. We measure whether it pays.
Any vendor can spend your budget on tokens. The harder question is what those tokens produced. So we built TIER — the Token Impact & Efficiency Ratio — to answer it.
Most AI vendors sell capability. We measure it.Outcomes, not compute
What it is
A yardstick for the yield of AI-assisted engineering
TIER is a self-hosted, open-source meter for the yield of AI-assisted engineering: quality-weighted outcomes per $1,000 of spend, per team, over time. Deterministic, no conversation content stored, and it ships its own adversarial failure-mode analysis.
The tagline says it plainly: measure what your AI spend produces, not what it burns. If DORA gave engineering teams a shared language for delivery performance, TIER does the same for the era of AI-augmented engineering — turning “we’re using AI” into a number you can track, defend, and improve.
It runs as a single Go binary with an embedded SQLite database — single-tenant, deterministic, and private by construction. It records the outcome, not the transcript, so measuring your engineering never means handing your prompts or code to anyone else.
Why it matters
Turn AI spend from an act of faith into a number
Accountable ROI
One honest ratio — outcomes per $1,000 of AI spend — that a team, a director, and a board can all read the same way. Buy results, not compute.
Honest by design
Deterministic scoring and a built-in adversarial failure-mode analysis — TIER is engineered to resist the vanity metrics that make AI look good on a slide.
Yours to run
Open-source (Apache-2.0) and self-hosted. No data leaves your boundary, no conversation content is stored, and there is no vendor between you and your own numbers.
Dogfood, in the open
We point TIER at ourselves first
We don’t ask you to measure what we won’t. Here is Intenteon’s own TIER dashboard for AI-assisted engineering — self-reported, and shown with the honesty TIER is built for.
as of 2026-07-18 · trailing 30-day window Self-reported from Intenteon’s own engineering. These figures move week to week — that’s the point of measuring them.
221
quality-weighted outcome points per $1,000 of AI spend
$4.52
cost per weighted outcome point
152 → 723
merged pull requests to weighted outcome points
$3,268
measured AI spend over the window
68%
exploratory work, by TIER’s own classification
Figures are Intenteon’s self-reported results over the stated trailing window and are expected to drift over time. TIER does not produce cross-organization benchmarks — it measures your team against your own trend, not against anyone else’s.
In our practice
The same discipline, on your AI initiatives
TIER is how we hold our own AI work accountable, and it is the discipline we bring to yours. When we help an organization adopt AI, “did it work?” is not a feeling at the end of the engagement — it is a measured ratio we can point to from the first sprint. That is what we mean when we say we deliver outcomes, not demos.
Open source
TIER is free, self-hosted, and backed by Intenteon
TIER is its own open-source project with its own identity — and it is stewarded by Intenteon. It is built to be run by the people it measures: self-hosted, source-readable, and pointed at your own AI-assisted engineering the way we point it at ours.
TIER’s home is tiermetric.org — not yet open.
Backed by INTENTEONQuestions
Frequently asked questions
What is TIER?
TIER — the Token Impact & Efficiency Ratio — is a self-hosted, open-source meter for the yield of AI-assisted engineering: quality-weighted outcomes per $1,000 of spend, per team, over time. It is deterministic, stores no conversation content, and ships its own adversarial failure-mode analysis. Think of it as DORA for AI-augmented engineering.
Is TIER an industry standard?
TIER is an open, Intenteon-authored methodology and open-source tool — not an industry-anointed standard. We built it to measure our own engineering honestly, and we’re opening it up so other teams can measure theirs the same way.
Does token efficiency actually affect my costs and ROI?
Yes. AI spend is easy to grow and hard to justify. TIER ties every $1,000 of AI spend to the quality-weighted engineering outcomes it produced, so you can see whether more compute is actually buying more results — and buy results instead of compute.
Is my data safe with TIER?
TIER is self-hosted and single-tenant — a single Go binary with an embedded SQLite database. It is deterministic and stores no conversation content, so measurement never means shipping your prompts or code to a third party.
Ready to measure what your AI actually delivers?
Tell us what you want your AI to accomplish. We’ll show you how to prove it’s working. Your Intent. Delivered.