Phase 1 — Quoting Automation.

Deterministic. Auditable. Priced with your own rates.

Prepared for Sean Muscat, Alectrite Electrical Services.

Real jobs run3 · all yours
Schedule capture126 / 126 · JEC Air
PricingReproduced to the cent
Measurement+0.0007% on a known grid
DeliveryFour staged milestones
CommercialsFixed price · §7
Summary

Executive summary

From our discovery meeting we found that your estimating function is the principal constraint on the business: a standard quote absorbs approximately six hours of senior time, larger tenders up to a week, pricing knowledge sits almost entirely with the owner, and quoting capacity is capped at roughly ten quotes per week against a target of twenty. You had already validated the underlying concept yourself — an AI producing the correct cable size and maximum demand to AS/NZS 3000 from one of your switchboard schedules. Phase 1 develops that into a working estimating platform: deterministic, auditable, priced with your own rates, and with a reviewer in control of every number that reaches a quote.

Following the pilot build, we have run three of your real jobs through the platform — JEC Air priced to the cent against your own itemised build-up; Coles Waratah processed blind end-to-end; and Mount Druitt PCYC, where blind machine counts matched your quantities 8-for-8 and repriced to your own costing to the cent (§2).

This document sets out the discovery findings the project is scoped against (§1), the results already achieved on your jobs (§2), the platform being developed and where the investment goes (§3), the working demonstration you will be given (§4), the delivery milestones (§5), data security and governance (§6), the commercial terms (§7), the stages beyond Phase 1 (§8), what we need from you (§9), and the success criteria (§10).

§1 · Discovery

Discovery findings

From our discovery meeting of 6 July 2026, we identified the following operating constraints. Each maps directly to a platform capability in §3.

Quoting throughputA standard quote requires ~6 hours; large tenders (~$800k scale) up to a week. Current capacity is ~10 quotes per week against a stated target of ~20, without additional hires.
Knowledge concentrationPricing is derived from experience-based benchmarks held almost entirely by the owner — approximately 1,000 installation hours for a new Coles-format store, with board and equipment pricing carried from repetition. This knowledge is a business asset that currently exists in one place.
Margin exposureA recent large project returned compressed margins traced to estimate refinement — a direct, quantifiable cost of the current process.
Document burdenProject specifications run 17–235 pages, the majority covering mechanical scope outside your package. Isolating the electrical scope is manual and slow.
Take-off effortManual scaling of cable runs is the single largest time cost — performed on mechanical drawings where the electrical runs are not drawn and must be inferred from equipment and board positions.
Terminology riskThe same board appears across consultants as MSSB, MCC, CMSSB or "mechanical control centre" — inconsistent naming that creates scope-reading risk on every tender.
Strategic constraintThe owner's capacity is consumed by estimating rather than business development, and operations slow materially when key staff are unavailable.
§2 · Proven capability

The pilot AI engine — developed and proven

The core engine has been developed and demonstrated in operation. All results below were produced on Alectrite job packages — your own drawings, schedules and pricing structures, not demonstration data.

JEC Air — 56–60 Albion St, Surry Hills

Full end-to-end run  ·  ✓ Proven

Schedule items captured126 / 126
Your build-up, to the cent$155,478.33
Error on a known 9,000 mm grid+0.0007%
  • Schedule read and an independent tag count agree exactly — 90 indoor units, 7 condensers, 29 fans.
  • Full 126-circuit MSSB schedule generated to AS/NZS 3000.
  • Drawing QA identified fan TEF-L1.01 scheduled but absent from the Level-1 plan — ≈$1,070 of scope, at your installed rates, that a drawing-only take-off would have missed.
PROVENANCE — one job (n=1) · counts verified two independent ways · pricing reproduced from your itemised mechanical build-up · re-confirmed on the current platform 20 Jul 2026.

Coles Waratah

Blind end-to-end run  ·  ◌ Indicative

  • A different consultant and a different drawing layout, processed with the equipment count locked blind — before any reference to your figures.
  • Priced end-to-end to an indicative figure; items that could not be priced with confidence were flagged for review, not defaulted to zero.
  • Final pricing pending rate-card confirmation with you (§9).
PROVENANCE — one blind run · count locked before comparison · price withheld here pending confirmation of the rates it stands on.

Mount Druitt PCYC

Blind reproduction  ·  ✓ To the cent

  • Run blind — your submitted quote and your workbook's filled quantities excluded from every input.
  • Machine counts matched your quantities 8-for-8.
  • The single variance traced to one distance band; deterministic measurement from the drawing's own coordinates confirmed it, and the repriced result matched your costing build-up to the cent.
  • The run paused with 30 items flagged for human review — nothing was auto-confirmed.
PROVENANCE — one job · a reproduction pass against your costing workbook and a blind pass with your figures excluded · dollar figures presented against the job file at the walkthrough.
§3 · Scope of development

What Phase 1 develops — and where your investment goes

Phase 1 delivers one system: job package in, reviewed and priced quote out. The proven engine in §2 is the foundation; Phase 1 builds it into the complete platform below — the workflow in plain terms first, then the engineering behind it, so the scope of what you are funding is clear and measurable.

The quoting workflow, as the platform runs it

Each step, in plain terms

  1. 01Job package in.Specification, drawings and schedules — including 200-page specifications. The electrical scope is separated from the mechanical volume automatically.
  2. 02Schedule read — exact.Equipment and switchboard schedules read directly from the drawing's data layer, character for character. Every read produces a report; a read that could drop data is flagged for review rather than passing silently.
  3. 03Independent verification count.A second, independent count of every tag on the plans, cross-checked against the schedule read. The two must agree, or the discrepancy is raised with the dollar value at stake.
  4. 04Drawing QA.Schedule-to-plan reconciliation on every job — the check that identified TEF-L1.01. Errors in the consultant's drawings become an RFI you issue, not a cost you absorb. MSSB, MCC and CMSSB are resolved by the terminology layer.
  5. 05Locate, measure, confirm.Your drawings are mechanical — the runs aren't drawn. The platform infers each cable run (equipment to board), presents it on the drawing, and the reviewing estimator confirms or corrects it before pricing — with your routing allowance and ~10% wastage applied to the confirmed metres, consistent with your current method.
  6. 06AS/NZS 3000 calculations.Maximum demand, cable sizing and the full board circuit schedule generated from the standard's own tables, every constant cited to its clause and edition — reproducible and auditable.
  7. 07Priced with your rates.Your own costing workbook becomes the platform's master rate card — 56 rates, validated line by line against your sheet — with margin applied the way your workbook applies it. Your pricing knowledge becomes a structured business asset.
  8. 08Review, then out.Everything lands in the estimator workspace: a review queue ordered by value at risk, single-action confirm or correct, immediate reprice. Out the other side: the run sheet, the quote, the Drawing QA report, and an Excel/CSV export with a provenance trail on every line.
Estimator workspaceAlectrite · Job: [pilot job] · Quote: $ — · 12 items in review
Review queue · by $ at risk
Drawing viewer · evidence highlighted
Item detail · confirm / correct
Drawing QA · schedule ↔ plan Export: quote + provenance (xlsx/csv)
Illustrative wireframe — the estimator workspace. The working build is demonstrated live in §4.

For the technically minded — this is the depth of what is being built and funded. The architecture below is how the platform is actually organised: an orchestrator agent coordinating specialist agents and engines, each with defined tools, with a human gate between the agents and any number that leaves the building.

Agentic harnessOrchestrated AI agents with defined tools, deterministic guardrails and human-approval interrupts — the agents cannot act on a price without a human decision at the review point.
Deterministic extraction enginesThe schedule reader, tag counter and reconciliation engine are engineered code operating on the drawing's data layer — exact and repeatable, not model guesses.
AI model layerA major commercial AI model on locked, reproducible settings with confidence thresholds; configured and tuned on Alectrite's terminology, drawing conventions and pricing structures, so it reads your documents the way your estimator does.
Rules and knowledge enginesThe AS/NZS 3000 calculation engine with clause-cited constants, and the versioned rate-card engine built from your pricing formulas.
Workflow and review layerThe estimator workspace, value-at-risk review queue, evidence linking, and an append-only record of every approval and correction with the named person who made it.
Observability and qualityFull tracing of every agent action and model request, run records per job, and a regression suite of 2,500+ automated tests that re-verifies the proven results before any change is released.

Developed within Phase 1

Live cable-run measurementThe measure-and-confirm workflow is built and verified in testing; proving it on live jobs — starting with your pilot job — is a Phase-1 development objective.
Layout generalisationExtending proven schedule capture across additional consultants' drawing formats.
Rate-card completionFull structuring and confirmation of your master rate card and pricing formulas (§9).

Deliverables — clear and measurable

§4 · The platform walkthrough

The working demonstration

The platform walkthrough is a live, supervised demonstration of the current build, run on your own jobs and operated by Zero Point — nothing is installed or left running on your side. It serves two purposes: it shows you the working platform at each stage, and it operates as a requirements session — we capture what your team needs at each point of the estimating and take-off process, and the Phase-1 build develops those features with your input.

  1. 01Your job in the estimator workspace.Your company's drawings, schedule and pricing on screen — live, not pre-recorded.
  2. 02The take-off, with evidence.Select any figure and the view moves to the exact region of the drawing it was read from.
  3. 03The review queue, ordered by value at risk.The most material open item first; anything the platform could not price with confidence is flagged for manual measurement rather than hidden as a zero.
  4. 04Confirm or correct.A single action; the quote reprices immediately, and corrections are retained for future jobs.
  5. 05Drawing QA.The schedule-to-plan reconciliation screen — the check that identified the missing fan — run on your job.
  6. 06The audit trail.Every agent action, model request and decision on the job is recorded and reviewable end to end — full traceability and observability across the agents we build, down to the individual read.
  7. 07The export.The quote out to Excel/CSV, with a provenance column on every line.
PROVENANCE — every screen above is the current working build, not a mock-up; the workspace ran a live on-site demonstration for our stone-industry client on 24 Jul 2026.
§5 · Delivery

Delivery milestones

Phase 1 is delivered in four stages. Each stage concludes with a review at which progress is demonstrated working on your own jobs before the next stage proceeds, and payment is aligned to these stages (§7).

Stage 1 Setup

Pilot job selected; rate-card and process-mapping sessions completed with you (§9). We take the job package from there.

Decision point: which job the pilot proves the concept on.
Stage 2 Pilot

The pilot job runs end-to-end: take-off, Drawing QA report, review in the workspace with our principal estimator, and a priced quote with provenance export. Results are assessed against the proof-of-concept criteria (§7).

Decision point: proceed to the Core build on the proven concept.
Stage 3 Core build

The platform is developed capability by capability, each demonstrated on your jobs at a scheduled job review.

Decision point: each increment reviewed before the next begins.
Stage 4 Launch

Your team begins using the platform, with training and guidance provided on operating it. Day-to-day access arrangements and ongoing running costs are confirmed at this stage (§7).

Decision point: go-live, and scoping of the next stages (§8).

Stages are sequenced by demonstrated delivery rather than calendar dates; at each review you will have full visibility of what is complete, what is next, and expected timing for the stage ahead.

§6 · Data & governance

Data security & governance

Data securityAll Alectrite data is held in an isolated, client-segregated environment with access restricted to Zero Point principals. There is no standing public deployment — the platform is operated in controlled, supervised sessions.
Local storageJob files, rate cards, quotes and correction history are stored in Zero Point's managed environment, segregated per client and excluded from the software codebase by an enforced technical control.
DatabaseThe operational registry — job records, review events and approvals — is a client-stamped, append-only store: every record carries the client identifier and the named person who acted.
AI processing & residencyDocuments are processed by a major commercial AI model under locked, reproducible settings. Operational run records, including the document content processed, are retained in Australian data centres (Sydney) with restricted access. Your data is not used to train public or shared AI models.
PrivacyAlectrite data is used solely to deliver the Alectrite platform. On exit, your data is returned and working copies are deleted.
GovernanceHuman approval is required for every rate, correction and sign-off — nothing in the platform can approve itself. Reads below the confidence threshold route to review rather than into a quote, and every model or configuration change re-runs a regression suite of 2,500+ automated tests before release.
§7 · Investment

Investment — fixed-price Phase-1 development

Phase 1 is delivered at a fixed price. There is no hourly billing.

Investment specification

Pilot — proof of concept on your chosen job (Stages 1–2)$7,500
Core build — Phase 1 quoting automation, list$35,000
Partner discount — first electrical vertical−$10,000
Core build, net$25,000
Pilot fee credited in full against Core−$7,500
Payable on Core after pilot credit$17,500
Total Phase-1 investment$25,000

Payment on the Core build is staged against delivery: 50% on commencement, 25% at mid-build, 25% on acceptance at launch. The partner discount is one-time and has a clear basis: Alectrite is our first electrical-vertical partner, and your jobs built the proof base in §2.

Ongoing running costs from launch are $1,500 per month, plus AI usage costs at actuals — confirmed once the completed platform's operating profile is known. Monthly costs vary with usage. There are no usage fees during the Phase-1 build.

Pilot terms — proof of concept

The pilot exists to prove, on a job you choose, that your quoting process can be automated end to end. If the pilot identifies that a required capability cannot be automated, the pilot fee is refunded. Where the concept is proven — as the engine results in §2 indicate — development continues on that foundation and the pilot fee is credited in full against the Core build. A decision not to proceed following a successful pilot does not attract a refund; the pilot deliverables — the end-to-end run, the Drawing QA report and the reviewed, priced quote — remain with Alectrite in either case.

§8 · Next stages

Beyond Phase 1

Your discovery brief extended beyond quoting. Once Phase 1 is delivered, the following are scoped and priced as their own projects — Phase 1 carries no obligation to proceed with them:

Phase 2 — switchboard procurementRFQs issued to your suppliers, your approval on every send, purchase orders raised and pushed to Xero. The procurement workflow has already run end-to-end on a real project for our stone-industry client — the platform's estimate landed within about 3% of the real supplier quotes (−3.0%, one job) — so this is an adaptation of proven capability, not an experiment.
Phase 3 — the operations layerJob-history analysis and loss review over your Xero data, FLIR thermal report drafting, a company knowledge base, and government tender responses — sequenced in the order that matters to you.

Commercial terms summary (to be formalised in the services agreement)

§9 · Working together

What we need from you

The platform is built from your knowledge of your trade. Much of that knowledge currently exists as experience rather than documented process — so we guide its capture, step by step, and it becomes both the structure the platform runs on and the material the agents are trained against.

Pilot job selectionA live or recent project where the correct outcome is known.
Rate-card development — guidedWorking sessions in which we help structure your pricing formulas — per-point rates, build-ups and margins — into the master rate card the AI prices from. You bring the numbers; we provide the structure and do the build.
Process mapping — guidedA step-by-step walkthrough of how you read a set of drawings: what you look for first, how you identify your scope, how you interpret cable runs, and what you check before you price. We document this with you, and the agents are trained on it — what to look for, what to verify, and when to escalate to a person.
Job reviews during the pilotScheduled reviews where we present what the platform has produced on your job, capture your feedback, and define the next development steps from those outcomes.
§10 · Success criteria

Success criteria

Phase 1 is measured against defined criteria, benchmarked to your current process times and validated during the pilot alongside industry data:

A platform shaped around your business

The platform is developed around the way Alectrite already operates — your terminology, your pricing structures, your review judgement — rather than requiring the business to adapt to generic software. Phase 1 delivers that foundation for quoting. The same architecture then extends across the business in the next stages (§8): documented workflows and automation in place of head-held process, visibility of every job's status and history, and — as the platform grows — field access, putting job, client and compliance information at your team's fingertips on site. The outcome is a business that is systemised, measurable and resilient: proven workflows, automation and agentic capability, tailored to how you work.

Next step

What happens next

  1. 01Choose the pilot job.A live or recent project where the correct outcome is known (§9).
  2. 02We run the platform walkthrough on it, live.The working demonstration, on your own drawings (§4).
  3. 03Stage 1 begins.Rate-card and process-mapping sessions with you, then the pilot runs (§5).

Reply to David · diannello@zeropointai.io