See the decision materials—not just the promise.

Pre-assessment: Meridian Fab

This is an illustrative fixture derived from reviewed public-evidence examples. It demonstrates the product format; it is not a claim about a real company.

Offering surface

• Custom fabricated assemblies — serves food and beverage plants and construction and architectural buyers.

Account and segment map

• Food and beverage plants — breweries, dairies, and processors needing custom stainless assemblies.
• Construction and architectural — contractors and architects needing structural or ornamental metalwork.

Top accepted signals

• Customer traceability requirements are tightening — regulatory pressure, rising trajectory, observed 2026-07-15.
• Capital projects increasingly require digital compliance attachments — economic pressure, stable trajectory, observed 2026-07-12.
• Experienced estimating talent remains scarce — supply-side talent pressure, rising trajectory, observed 2026-07-10.

The sample signal wording above is fictional. In a live pre-assessment, each item links to the exact public page that supported it.

Ranked business-priority questions

Scale traceable bid-package assembly

• Status: confirmed against current structure · medium confidence · horizon now.
• Implied demand: customers need quote packages with structured material, origin, and compliance evidence.
• Projection: the pressure lands on Custom fabricated assemblies and the RFQ-to-quote workflow.
• Touched graph nodes: RFQ to quote turnaround, senior estimator judgment, quote and final-cost history.
• Supply tension: scarce estimator capacity raises the value of augmentation while preserving human sign-off.

Establish reusable compliance-attachment capability

• Status: blind spot · medium confidence · horizon quarters.
• Implied demand: digital evidence packs are becoming part of the product rather than optional paperwork.
• Projection: no recorded process or asset currently owns reusable compliance attachments.
• Touched graph nodes: none — missing capability.

Seeded interview backlog

1. Which customers already request material-origin, inspection, or compliance attachments with a quote?
2. Who owns those attachments today, and how often do they delay submission?
3. Which fields can be assembled from existing quote, CAD, purchasing, and final-cost records?
4. What evidence would show that reusable compliance attachments are not a real priority?

How to read this output

The pre-assessment does not propose an AI project from public evidence alone. It starts with company-specific evidence and adds sourced external pressure only when that first read is too thin. Those hypotheses test whether outside pressure affects the company’s products or work and highlight the internal evidence still needed.

Illustrative memo structure

This is a sample document that shows the format of a diligence memo. It is not a claim about a real company.

Investment question

Can this company create differentiated value from AI within 12 months, or are the current claims mostly generic automation upside?

Headline findings

• Grounded near-term lever — one workflow shows repeated pain, reachable evidence, and a clear owner.
• Strategic upside — there may be a reusable capability, but it depends on rights to data and repeatable operating loops.
• Core uncertainty — management still needs to prove whether the workflow evidence is broad enough for portfolio-scale rollout.

Management questions that would most change the decision

1. Which customers or internal teams already feel the pain strongly enough to sponsor a pilot?
2. What data can the company actually access, retain, and evaluate?
3. Which operator owns rollout, approval boundaries, and post-pilot adoption?
4. Where would a failed pilot create legal, customer, or operational downside?

Portfolio comparison lens

First 100 days

1. Confirm one high-value workflow and the owner who can sponsor it.
2. Request representative data, approval constraints, and current process evidence.
3. Reject or demote generic AI narratives that cannot survive the evidence review.
4. Produce a focused pilot plan only for ideas that remain well supported.

What this document is for

The memo helps a deal or operating team challenge the claim, not merely admire it. It should expose where the company has a real path to AI value creation and where further management evidence is required.

Buyer question this answers

How do we avoid funding generic automation while still moving quickly on painful manufacturing workflows?

Why this roadmap is credible

• It starts from observed problems and available records, not an abstract brainstorm.
• It makes the order clear: measure first where evidence is weak.
• It protects larger bets by showing why they differ from common automation.

What to copy into your own assessment

1. Measure one recurring operational problem.
2. Link it to process owners and accessible evidence.
3. Turn missing evidence into named preparation work.

Start before the workshop with the sample pre-assessment.

This is an illustrative example. Atlas Ledger is a fictional company; the ratings, reasons, and value ranges below are Use Case Foundry's output for its company model, not claims about a real business.

The top three, scored

Value ranges are estimated from the recorded cost of the pain each opportunity addresses. Every rating can be challenged, and it changes when the evidence changes.

1. Audit and control evidence copilot

• Why it ranks first: the firm's books must be audit-ready, and the linked pain is high-severity and high-frequency: about 2,400 invoices a month, with duplicate payments and miscoded entries forcing client restatements.
• Company advantage: compliance capability is itself a barrier, and the copilot is grounded in the firm's own invoice and posting history.
• Before a pilot: agree the evidence each control needs and name the reviewer who signs off. Use Case Foundry has a tested pilot path for evidence assembly.

2. AP matching and GL coding agent

• Why it ranks second: the same $450,000-a-year pain, treated as a match, reconcile, and explain problem across invoices, purchase orders, and receipts.
• Company advantage: three years of the firm's own matching history, an existing close-automation project it can reuse, and the ERP it already runs.
• Before a pilot: the agent drafts postings only, and a controller approves each one. Price mismatches, duplicate invoices, and missing receipts go to an exception queue. Use Case Foundry has a tested pilot path for financial reconciliation.

3. Senior-controller copilot for the close

• Why it ranks third: the 10-day close costs about $300,000 a year and waits on senior controllers, who are hard to hire.
• Why it ranks below the first two: encoding scarce expertise gives a medium rather than high company advantage, and there is no tested production path for this pattern yet.
• Before a pilot: capture how controllers clear reconciliation breaks, and agree how junior staff's work is checked.

Buyer question this answers

Which finance AI work deserves a pilot first, and what has to be proven before it scales?

This ranking starts from internal company evidence. See how a first read is built before any internal evidence exists in the sample pre-assessment.

This is an illustrative example. Beacon Pay is a fictional company; the ratings, reasons, and value ranges below are Use Case Foundry's output for its company model, not claims about a real business.

The top three, scored

Value ranges are estimated from the recorded cost of the pain each opportunity addresses. The six highest-ranked opportunities all target the same KYC backlog, so Use Case Foundry groups them into one investment theme: the sponsor makes one funding decision and compares approaches, instead of approving six overlapping projects.

1. KYC compliance evidence copilot

• Why it ranks first: lending and payments are heavily regulated, and the linked pain is high-severity and high-frequency: manual KYC review is slow and costly at scale, and backlogs delay onboarding.
• Company advantage: compliance capability is itself a barrier, and the copilot is grounded in the company's own data.
• Before a pilot: the copilot assembles evidence and drafts findings, and a named compliance reviewer approves every case; it does not make the onboarding decision. Use Case Foundry has a tested pilot path for evidence assembly.

2. Onboarding qualification for applicants and merchants

• Why it ranks second: the same $1.2 million backlog, treated as a qualification problem across consumer applicants and online merchants.
• Company advantage: five years of labeled transaction and fraud outcomes, an existing capability it can reuse, and a place inside the onboarding flow customers already use.
• Before a pilot: decide which checks it may pre-fill and which stay with reviewers, and confirm how it connects to the core banking system. Use Case Foundry can produce the pilot package; there is no tested production path for this pattern yet.

3. AML reviewer copilot

• Why it ranks third: AML review is the scarce skill that KYC onboarding waits on.
• Why it ranks below the first two: encoding that expertise gives a medium rather than high company advantage.
• Before a pilot: capture how experienced reviewers clear cases, and agree how the copilot's suggestions are checked and explained for audit.

Buyer question this answers

Where should a regulated lender or payments company start with AI, and what must be in place before automation touches onboarding, credit, or fraud decisions?

This ranking starts from internal company evidence. See how a first read is built before any internal evidence exists in the sample pre-assessment.

Buyer question this answers

How do we prioritize support AI opportunities that improve outcomes, not vanity metrics?

Why this structure helps

• It keeps support leaders focused on resolution quality and repeatability.
• It exposes what evidence is needed before scaling automation.
• It reduces roadmap noise by removing low-differentiation ideas early.

Compare the internal roadmap with the public-evidence sample pre-assessment.

What readiness means

An AI agent is ready only when it has a clear job, representative tests, minimum permissions, required human approvals, safe failure behavior, monitoring, accountable owners, and a practical way to run before it touches live work. The adoption plan should already show the foundations, owners, controls, and rollout route around the selected use case.

The seven steps

Each step builds on the one before it:

1. Agent design — define the job, work loop, tools, permissions, and what evidence is still missing.
2. Starter code — add the service structure, test suite, missing-evidence list, and pilot package.
3. Safety review — check that the package is complete and safe enough for offline testing.
4. Offline test run — run expected, difficult, and unsafe cases without calling live customer systems.
5. Observe-only simulation — test behavior on approved historical or clearly labeled synthetic cases, with no live actions.
6. Supervised pilot plan — define pilot users, sample data, success measures, monitoring, stop rules, and approvals.
7. Production handoff — give engineering the integration list, operating owners, monitoring plan, rollback steps, and final approval checklist.

Tested pilot paths

Support copilot

The tested read-only Cloud Run path answers from an approved support corpus with source-aware retrieval and access controls. Each customer corpus, identity setup, and release needs its own acceptance testing.

Evidence assembly

The tested read-only Cloud Run path assembles evidence packets and gap tasks. It cannot declare a control effective or submit an attestation. The lasting reviewer workflow still requires customer validation.

Financial reconciliation

The most mature tested path adds lasting human review, PostgreSQL, application identity, multiple service copies, secure secret references, and monitoring. It proposes matches and queues exceptions but cannot post journals or move money.

What ships today

For supported patterns, Foundry can export real starter code with integration templates or bundled spreadsheet data where appropriate, configuration placeholders, safety controls, automated tests, and a handoff guide. The tested cloud paths apply only to the specific workflow and software release tested; they do not make every generated agent production-ready.

Why this matters

A convincing agent description is easy to generate. Use Case Foundry ties readiness to reviewed evidence, adoption planning, simulation, and the delivery package, so “ready” has a clear meaning: required tools are covered, risky actions are tested, human approval exists, and the test threshold was met.

What it does not do

Use Case Foundry provides readiness checks and starter code for supported patterns. It does not create customer credentials or provide universal one-click production deployment.