# The AI Assessment: a value assessment brief

For a finance or operations executive who owns a number and runs two or more systems of record for the same objects. October 7, 2026.

The AI Assessment is the first step of one way to buy EOS, the Enterprise Operating System: a fixed two-week assessment, a pilot that runs to day 90, then a commercial term paid as a share of value your own finance function has verified. This brief states what the assessment produces, how the two weeks run, what you provide, how the fee is credited, how the value share is measured afterward, and the guarantees in writing. Every term below is stated on a published page of this site, named at the end of its paragraph, and no term appears here that those pages do not state.

## 1. The assessment in one paragraph

The assessment is $15,000, fixed, for two weeks. It connects read-only to two systems of record that you name, with up to 10 named users and one agent key with read and propose authority, in a private workspace. It finds where the two systems disagree about the same object, quantifies each disagreement in currency on your valuation basis, and writes the targets with you: what will be measured, from which records, against which baseline, and what counts as achieved. The $15,000 is credited in full against the $50,000 pilot. The assessment changes nothing in your systems. Source: /pricing.html, /pilot.html, /request-assessment.html.

## 2. What the assessment produces

You hold five things when the two weeks end, whichever way it ends. The register and the decision log are the charter's deliverables; the first finding, the baseline and the targets are the commercial steps the pricing page describes.

| Deliverable | What it is | Where the term comes from |
|---|---|---|
| The finding register | Every contested figure where two systems of record disagree about the same object, each value traced to a dated record, with its source, timestamp and record reference from the first day. Your named authority sees it as it grows, not at a reveal meeting. | /pilot.html (Shape, Deliverables) |
| The first finding | The first contested figure that meets the charter's definition of a finding: each value traces to a dated record, the gap is quantified in currency on your own valuation basis, and your named authority confirms it was previously unknown or unresolved. It is delivered within 14 days. | /pilot.html (The standard), /data/pricing.json (first finding, 14 days) |
| The value baseline | Yours, validated jointly at Day 0 with your controller or equivalent, before anything is measured. Cycle time is measured the same way: the baseline is your own historical timing, taken at Day 0. | /pricing.html (What is fixed for everyone), /velocity-attribution-rule.html |
| The decision log | Figures validated or rejected in writing, and decisions recorded with a name and a timestamp, handed to you with the register. | /pilot.html (Shape, Deliverables) |
| The targets and the written outcome | The targets written with your finance function, auditable by that function before anything else is signed, and the assessment outcome as a written document. If you read it and decide not to proceed, the engagement ends there and nothing further is owed. | /pricing.html (The enterprise engagement, step 1) |

If the assessment finds nothing material, the same register says so plainly: a clean estate is a legitimate finding about your estate, and it is reported as exactly that. Modeled savings, projected annualizations, data-quality observations without a currency figure, and anything your team has not confirmed in writing are not findings. An evidence pack that verifies offline without us is specified, not built yet. Source: /pilot.html.

## 3. The two weeks, day by day

The order of the work is the charter's: connect and normalize, reconcile, then findings and decisions. The day assignments are the working plan, set with you at the scoping call; the terms that do not move are the two weeks, the read-only access, the two systems and the day-14 first finding.

| Day | What happens | Source |
|---|---|---|
| Before Day 0 | The thirty-minute scoping call, which is free. We reply to a request within one business day with a scope and a date. You keep everything delivered. | /pilot.html (Commercials), /request-assessment.html |
| Day 0 | You name the two systems and the object family (purchase orders, inventory positions, supplier balances). Your security team chooses the access route. Your named authority, your named users and your controller are named. The baseline is validated jointly with your controller or equivalent. | /pilot.html (Shape), /pricing.html |
| Day 1 | Read-only, minimum-privilege access is in place, as exports or scoped API credentials, your security team's choice. Nothing is installed inside your systems and no write permission is requested. | /pilot.html (Shape) |
| Days 2 and 3 | Both sources land in one canonical shape. Every figure carries its source, timestamp and record reference. The context service runs in a private workspace for your named users, and your one agent key is issued with read and propose authority. | /pilot.html (Shape), /pricing.html |
| Days 4 to 8 | The contested figures surface. Each is preserved with both values and traced to the dated records that produced it. Your named authority sees the register as it grows. | /pilot.html (Shape) |
| Days 9 and 10 | Each gap is quantified in currency on your own valuation basis. The evaluation battery runs and the report is drafted. | /pilot.html (The standard), /pricing.html |
| Days 11 and 12 | The register is walked with your team. Figures are validated or rejected in writing; decisions are recorded with a name and a timestamp. Our arithmetic never validates itself. | /pilot.html (Shape, The standard) |
| Day 13 | The targets are written with your finance function: what will be measured, from which records, against which baseline, and what counts as achieved. | /pricing.html |
| Day 14 | Handover: the finding register, the decision log, the targets and the written assessment outcome. If you proceed to the pilot, the day-90 measurement is scheduled before we leave. Whether to proceed is your decision afterward, not a commitment made now. | /pilot.html (Shape), /pricing.html |

## 4. What you provide

- The two systems of record by name, and the object family to examine. Source: /pilot.html.
- Read-only, minimum-privilege access: exports or scoped API credentials, your security team's choice. Your security team may run the public verification harness first, cold and offline. Source: /pilot.html.
- Up to 10 named users, who sign in with their company email from the roster you share, with no consumer identity check. External parties you invite one by one, such as a supplier, a consultant or an account manager, take a named-user place in the assessment. Source: /pricing.html.
- One agent key with read and propose authority, for the agent you choose to connect. It can read and propose; a change is made only by your own executor, on a named person's approval, and EOS verifies the result. Source: /pricing.html, /pilot.html.
- A named authority: the executive who owns the number and who confirms a finding was previously unknown or unresolved. Source: /pilot.html.
- Your controller or equivalent, to validate the baseline at Day 0, and your finance function, to audit the targets. Source: /pricing.html.
- Written confirmation of each figure and its basis. Validated means your team confirms the figure and its basis in writing. Source: /pilot.html.

You are never asked for a write credential, a card or bank details on this site. Fees are settled under your signed agreement. Source: /pricing.html.

## 5. Users, systems and the agent key at each step

| Step | Systems of record | Named users | Agent keys |
|---|---|---|---|
| AI Assessment | Two | Up to 10 | One, with read and propose authority |
| Pilot | Up to 4 systems | Up to 25 | Up to 3 agent keys |

In the commercial term, every employee on your roster or in your verified email domains uses EOS with no seat fee. External parties you invite one by one take a named-user place in the assessment and the pilot, and are $100 per seat per month in the commercial term. Source: /pricing.html, /data/pricing.json.

## 6. How the fee is credited

- The assessment is $15,000, fixed. It is credited in full against the $50,000 pilot, so the pilot adds $35,000 more.
- The $15,000 assessment, credited against the pilot, is not refundable. The assessment is yours to keep and to act on with or without us.
- The $35,000 pilot balance is fully refunded if the mutually agreed target is not achieved by day 90. The refund is paid by ACH. In that case your one-time cost is the $15,000 assessment.
- If you read the assessment outcome and decide not to proceed, the engagement ends there and nothing further is owed.

Source: /pricing.html (Which path is yours, step 1, step 2), /pilot.html (Commercials), /data/pricing.json (refund method).

## 7. How the value share is measured afterward

After a pilot that achieves its target, the commercial term begins: $0 a year, with no seat fee for your employees, no modules and no per-connector fee. We are paid a marginal share of value your finance function has verified: 15% of the first $100,000, 10% to $500,000, 7.5% to $5 million, 5% above. Each rate applies only to the slice of value inside its bracket. Source: /pricing.html.

The amount is measured by two published rules that your own finance team can reproduce without us.

**The value attribution rule (dollars).** The attributable share is the lower bound of what the evidence supports, computed from a pinned evidence pack against a baseline captured before the intervention. A finding must pass six gates before any value is attributed:

1. The finding was surfaced by EOS, not already open in your own queue.
2. The baseline predates the intervention. With no earlier baseline, the attributable share is zero.
3. The evidence pack is pinned and intact. The pinned evidence pack that enforces this is specified, not built yet.
4. The outcome is measured at settlement, not at recording and not at decision.
5. A named person on your side confirms the measured figure.
6. Confounders have been assessed and deducted.

The confounder deduction is deliberately unfavorable to us: no plausible alternative cause, 100%; one that cannot be quantified, 50%; one that can be quantified, 100% minus its quantified share; two or more that cannot be quantified, 0%; and a cause you name that we did not, your assessment governs. Never attributed: value from a finding you had already identified, a decision you would have made anyway and say so, avoided cost with no settlement event, run-rate projections, and a period whose evidence cannot be reconstructed. Source: /value-attribution-rule.html.

**The velocity attribution rule (time).** Two metrics are defined precisely. Organizational latency is the time from a discrepancy becoming detectable to a decision recorded against it. Decision velocity is the time from evidence being ready to a decision recorded by a named authority. Four gates apply: both timestamps come from a system of record; a pre-EOS baseline exists from the same customer; the comparison is like for like; and your named person confirms which clock started. Published averages use the median, state n, include only measurements that passed all four gates, and are never published from fewer than five customers. Figures publish from measured deployments only, and only with the organization's written consent. Source: /velocity-attribution-rule.html.

**The dispute path.** You challenge a measurement in writing. We reproduce the calculation from the pinned pack and show every input. A confounder you identify is deducted without argument. If we disagree on a quantification, your figure governs for that period. Nothing is billed while a dispute is open. Source: /value-attribution-rule.html.

**Ending.** You may end the term at any time, for any reason, with thirty days written notice; only the share earned on value already verified before the notice is owed. Price increases are capped in the master services agreement. There is no exit fee, and no notice is required to export. Source: /pricing.html, /exit.html.

## 8. The written guarantees, as the charter states them

- **Published first.** Every term was written and published before any negotiation. Every pilot runs against the charter, and results are measured by its definitions, not by definitions invented afterward. Source: /pilot.html.
- **Read-only.** No write permission is ever requested, because the read connectors have no write method to use it. EOS holds no write credential to your systems of record. A change is made only by your own executor, for a request a named person approved, under a signed, single-use authorization from Minerva, and EOS then verifies the result. The two-week assessment changes nothing. Source: /pilot.html, /pricing.html.
- **Definitions fixed in advance.** A finding is a contested figure where two systems of record disagree about the same object, each value traces to a dated record, the gap is quantified in currency on your own valuation basis, and your named authority confirms it was previously unknown or unresolved. Validated means your team confirms the figure and its basis in writing. Captured means cash or working capital movement your team attributes to the resolution, measured at day 90 against the pre-engagement baseline. Source: /pilot.html (The standard).
- **You keep what we deliver.** The register and the decision log are handed to you whichever way the engagement ends. Source: /pilot.html.
- **The refund.** The $35,000 pilot balance is fully refunded if the mutually agreed target is not achieved by day 90. Source: /pricing.html.
- **The warranty.** Attached to every pilot; the threshold is negotiated and the mechanism is identical for every customer. Source: /pricing.html.
- **Publication.** Nothing about you is published by default. With your written permission, your name or an anonymized profile; the finding and its measured impact only under that permission. Source: /pilot.html.
- **Your data.** Yours. We do not sell it, and we train no model on it. It is pooled across customers only if your organization's owner opts in, which is off by default, and a pattern shows only when at least 5 organizations contributed. Source: /pricing.html.
- **Independent assurance.** The charter states what exists: a read-only architecture with no write method, a public verification harness your security team can run cold and offline, and a SOC 2 examination and a third-party penetration test, both planned. No certification is claimed on any page until its report is issued. Source: /pilot.html (Security posture).

## 9. Next step

Request the assessment with the form on /request-assessment.html. A principal reads every message, and we reply within one business day with a scope and a date. The charter is at /pilot.html, the prices at /pricing.html, and the two measurement rules at /value-attribution-rule.html and /velocity-attribution-rule.html.

<!-- signoff -->
[Your name]
Web3 Ventures Enterprise

© 2026 Web3 Ventures Enterprise