Federal program cost & performance evidence

AI prepares.
Humans approve.

TELOS takes public federal program data — spending records, budget justifications, performance reports — and walks it through a five-stage pipeline to analysis-ready imports. Every value keeps its receipts. Nothing exports without a named human sign-off.

See the pipeline Talk to MHG
Staged record · SBA 7(a) Approved
Value
$25.4B loans guaranteed, FY 2024
Source
FY 2025 Congressional Budget Justification, p. 42
Extracted
Run #128 · prompt v6 · reviewed line-by-line
QA
9 checks passed · 0 blockers
Approved by
Second reviewer — not the preparing analyst

Every number on screen can show this card.

The principle

AI does the tedious part. People stay accountable.

Cost and performance analysis stalls on data preparation — hundreds of pages of budget documents, thousands of spending rows. TELOS splits the work the only way that survives an audit:

1

AI proposes

Models extract figures from source documents, tag records, and map them to your cost-model structure — fast, and always as a draft.

2

Humans review

Analysts inspect and edit every staged record. Automated QA flags duplicates, bad numbers, and gaps before anything moves forward.

3

Only approved data ships

A second reviewer — never the preparing analyst — signs off. Exports contain approved records only, formatted for direct import into your cost-modeling platform.

The pipeline

Five stages, each one visible.

TELOS is a dashboard, not a black box. Work moves left to right across lanes you can watch, and every stage is timed — so you know exactly what the AI assist is saving you.

  1. 1

    Acquire

    Pull spending data live from USAspending.gov and register source documents — budget justifications, performance reports — with checksums.

  2. 2

    Analyze

    Profile what arrived: coverage, fiscal years, anomalies. Know your data before you trust it.

  3. 3

    Prepare

    AI-assisted extraction from source PDFs, tagging, and mapping to your target model — every proposal linked back to its page and prompt version.

  4. 4

    QA

    Rule-based checks catch duplicates, invalid numbers, and missing fields. Blockers stop a batch; findings are logged, not overwritten.

  5. 5

    Approve & export

    Two-person sign-off, an immutable audit entry, and a clean export with model-ready headers. A printable batch certificate documents the run.

Traceability

Every value can show its source.

Ask any figure in TELOS where it came from and it answers: the source document and page, the extraction run and prompt version that proposed it, the QA checks it passed, and the named reviewer who approved it.

Behind that sits a hash-chained audit log — each entry cryptographically linked to the one before it, so history can be verified and can't be quietly edited.

  • Source document, page, and checksum on every extracted value
  • Extraction run and prompt version recorded — AI output is never anonymous
  • QA findings preserved alongside the data they flagged
  • Approvals recorded by name, enforced two-person rule
  • Hash-chained audit log — tamper-evident by construction
  • Stage timers quantify the AI speedup on every batch

Built on public data

Start from the program inventory. End at your cost model.

Pick programs straight from the OMB Federal Program Inventory; TELOS wires up the corresponding spending data automatically and organizes work by agency and program portfolio.

OMB Federal Program Inventory

Program-first entry point — browse and select federal programs by agency and category.

USAspending.gov

Live award and assistance data, pulled by Assistance Listing number for each selected program.

Budget justifications & performance reports

Agency CBJs and annual performance reports, registered as sources and mined page-by-page.

CostPerform integration

Built to feed CostPerform.

Mason Harriman Group is a partner and implementation provider of CostPerform, the cost and performance modeling platform. TELOS was designed around a deliberate division of labor: TELOS prepares and approves the evidence; CostPerform models the costs and computes cost-per-outcome.

Model-ready exports

Approved batches export with CostPerform-formatted headers — attribute symbols included — and allocation syntax where relationships are known, so imports load without manual rework.

One-way, approval-gated

Data flows in one direction: from TELOS's approved staging tables into CostPerform. Nothing reaches your cost model that hasn't passed QA and two-person sign-off.

Traceability carried through

Because every exported value keeps its lineage in TELOS, any figure in your CostPerform model can be traced back to its source document, page, and approver.

The name

τέλος — end, purpose, outcome.

TELOS is Greek for the end a thing is for. Federal programs exist to produce outcomes; TELOS prepares the evidence that connects what a program spends to what it achieves.

  • TTagging
  • EExtraction
  • LLineage
  • OOversight
  • SSign-off

About Mason Harriman Group

See it run on your programs.

TELOS was built by Mason Harriman Group (MHG) to demonstrate a working principle: AI-assisted data preparation with human accountability, end to end, on real federal data. We'd be glad to show you.

[email protected]