Audit-grade reports.
From the messy data
your team already has.
Subfractal builds expert-system software that takes the data your operations team already exports — Airodis, CEMS, lab feeds, maintenance logs — and produces a draft regulator-ready report in the time it takes to make a coffee. Then your subject-matter experts review and ship.
Regulated reporting is the work nobody asked to do but everyone has to.
Senior environmental, compliance, and operations specialists spend half their week stitching exports into narrative reports. The data is right. The expertise is right. The bottleneck is the assembly.
The cost
A monthly tunnel report runs 14 hours of senior time from export to submission. Multiply by 12 tunnels. Multiply by every reporting category your operations team owns.
That time is being spent on data manipulation, not the analytical judgment your clients are paying for.
The trap
Generic AI tools fail here. They hallucinate against regulators. They don't understand your templates. They produce output your team has to completely rewrite.
What's needed is a system that reasons over your data, your templates, your historical reports — and gets sharper every cycle.
RIC March 2026 — drafted end-to-end in two minutes.
First production deployment is in flight with Assured Environmental on tunnel air-quality reporting. Phase 1 demo shipped April 2026: one tunnel, full report, generated from raw exports against the production template.
From airodis.xlsx to a 32-page draft, ready for senior review.
Inputs: monthly Airodis export, CEMS maintenance log, monthly Microcosm template. Outputs: a fully drafted Word document — 12 narrative sections, 24 tables, 23 chart figures — formatted to spec, audit-trailed, ready for the team to review and refine.
A different way of writing software for reporting.
Most automation breaks the moment your inputs change. Subfractal's substrate is built on DSPy — a declarative architecture that learns from your historical reports rather than depending on hand-written rules.
What good looks like.
We declare the inputs (your exports), the outputs (your template), and the voice (your historical reports). No procedural code that breaks when a column moves. No prompt-engineered chatbot.
Expert systems, not templates.
The substrate composes specialised reasoners — one per narrative section, one per analytical step — each backed by an evaluation harness measured against your domain corpus. Every output traces back to inputs.
Sharper every cycle.
As your team reviews and edits drafts, the system tunes against the deltas. Month two reads more like Assured than month one. Month six reads like the team's senior reviewer.
Expert systems for the reporting categories your clients already operate in.
The Phase 1 substrate ports across regulated reporting domains. Same architecture, different signatures, different evaluation corpora.
Tunnel air-quality reporting
Airodis exports → monthly compliance report. Phase 1 reference deployment with Assured Environmental.
Water + contamination monitoring
Lab data feeds + sampling regimens → quarterly statutory reports. Same substrate, different signatures.
Workplace health + safety
Incident systems + audit findings → board-ready risk reports. Audit trail integrated.
Mining + maintenance audits
Equipment registers + calibration logs → regulator-bound asset reports. Evidence-linked.
Sustainability + carbon
Operations data + framework templates → TCFD- and NGER-ready submissions.
Custom domains on request
If the shape is messy data → narrative report → regulator, the substrate fits. Discovery scopes it.
A joint venture between operating expertise and AI substrate.
Subfractal is the commercial vehicle that wraps co-built IP from HYPRCORP and Assured. Each party brings what they're best at; the JV monetises the combination.
- Sector brand and regulator-facing authority
- Existing client base across regulated reporting
- Domain expertise — what good looks like
- First-line client relationships
- Co-owned product layer + commercial IP
- Licensing operations and pricing
- Roadmap across regulated reporting domains
- Recurring licence revenue and enterprise contracts
- Engineering side: Expert Systems Consulting + Forward Deployed Engineering
- Human side: Organisational Transformation + Pastoral Care
- External or embedded — buyer chooses how we engage
- Each quadrant stands alone or composes with the others
Four quadrants. Engage one, some, or all.
HYPRCORP brings four distinct offerings to the JV. They organise on two axes — how we engage (external or embedded) and who we work with on your side (engineering or human). Each cell stands alone. Combinations are common.
Expert Systems Consulting
We encode your domain experts into a system that runs. DSPy substrate, declarative pipelines, eval harnesses tuned to your historical work. Not a GPT wrapper.
Fixed-scope engagement · from $50kOrganisational Transformation
Programme-shaped change of how the business operates around the new systems. Sequencing, governance, capability uplift — bridging the gap between AI shipped and AI used.
Programme fee · scoped per engagementForward Deployed Engineering
Senior engineering inside your organisation, shipping with your team — not over the fence. The build velocity behind every reference deployment HYPRCORP signs.
Day rate · $1.5–2.5k · or weekly retainerPastoral Care
Direct attention to the people whose work is changing as the systems land. The reason HYPRCORP's adoption rates beat the consultancy average — adoption is a metric, not a hope.
Retainer-embedded · or standalone monthlyThree layers — explicit, contractually clean.
Client work continues uninterrupted under existing terms. The JV layer is additive — productisation of what gets built, not redirection of it.
| Layer | Owner | Why |
|---|---|---|
| Client implementations | Client owns | Code, configurations, tuning, and templates shipped to a specific Assured client. Same as today's commitment — no regression. |
| Methodology layer | HYPRCORP retains | DSPy substrate, eval-harness architecture, signature patterns, expert-system templates. Reusable across engagements; preserves R&D incentive treatment. |
| Product layer | Subfractal (JV) | Subfractal-branded productised offering, pricing, licensing terms, sales collateral. Co-built. Co-owned. Co-monetised. |
From Phase 1 reference to Subfractal v1 — 90 days.
The current engagement is the foundation. Discovery and JV scoping run in parallel, in a structured sequence with explicit decision gates.
Phase 1 build-out
RIC reference deployment scales to all 12 tunnels. Existing Phase 1 engagement and weekly retainer continue exactly as scoped.
In flightJV scoping
Sector pattern validated across Assured's existing client base. Founding product scoped. Term sheet drafted — equity, IP carve-outs, governance, milestones.
Working session opens thisSubfractal v1
JV incorporates. Founding IP contributions go in. First commercial pilots ship to Assured's existing clients beyond the founding deployment.
Recurring licence revenueWorth a 90-minute working session.
Phase 1 work continues either way. The JV question is about what we build into on top of it.