See it before you sign up
From a problem to a decision. No account, no call.
Start with a business problem. Foundry suggests the candidate AI use cases — and where AI isn’t the answer, it says so — then evaluates each one. Everything below is computed by the real engine; it’s exactly what the free tool gives you.
The problem · Finance & accounting
“Month-end close drags on — staff spend days rekeying invoices and chasing odd transactions.”
Before you reach for AI: Much finance pain is process and data-quality — reconciliation rules, master data, ERP config. If the numbers are wrong at source, AI scales the errors rather than fixing them. Often a workflow fix comes before any AI.
Candidate use cases, evaluated:
Invoice & document data entry
Reads invoices, receipts and statements and pulls the fields into your finance system so people stop rekeying them.
Worth pursuing now.
Spend & expense anomaly detection
Scores transactions and expense claims to flag the unusual ones for a human to review.
Promising, but tighten it before committing.
Finance policy assistant
Answers staff questions about expense, procurement and finance policy from your own documents.
Worth pursuing now.
The problem · Customer service & support
“Support volume keeps climbing and customers wait too long for a first reply.”
Candidate use cases, evaluated:
Draft first-pass replies
Drafts a first response to routine enquiries from your knowledge base for an agent to review and send.
Worth pursuing now.
Self-service answers (deflection)
Lets customers get answers directly from your help content, deflecting the routine tickets.
Promising, but tighten it before committing.
Triage & route incoming tickets
Classifies and routes incoming tickets to the right queue with the right priority.
Worth pursuing now.
The problem · HR & recruitment
“Hiring is slow, and HR answers the same policy questions all day.”
Before you reach for AI: Anything that screens, ranks or filters people is a high-impact use of AI — it can significantly affect individuals and carries real fairness, transparency and record-keeping obligations. Go in expecting a full governance assessment, not a quick win.
Candidate use cases, evaluated:
HR policy assistant
Answers staff HR questions (leave, entitlements, policy) from your own documents.
Worth pursuing now.
Screen or rank candidates
Scores or shortlists job applicants against role criteria for a recruiter to review.
Promising, but tighten it before committing. It likely engages APP 1.7 transparency obligations.
A full evaluation, in detail
Loan-approval decision support
An assistant that helps loan officers decide whether to approve personal loan applications from an applicant's financial history — a high-stakes, credit-domain use of personal information.
Verdict
Refine
59/100
confidence: high
Promising, but tighten it before committing. It likely engages APP 1.7 transparency obligations.
How it scores
A moderate expected productivity benefit, adjacent to the strategy.
Data readiness partial, skills some, delivered by configuring.
medium confidence that employees will actually adopt it.
Value weighed against a ~$500–$800/mo run cost.
High-tier: viable, but expect a full governance assessment before deployment.
Adjacent to where the organisation is heading.
Risk preview
Governance tier: High · engages APP 1.7
- APP 1.7 trigger — significant effect on rights or interests
- High-impact categories
- APRA-regulated entity, material AI
Cost to own (indicative)
Run: A$500–A$800/mo
Build: A$20,000–A$80,000 one-off
- 1,000–10,000 / month × ~6,000 tokens/interaction on the balanced / mid model class (~$4/M tokens)
- raw inference ≈ $120/mo before ops overhead
- configure a platform carries a ~$300/mo non-inference baseline
Order-of-magnitude, to inform the decision — not a quote. Blended token rates by model class, as at 2026-09; excludes internal staff time and change costs.
Model guidance
Suggested class: Balanced / mid
AU-only residency: prefer a provider with an in-region option (e.g. AWS Bedrock au profiles).
What would change this
- Governance load is significant — factor an assessment and named owner into the timeline before committing.
- It sits away from the core strategy — sponsorship and a clear owner would strengthen the case.
- Adoption is uncertain — validating that the intended users actually want it would raise confidence.
Then govern it
What the Register produces
Decide to pursue a use case and it flows into the Register, which turns it into a defensible, standards-mapped governance pack:
Triage decision + full trace
The risk tier (Low / Medium / High) with every rule that fired — reproducible and auditable, never an opaque score.
Tier-appropriate assessments
Privacy impact, bias & fairness, human oversight, vendor and more — AI-drafted, then reviewed, edited, and approved by a named human.
Approval record
Who approved what, when, and on what basis — with any conditions tracked to completion.
Board-ready export
A PDF for the board or auditor plus a structured data file, containing only human-approved content.
AI drafts; a named human reviews and owns every approval — there is no “skip review” button. See the full workflow on How it works.
Sector example
A GP practice piloting an AI scribe
The risk level, the AI scribe checklist and the assessments it needs, worked out by the same rules the product uses.
Try it on your own problem.
Indicative decision aid, not legal or financial advice. Cost figures are order-of-magnitude; the risk tier is a preview of Obsignor’s governance triage, not a completed assessment.