Fixed-scope engineering for teams that cannot afford invented evidence, missed requirements or an opaque automation nobody can operate after handover.
WEDGE
Tender intake
BUYER
Bid and operations lead
SCOPE
One real tender
GATE
Human release
PRIMARY WEDGE / S01
From AUD 6,500
15 business days
Buyer fit, starting price, duration and processing target are validation hypotheses until agreed in writing, not promises of a tender win or autonomous submission.
Tender evidence pilot
Exact buyer
Victorian engineering consultancies with 15 to 100 staff, an existing Microsoft 365 or SharePoint bid library, and at least two substantial tenders per month.
Narrow outcome
Pilot hypothesis: after setup and source validation, a configured, conforming client-authorised tender pack can produce a page-cited compliance matrix, deadline and owner plan, clarification questions, missing-evidence list, and first-draft answers within one business day, grounded only in approved material.
01Authorised pack
02Obligation map
03Approved evidence
04Human review
+Tender-pack ingestion for PDF, DOCX and XLSX
+Page-cited mandatory-requirement matrix
+Deadline, owner and clarification-question register
+Approved-library evidence retrieval with source links
+First-draft responses that expose missing evidence
+Runbook, acceptance checks and operator handover
Operating boundary
Client-authorised files only. No restricted-portal scraping, invented credentials, bid-strategy decisions or automated submission. Human review before any submission. Retention, access and deletion rules are agreed before representative material is processed.
S0210 business days
AI workflow reliability sprint
From AUD 4,500
For an n8n, Python or LLM workflow that works in a demo but fails under operational pressure.
+One priority workflow rebuilt around explicit state
+Retries, deduplication, timeouts and failure visibility
+Human approval at the consequential-action boundary
+Tested rollback, runbook and handover
S035 business days
Agent architecture and risk review
From AUD 2,000
For a team deciding whether an agent, RAG system or automation is safe and economical enough to build.
+Architecture and capability-boundary map
+Privacy, failure-mode and side-effect review
+Costed implementation sequence and dependencies
+One focused spike for the highest-risk unknown
PROOF / MERIDIAN 33
Evidence before ornament.
The same architecture used in Meridian 33 binds claims to dated sources, runs deterministic and adversarial checks, and refuses release until a human approves the exact output. Its corrected v2 verification, completed 24 July 2026, recorded 146 tests and 29 of 29 adversarial checks across a 97-source evidence corpus.
A named owner, a real workflow, representative test material and authority to involve the people who operate it. Sensitive data can remain local or inside approved systems.
Not a fit
Unsupervised mass outreach, fabricated evidence, autonomous hiring or trading decisions, or a vague promise to add AI everywhere. Consequential actions stay visible and approval-gated.
NEXT STEP
Bring one real pack or one broken workflow.
Send a short description of what happens today, where it fails and what a successful result would change. I will reply with the smallest credible pilot, its assumptions and a fixed-scope price.
Starting prices are indicative and exclude third-party software, data, cloud, travel and tax. Scope, data handling, schedule and acceptance checks are agreed before work begins. Following a successful pilot, ongoing support starts from AUD 2,500 per month.