Model decision · March 18, 2026
Your Model Mix Is a P&L Decision, Not a Benchmark Decision.
Frontier or open-weight, the choice shapes cost, control and recovery — not just token price.
The System of Record for AI
What your AI costs. What it returns. What you can prove.
Oabo Scorecard is the System of Record for AI: a living AI system inventory linked to cost, value, governance, performance, and the evidence behind every ROI claim.
30 days. Up to 3 AI systems. No card.
IT
invoice-triageone agent, one monthCounted
On record
Claimed
$40,732
Verified
$23,932
Graded
Two axes
Readout
Two units
The verified share above comes from these five stages.
IT
invoice-triageone agent, one monthCounted
Runs on record
Claimed
$40,732
Verified
$23,932 · 59%
Graded
On two axes
On the readout
In two registers
The verified share above comes from these five stages.
One number, on trial
Every figure carries its own paperwork.
One sample claim, opened all the way down — and the estimate that is kept out of the total. The record it opens into is the AI value ledger.
Hardened cash · sample month
$48,312.67Claim CLM-2291 · Invoice exception triage · GL 6410 Contract Labor
Not counted
One assisted estimate, $61,940.00 — labelled, kept out of this total.
Sample data
Cash ledger · sample workspace
Claim
Description
Amount
CLM-2287
Duplicate-invoice recovery
$12,904.15CLM-2291
Invoice exception triage
$48,312.67GL 6410 Contract Labor
JE-88214
Contract labour — September invoices
$21,480.00JE-88257
Overtime reversal, AP team
$12,118.42JE-88301
Temp-staffing cancellation
$9,940.25JE-88342
Late-payment penalties avoided
$4,774.00Foots to
$48,312.67CLM-2304
Vendor statement reconciliation
$9,731.40Claims · invoice-triage
Claim
Value class
Amount
CLM-2291$48,312.67
Ledger-backed
Evidence level
Self-reported
Validator-confirmed
You never grade your own claim.
Claims · scroll one row
Claim
Value class
Amount
In total
CLM-2291
Ledger-backed
$48,312.67Counted
CLM-2318$61,940.00
Assisted estimate
Excluded
Hardened total
$48,312.67unchanged
An estimate is a real thing to record. It is not a thing to add up.
Audit log · linked history
2026-09-30 14:02:11a.riveraclaim.validatedCLM-2291
hash 9f2c1a7b3e04
2026-09-30 14:06:48s.okaforevidence.attachedCLM-2291
preceding 9f2c1a7b3e04
hash c07d55ae91f2
Each entry is hashed with the one before it. Remove a row and every hash after it breaks.
The readout · two registers
Cash · hardened
$48,312.67Realized dollars
Capacity
412 hrsHours, pending conversion
Two units. They are never added together.
Count coverage
A92%
Rate provenance
B84%
Reported as a pair. Never averaged into one letter.
The problem
AI spending is growing faster than companies can explain the return.
Everyone can list their AI projects. Few can prove what they return. Closing that gap is AI ROI measurement: a return with the evidence behind it.
- Have an AI strategy100%
- Know their AI spend76%
- Can explain the spend43%
- Can tie spend to ROI14%
3.5–5×
More likely to report established ROI
With clear accountability and strong cost visibility.
KPMG Global AI Pulse, Q2 2026
Sources: KPMG Global AI Pulse (2026); McKinsey State of AI (2025).
The value gap
Paying for AI is now normal. Saying what it earned is not.
Paying for AI rose 6.8 points across eight monthly prints — 43.8% to 50.6% of US businesses on Ramp’s platform. 2% of S&P 500 companies put a number on what AI did to their earnings last quarter — the same 2% as the quarter before, and 1% the quarter before that. We track both, each on its own scale.
Paying for AI
50.6%of US businesses on Ramp’s card and bill-pay platform paid for at least one AI tool in April 2026: up 6.8 points from 43.8% in September 2025, against about 35% a year before.
Ramp AI Index, May 2026 edition — Apr 2026 data · source
Can name the return
2%of S&P 500 companies put a number on what AI did to their earnings last quarter — 1%, then 2%, then 2% across the three published quarters.
Goldman Sachs, via Investing.com, August 17, 2026 · source
Sources: Ramp AI Index and Ramp Economics Lab, dated monthly editions from October 2025 to May 2026 (businesses on Ramp’s card and bill-pay platform); Goldman Sachs’ count of S&P 500 earnings calls, reported by Fortune (March 2026) and Investing.com (August 2026). Every figure carries a link to the page it was read from.
Last updated August 17, 2026. How we source this. Each panel keeps its own scale because the two series count two different populations on two different clocks — businesses buying AI tools month by month, and S&P 500 companies speaking on a quarterly call. Neither contains the other, so the distance between the panels is a contrast, not a subtraction.
The record
A month from the sample workspace
Quality grade
Credible
On current evidence
Verified through source checks
Sample cash saved
$20,704After this claim's signed 20% hold-back
PotentialClaimedVerifiedHardened
$
Realized cash$20,704
Verified
H
Capacity created298 hrs
Pending
~
Assisted estimatesSeparate
Labeled
Illustrative sample · As of Aug 16, 2026
Every figure traces to a source · cash, capacity, estimates, and spend remain separate
The product
See the record your team works in
These are screenshots of the sample workspace: cash after the hold-back, hours kept apart from dollars, and how much of it is verified.
One readout your board can trust — cash, hours, and how much is verified.
Every AI system on one record — owner, lifecycle, risk, spend, and which models it calls.
Cash anchored to your ledger. Hours in their own register. Every figure carries its state.
The trail a reviewer reads: who recorded what, when, and the evidence behind each number.
Use cases
Find the company that looks like yours.
Ten shapes of company, from fifteen people to four thousand. Each one says where the AI runs, what Oabo puts on the record in the first thirty days, and what to expect in cash and in hours.
Agency
Your agency bills hours. AI just handed some back.Writers, designers, and editors bill against retainers. AI now does part of that work. The retainer did not change.
Cash saved
$41,000
Capacity created
2,100 hrs
SaaS
Support answers cost money. Now you know how much.The assistant deflects tickets all day. Finance sees one API bill and no idea what it bought.
Cash saved
$96,000
Capacity created
5,400 hrs
Accounting
Busy season is the test. Prove the AI helped.One compressed quarter, seasonal staff on every engagement, and a tool the partners approved on a promise.
Cash saved
$188,000
Capacity created
6,900 hrs
Logistics
Every exception costs cash. Count the ones AI cleared.Twelve hundred people move freight. A handful of models now triage the problems, and nobody has priced the difference.
Cash saved
$610,000
Capacity created
19,000 hrs
Insurance
Claims move faster. Finance still needs the working.Intake, triage, and first review now run with AI. The finance business partner wants that in the operating plan.
Cash saved
$430,000
Capacity created
14,500 hrs
Regulated bank
Your examiner will ask. Have the evidence linked.Four thousand people, a model inventory that grew faster than the review process, and a supervisory letter that is now current.
Cash saved
$1,240,000
Capacity created
11,800 hrs
Legal
Document review got cheaper. Bill it honestly.Four practice groups, a fixed-fee book that keeps growing, and first-pass review that no longer takes a week.
Cash saved
$74,000
Capacity created
3,300 hrs
Healthcare back office
Denials, coding, prior auth. Which line moved?AI now drafts appeals and pre-checks authorisations. The CFO wants that in the operating statement, not in a slide.
Cash saved
$355,000
Capacity created
12,600 hrs
Manufacturing shared services
Shared services runs on volume. Nine plants want the number.Finance, procurement, and HR operations across nine plants, and AI arrived one team at a time.
Cash saved
$720,000
Capacity created
24,000 hrs
Consultancy
Fifteen people. Nine subscriptions. One bill nobody reads.A small consultancy bills expertise. AI is now inside every deliverable, and every consultant expensed their own.
Cash saved
$9,400
Capacity created
780 hrs
Every figure on these cards is an expected pattern for a company of that shape, not a named customer’s claim. Cash and hours are two measurements and stay in two blocks.
The record
See how Oabo connects cost, evidence, and value.
See the evidence behind every cost and benefit.
Recorded
Reportable
Defensible
Why now
Why companies need this now
“Building is easier than it has ever been. Generating real value is still hard.”
“A currency where you have no instinct for what you’re using, and the accounting practices aren’t even there for it … treated as an investment, but a risky one in case it has no return.”
What Oabo gives you
Tied to financial records Backed by source evidence Easy to recalculate
Insights
See which AI results are supported by evidence.
Oabo Insights shows how leaders can check the sources and calculations behind AI costs and results.
Operating model · February 11, 2026
Stop Reviewing AI Use Cases. Review the Value Stream.
Once AI changes real work, a project list is too small a unit of management.
Agentic shift · December 17, 2025
A Chatbot Answers. An Agent Commits the Enterprise.
Not intelligence — whether it can alter work, spend or a customer record.
How it works
How Oabo works
It starts with an AI system inventory: one record per AI system, with a named owner, a clear purpose, and its cost.
Your AI systems
01Connect tools
02Assign an owner
03Link the evidence
04Share the readout
$ · hrsBoard readout
01
Connect your AI tools
Send one usage record so Oabo can identify each AI system and begin tracking its cost.
02
Assign an owner
Give every AI system one record with a named owner, a clear purpose, and its cost.
03
Connect results to source data
Show where each result came from and how it was calculated. Estimates remain clearly labeled.
04
Share the result and its evidence
Give finance, operators, and leadership the result, its source, and the calculation behind it.
See how Oabo calculates and checks each result.
See what counts as cash, what stays in hours, and what remains an estimate.
Who it is for
What each team can see
Finance
See AI spend, evidence, and value in one place.
AI leaders
See every AI and agent, who owns it, what it costs, and what it returns.
Risk & compliance
Trace each claim to its source and review history.
Leadership
See what is working, what is not, and where to invest next.
Connects to the tools you already use
OpenAI
Anthropic
Azure
AWS
Snowflake
NetSuite
Workday
Stripe
and many moreWhere Oabo sits
What Oabo adds
Each neighboring tool sees one slice — spend, risk, or execution. Oabo connects them, so cost, value, and AI governance share one record.
FinOps
CloudZero · Vantage
Tracks
Spend
Does not show
What AI did, what it returned
GRC / inventory
OneTrust · Credo AI
Tracks
Existence, risk
Does not show
Cost, output, value
Observability
LangSmith · Arize
Tracks
Execution, quality
Does not show
Whether the output improved the business or paid off
ROI / value
Internal tools · decks
Tracks
Self-reported ROI
Does not show
Source evidence and calculations finance can check
Get started
Start your own record
Your free 30-day workspaceYour own copy of the whitepaperA read-only sandbox
Or take it slower
See what the method covers. Then talk to us.
The method
What every figure is built from, how it is sourced, and how it is checked. The write-up itself comes with your access link.
See what it coversWho you will actually work with
You set Oabo up yourself — connect a tool, name an owner, and the record starts. When you want a second pair of eyes on a number, you get the founder, whose background is advising executive teams through performance and restructuring work: the discipline of making a figure survive scrutiny, now pointed at what AI costs and what it returns.