Skip to main content
Neurastruct
Download the overview (PDF)
Capability 2 · AI & Automation

AI that does the admin, not the deciding.

We build AI workflows the Neurastruct way — structure-first, human-reviewed, audit-ready. Messy inputs in (PDFs, emails, scans), structured outputs out, action only on rules you set and a person's tick. Fixed-price discovery, fixed-price build, optional run retainer. AU-hosted by default.

The Neurastruct Method: structure-first, human-reviewed, audit-ready — AU-hosted, every line reviewed before it goes out.

Book a consultationSee pricing

Proof we ship products, not slideware

The method that estimates a slab is the method that triages your inbox.

We don't just wire together tools you already own — we ship our own products, built structure-first and reviewed line by line. SlabLogic, our vertical SaaS for Australian concreters, is the proof. The same discipline underpins every workflow we automate here.

Book a consultation

Plain English

What's an agentic workflow?

A system that doesn't just answer questions, it takes actions on rules you set. Read a supplier price-list email → propose updated prices in your POS → wait for your tick → push them through. The 'agent' is the loop, not the chat.

Workflow automation is the same shape without the loop — ingest a messy input, produce a structured output, hand it back to a human for the decision. Agentic workflows add a controlled action step on top.

Example workflows

Here's what we'd build for your industry.

Two illustrative workflows. Both are pattern walkthroughs, not deployed case studies — the real workflow gets scoped against your tools and your messy inputs.

Example workflow — based on patterns we've seen, not yet a deployed case study.

Accounting & bookkeeping practice · BAS prep + supplier reconciliation

The problem. Quarterly BAS prep means downloading bank statements, MYOB/Xero exports, and supplier invoices that arrive as PDFs, scanned receipts, email attachments, and the occasional fax. The bookkeeper spends two solid days per client reconciling line items — most of it is matching, not judgement.

What we'd build. A workflow that watches the client's inbox (or a forwarding alias), parses supplier invoices on arrival, matches them against MYOB/Xero bills already in the system, flags mismatches, and drafts the GST coding for the bookkeeper to confirm. Everything is held in a review queue — nothing posts to MYOB until the bookkeeper ticks it.

Realistic outcome. Two days of reconciliation down to half a day, on a typical 100-supplier quarter. The bookkeeper still does the judgement calls; the AI does the typing.

Example workflow — based on observed patterns at owner-operator factories.

Small manufacturer · supplier-email parsing + job updates

The problem. A 12-person job shop in regional Victoria takes 30–50 supplier emails a day — price changes, ETA updates, 'back-order on the M12 bolts.' The production manager reads each one, decides whether it affects any open job, then walks out to the floor to tell whoever needs to know. Real cost: the things that don't get walked out on time.

What we'd build. An agent that reads each supplier email, cross-references the open job list, identifies which jobs are affected (by component code, by supplier, by delivery date), and posts a structured update — 'Job #4471 — M12 bolts delayed 4 days, affects Friday's assembly' — into the existing whiteboard channel (Slack, Teams, or a wall-mounted tablet). No invoices touched. No POs raised. Just the read-and-route work.

Realistic outcome. Most days the production manager opens the channel instead of opening 50 emails. Some days the agent's read is wrong and the manager corrects it. The wrong reads are useful — they show what the agent missed and what to fix next sprint.

How we engage

Discover · Build · Run.

A workflow audit first, then a fixed-price build of the automation, then an optional run retainer. Fixed-price at each stage; most clients start at Discover.

Discover

$4,500

2-week workflow audit. We learn your business, map the highest-impact automations, and hand back a written technical proposal with a fixed-price Build quote. No commitment to proceed. The fee is credited back against your Build, so starting isn't the expensive option.

  • 2-week engagement
  • Written technical proposal
  • Fixed-price Build quote
  • Fee credited back against your Build

Build

from $15,000

The actual build of the workflow identified during Discover. Tight 1-week iterations with weekly progress demos, hand-off documentation, and a 30-day post-launch support window.

  • Single-workflow $15k–$30k · Multi-workflow $25k–$80k
  • Fixed-price after Discover
  • Weekly progress demos
  • Hand-off documentation
  • 30-day post-launch support

Run

from $750/mo

Optional monthly retainer for monitoring, model updates, and incremental feature work. Three tiers depending on how active the system needs to be.

  • Entry $750 · Mid $1,500 · Senior $2,500
  • Monthly retainer, month-to-month
  • Monitoring + model updates
  • Minor feature work included
  • Cancel any month

Fixed-price AI packs

Pre-scoped automations that solve one named problem and ship in weeks — named price, named scope, no discovery fee. Inbox & Quote Triage is the flagship; Document Extraction, AI Readiness & Governance and Rescue round out the set. Book one from a free 30-minute call. Bigger or bespoke work is a custom Build with a paid Discover sprint.

Inbox & Quote Triage pack

6,500 dollars

Ships in 2–3 weeks

Incoming email read, sorted, and turned into a draft reply or a structured record before anyone on your team opens it — with a person approving before anything sends.

The problem

The same handful of questions and quote requests arrive dozens of times a week. Someone reads each one, works out what it is, and either replies or copies the details somewhere. It's constant, and it's mostly pattern-matching, not judgement.

What you get

A workflow that watches one mailbox, classifies each message into categories you define, and produces either a drafted reply or a structured record. It drafts; a person approves before anything sends or posts. This is the automation that's hardest to buy off the shelf — it lives inside your own categories, your tone, and your systems — which is why it's the one most trades and retail teams start with.

Intake from one mailbox

The workflow watches one nominated inbox or forwarding alias. Additional mailboxes are a separate scope line.

Classification into your categories

Messages get sorted into categories you define — quote request, supplier update, support question, and so on — so the right ones reach the right place.

Drafted reply or structured record

Per message, the workflow produces either a drafted response in your tone or a structured record (into a spreadsheet or CRM), depending on the category.

Human sign-off before it sends

The workflow drafts; a person approves. Nothing sends or posts on an unchecked pass. Send-without-review is a deliberate exclusion, not an oversight.

Not included

  • Fully autonomous sending without review — that's a deliberate choice. If you want send-without-approval later, it's scoped separately once you've seen the draft quality.
  • More than one mailbox — additional inboxes are a separate scope line.
  • Complex multi-step actions (raising POs, processing payments) — this triages and drafts; taking action on rules is a custom Build.

Upgrade path: Pairs naturally with Document Extraction into a shared review queue, which is the entry pattern for a multi-workflow custom Build.

Book a free 30-min callFull details

Document Extraction pack

6,500 dollars

Ships in 2–3 weeks

Your supplier invoices, quotes, or forms read and written straight into your accounting or job system — validated against your own rules, AU-hosted, with a review queue for anything the workflow isn't sure about.

The problem

A stack of PDFs, scans, and email attachments gets re-keyed by hand into a spreadsheet or your accounting system. It's slow, it's error-prone, and it's the same shape of task every single day.

What you get

The outcome is a document type off your team's plate, not a pile of raw text. The workflow reads one nominated type, pulls the fields that matter, checks each one against the rules you set, and writes the result into the system you actually work in — the reach the cheap off-the-shelf tools don't have. It runs AU-hosted, and anything below a confidence threshold you control goes to a review queue instead of being written on a guess.

One document type, handled end-to-end

Pick the document type that's eating the most time — supplier invoices, say. The workflow handles that type from arrival to structured output. Additional types are priced per type at scoping.

Extraction to your destination

Fields land where you need them: a spreadsheet, your accounting system, or a database. One destination is included; a second integration is a separate scope line.

Confidence threshold you control

You set the bar. Above it, the workflow proceeds; below it, the document is held for a human. The threshold is yours to tune — the workflow never writes a low-confidence guess into your system of record.

Review queue for low-confidence items

Anything the workflow isn't sure about surfaces in a simple queue for a person to confirm or correct. The corrections also show what to tune next.

Not included

  • More than one document type — each additional type is priced separately at scoping.
  • Write-back that needs a paid API tier on your side (some accounting systems gate this) — we'll tell you if that applies before you commit.
  • Hand-scribbled or heavily-degraded source documents may not clear a useful confidence threshold — we flag this honestly at scoping rather than promise accuracy we can't hit.

Upgrade path: The natural first step into a multi-workflow custom Build — e.g. invoice extraction plus supplier-email triage plus reconciliation, sharing one review queue ($25k–$80k band).

Book a free 30-min callFull details

AI Readiness & Governance pack

3,500 dollars

Ships in 1–2 weeks

A written, plain-English assessment of what's safe to automate in your business, what isn't, where your data would go, and the documentation an insurer or auditor will ask for.

The problem

You know AI could help, but you can't move until you know it won't put client data or your compliance position at risk. For professional services, allied health, and bookkeeping, that sign-off has to happen before any build — not after.

What you get

A written readiness report you can hand to your PI insurer or compliance person: what's safe to automate, what isn't and why, a map of where your data lives and who touches it, and a ranked list of automation candidates with rough effort against each — the kind of record an insurer, or an AHPRA, APRA, ATO or NDIS reviewer, will actually accept. It's an assessment, not a build.

Written readiness report

A plain-English document — no slideware — covering what in your day-to-day is a good automation candidate, what isn't, and the honest reasons for each. Written so a non-technical owner and a technical reviewer both get what they need from it.

Data-handling map

Where your data currently lives, what leaves your business and to where, and which parts would touch an AI model under each candidate workflow. This is the part your insurer and your auditor actually want to see.

Ranked automation candidates

The repetitive work most worth automating, ranked, with a rough effort band against each so you can see what's a small pack and what's a larger custom Build before you spend anything on scoping.

Governance documentation starter

A short, reusable record of what AI is used for, what data it sees, and the human-review step on each workflow — the paper trail that makes an AI engagement defensible to a regulator or a client.

Not included

  • Any build work — this is an assessment. If it surfaces work worth doing, the report becomes the scope for a fixed quote.
  • A formal legal or regulatory opinion — we're not your lawyer or your compliance auditor. We give you the map; sign-off stays with your professional advisers.
  • Penetration testing or a security audit of existing systems — that's a separate, specialist engagement.

Bundle credit: buy any AI pack or start a custom Build within 60 days of a Readiness assessment and the $3,500 comes off the price. You don't pay twice for the same thinking. (Mirrors the credited-to-Build logic on custom Discover fees.)

Upgrade path: If the report surfaces a workflow worth building, it becomes the scope for a fixed-price pack or a custom Build — no separate Discover needed for anything already scoped in the report.

Book a free 30-min callFull details

Rescue pack

from 4,500 dollars

Ships in 2–3 weeks

You bought an AI tool or automation that stalled, broke, or never landed. We audit what you've got, put a fixed price on the fix, then repair or rebuild the piece that matters — and hand it back as yours.

The problem

Someone sold you an automation and it didn't stick. It half-works, it broke after an update, or it never went live at all — and now no one's sure what it does or whether it's safe to keep running. You've already paid for it once and you don't want to write it off.

What you get

A scoped-per-case engagement: we find what's actually running and where it fails, give you a fixed-price plan for the fix, then repair or rebuild the highest-value piece — human-reviewed, AU-hosted, with an audit trail — and hand it over as yours. You approve the number before any remediation starts.

Diagnostic of what you have

We work out what's actually deployed, what it touches, and where it breaks — the parts running, the parts that never landed, and the data moving through it. You get a plain-English picture of the thing you already own, not a sales pitch for a replacement.

A fixed-price remediation plan

The diagnostic sets a fixed price for the fix. You see exactly what we'd repair, what we'd rebuild, and what we'd retire — with the number against it — and you approve it before any remediation work begins.

Fix or rebuild the piece that matters

We repair or rebuild the highest-value part — human-reviewed before anything goes out, AU-hosted by default, with an audit trail a reviewer will accept. We fix what's worth fixing rather than gold-plate the whole thing.

Handed over as yours

You get the working piece and its source, documented — yours to keep, no per-seat licence and no lock-in to us. If it needs minding after handover, that's a Hosting & Care retainer, not a condition of the fix.

Not included

  • A promise to save any specific tool — if the honest answer is that a piece is cheaper to rebuild than repair, we'll say so at the diagnostic.
  • Third-party licence or hosting costs — those stay with the provider you bought them from; we don't resell or mark them up.
  • Forcing a non-cooperating provider to release access or source — if a vendor won't hand over what we need, we tell you before you commit rather than promise a fix we can't reach.

Scoped per case: the diagnostic sets the fixed price for the fix, so you approve the number before any remediation work starts.

Upgrade path: Rescue is often the first piece of a larger custom Build, or steps onto an Hosting & Care retainer so the fixed thing stays fixed.

Book a free 30-min callFull details

Fixed-price rescue

Rescue — the AI tool that stalled.

A fixed-price, scoped-per-case engagement to audit and fix an AI tool or automation you already bought that stalled, broke, or never landed. We find what's running, work out where it fails, fix or rebuild the highest-value piece the Neurastruct way — human-reviewed, AU-hosted, with an audit trail — and hand it back as yours.

The diagnostic sets the fixed price for the fix, so you approve the number before any remediation work starts. Rescue starts at $4,500.

See the Rescue pack

Compliance-record automation — the record the regulator asks for.

A custom Build that produces the record the regulator actually requires — Medicare/MBS/DVA claim prep, AHPRA record-keeping, NDIS reporting, SafeWork/OH&S documentation — with a confidence-gated review queue and an audit trail an insurer or regulator will accept. This is not meeting notes; it's the artefact a reviewer signs off on.

AU-resident inference is non-negotiable here. A person keeps the final call, and every record carries the trail of how it was produced.

Talk to us about a compliance record

Price-book intelligence — your rates, applied automatically.

Your price book, applied automatically to any quote input — the pricing engine behind SlabLogic, generalised for any trade that quotes against a book. It plugs straight into an automation Build here: ingest your rate card once, and it prices extracted quantities, a spec, or a measurement set — confidence-scored and human-reviewed before anything goes out.

Talk to us about your price book

When AI Automation from Neurastruct is the right fit.

  • You're a small or medium Australian business (1–50 staff typical)
  • You have repetitive admin work that follows predictable patterns
  • You want fixed-price engagements, not open-ended hourly billing
  • Data privacy and AU residency matter to you
  • You want generic chatbots or 'AI strategy' PowerPoint decks
  • You're looking to replace your team rather than augment them

What's outside scope for AI Automation.

Things we deliberately don't sell — so you know what to expect.

  • AI strategy PowerPoint decks. We build working software, not slideware.
  • Generic chatbots. Every workflow gets tailored to your inputs and your tools.
  • Replacing your team. We automate the repetitive admin so people do higher-value work.
  • Hourly billing. Every engagement is fixed-price after the Discover sprint.
FAQ

Common questions about AI Automation engagements.

Do you work with non-tech businesses?

Yes — most of our work is with family-run AU businesses (trades, retail, professional services, allied health, manufacturing, in-house finance teams) that don't have an IT department. We bring the tech; you bring the domain knowledge.

How do you handle data privacy and AU residency?

AU-hosted by default (AWS Sydney). Tenant isolation. . We never send customer data to public LLM providers without a signed per-workflow consent record. See the Trust centre for the full breakdown.

What if you can't deliver what was scoped?

We use fixed-price engagements — timeline overruns are our risk to wear, not yours. If a Build proposal exceeds your budget, we split scope into phases. If something genuinely can't be built, we stop early and you only pay the Discover fee.

Can we skip the Discover sprint?

Yes if your scope is already clear (e.g. you've done discovery work elsewhere or it's a follow-up engagement). For first-time clients we strongly recommend Discover — it's how we earn the right to quote a Build accurately, and the fee is credited back against your Build, so it isn't a sunk cost.

Do you actually build your own AI products, or just wire up tools?

Both, but the products are the point. Most AI shops connect apps you already own. We've built our own vertical product, SlabLogic — concrete-estimating software for Australian contractors, launching soon — and we bring that same structure-first, human-reviewed method to every workflow. You get software that holds up when your accountant, insurer, or regulator asks to see it.

Ready to scope your first workflow?

The first consultation is free and takes 30 minutes. You'll leave with a clearer sense of what's automatable in your business — even if we're not the right fit to build it.

Book a 30-min consultation