Tallgrass Components
The model that leaves out what it cannot claim
An $85M distributor with a $450K technology budget. Six AI use cases surface; four carry verified dollars; and $335,000 in unevidenced savings gets withdrawn by the instrument itself.
Demonstration data. Every company, person, and transcript in these case studies is fictional, and none is a client engagement. Every figure is real output from a reproducible MCOS run, graded against a written answer key before publication.
$800,000
Grounded annual savings, 4x verified
178.6%
ROI on $287,100 invested
4 months
Payback
$335,000
Unevidenced savings withdrawn
The engagement
The discovery session
One structured discovery conversation with the COO, VP of supply chain, and controller. The operating pains carry stated dollars, and the COO states the frame on the record: $450,000 approved for supply chain technology, with a $75,000 approval threshold. The instrument treats the threshold as governance, not value; it appears in no money field.
The funded portfolio
Of six use cases, four carry dollars the source text supports, totaling $800,000 in grounded annual savings, each tagged with its verification path. The investment to capture them, $287,100, is the sum of the funded use-case costs, which yields an ROI expectation of 178.6% and roughly a four-month payback. The budget does not bound the claim; the report says so explicitly.
The withdrawal
Two more rows wanted to claim $335,000 in savings. The evidence was not in the transcript, so the guard withdrew the dollars and said so in the report. The rows survive as opportunities without valuations. That is the entire thesis of the product in one behavior: the model that leaves out what it cannot claim is the model you can trust.
What the instrument would not claim
- Every scenario figure in the financial risk analysis is labeled modeled, not spine-verified, in the output itself. The $800,000 spine and the modeling around it are never blended.
- The $75,000 approval threshold appears in zero money fields. Governance facts are not value.
- These are the same certified numbers the public Tallgrass demo ships. Re-verified 12/12 on current code the day this page was built.
Numbers of record
| Grounded annual savings | $800,000 (4 use cases, l2_verified) |
| Stated budget / threshold | $450,000 approved / $75,000 approval threshold |
| Investment | $287,100 (sum of funded use-case costs) |
| ROI / payback | 178.6% / 4 months |
| Withdrawn by the guard | $335,000 across 2 unevidenced rows |
Run of record
- Harness
- audits/10_ai_discovery_pipeline/run_tallgrass_aid.py
- Grade
- 12/12 answer-key claims hold
- Verified
- mcos-core e04445f on 2026-08-14
Deterministic rule-mode run of the production AI Discovery pipeline; the same certified fixture powers the live homepage demo.
Run AI Discovery on your own transcripts
The same pipeline that produced every number on this page, priced per run, with the same grounding guards on.