The disruption potential of ONYX1

Most job management platforms sell trade businesses a growing pile of features on a bill that grows with headcount. ONYX1 is built on a different premise: software for a trade business should behave like a utility. Secure, predictable, priced like a fixed cost you can plan a year around, and shaped around the one loop every trade business actually runs. This article lays out the three shifts behind that premise, what they mean in dollars and in practice for a New Zealand crew, and where to check every claim, because nothing on this site is sold before it exists.

The problem: job software priced and shaped for someone else

The tools on offer to a small or medium NZ trade business come in two shapes. At one end sit cut-down apps that stop at invoicing and leave safety, stock and the crew's hours in spreadsheets. At the other sit enterprise platforms priced per seat, sold with a setup fee, a demo call and an implementation project, carrying modules designed for a head office most trade businesses do not have. Both shapes are billed as though the business were much bigger than it is, and both quietly punish the exact thing a healthy business does: hiring.

That gap is not an accident of one vendor. It is the economics of the category. Per-seat pricing maximises revenue per client, so the interface accretes modules to justify the next tier, and the bill climbs with every apprentice. The result across the sector is a long tail of three-to-fifteen person crews running their operations on a mix of Messenger threads, paper pre-starts, spreadsheets and a shoebox of receipts, not because they are behind the times but because the software on offer was never priced or shaped for them.

Shift one: flipping the per-seat economics

The per-seat model taxes hiring. At around NZ$52 per user a month, the going rate among per-seat tools sold to NZ trades, a five person crew pays roughly NZ$260 a month and a seven person crew is near NZ$364, before anyone has quoted a job. Take on an apprentice, a sub-contractor for the busy season, or a part-time administrator to get the invoices out, and the fixed overhead jumps the day they start. The damage is not only the invoice. It is the decisions the invoice pushes owners into: sharing a login, leaving the apprentice off the system, keeping the part-timer on paper. Every one of those puts a hole in the records, and the holes surface exactly when the records matter most, in an audit or after an incident.

ONYX1 charges one flat price per plan: NZ$79 a month covers up to five people, NZ$149 covers up to eight, NZ$299 covers up to fifteen, all plus GST. The same seven person crew that was near NZ$364 in seats pays NZ$149 with every capability intact. Hiring never multiplies the bill, and every person gets their own login with a role that fits, so the records stay whole. Jobs, customers and timesheets are never capped on any plan; what the plans bound is the genuine cost tail of storage, document volume and AI usage, which is the honest version of a limit.

The commercial bet is deliberate. Flat pricing makes less from a big customer than a per-seat vendor would, and that is the point. The under-served micro-to-medium trade businesses that abandoned or never adopted job software over cost are most of the trade sector, and a utility price is what brings them in. A business that would not pay NZ$364 a month in seats will pay NZ$149 flat, forever, without renegotiating every hire.

Shift two: a blueprint for safe AI on the tools

Most AI in trade software is shallow: a text shortcut that rewrites an email, pads a short note into quote lines, or summarises a document. Useful, but it changes nothing about how the business runs. ONYX1 takes a structurally different route. It connects the Claude or ChatGPT a business already pays for directly to the working system over MCP, an open connector standard supported by the major AI apps, so the owner runs the day by asking rather than by clicking through screens.

The difference that matters is where the rules live. What a connected AI can see and do follows the role of the person who connected it, enforced in the database rather than in the interface. A tech's AI cannot read a price even if it is asked nicely, because the permission does not exist at the data layer, and no prompt can talk it into existence. Ask what is on today, who is working right now, or which invoices are overdue, and the answers come from the live system, not from a model's guess. The system recomputes every total itself; the AI narrates numbers, it never invents them.

The guardrails go further than read access. New connections start with reads plus safe writes, and money moves are opt-in per connection. A sensitive action returns a preview and a short-lived confirmation bound to that exact request, so a confirmation for a small invoice cannot be replayed against a large one. Every tool call lands in the audit trail, and writes are idempotent, so a retried request never doubles up. This design has a pedigree: ONYX1's founder and CTO publishes research on keeping AI agents answerable to the people who run them, and the product applies that thinking end to end. It amounts to a working blueprint for safe AI in field software: not a chatbot bolted on top, but an assistant operating inside the same rules as everyone else in the business.

Shift three: the lean single loop

Feature creep is how construction software dies. Platforms add sub-modules for enterprise operations, the navigation grows a second level, and the screen an actual tradie holds in a glove at arm's length in the sun gets busier every quarter. ONYX1 refuses the trade-off. It does one loop, deliberately, and leaves the rest out on purpose:

• Capture: a phone call becomes a job card in seconds, and leads convert straight to quotes or jobs so nothing gets lost between the call and the work.
• Quote: a branded, GST-correct quote the customer approves online with one tap, no account and no login needed.
• Schedule: the crew sees the day on a board, a schedule, a Gantt or a live dispatch map, all views of the same jobs underneath.
• Work: photos, notes and parts logged on site from the phone, with signed safety forms gating the clock-in, so safety happens before the timer starts.
• Invoice: the finished job becomes a GST-correct A4 invoice in one tap, numbered the way the business already numbers.
• Paid: payments are recorded against the job, Xero-shaped exports leave on every plan, and Pro organisations sync quotes and invoices two-way over Xero's API, with a direction switch per document type.

Everything in the loop reinforces the rest. The quote knows the job, the timesheet knows the site, the invoice knows the signed-off work, and the books leave clean for the accountant. Because the loop is also what the AI drives over MCP, the same six stages are available by asking, which is the quiet payoff of refusing feature creep: a small, coherent surface is exactly what makes an AI assistant reliable on top of it.

What this looks like for a real crew

Take a seven person plumbing and gas shop. On per-seat tools it pays around NZ$364 a month, its apprentice shares a login, and its safety forms live on a clipboard. On ONYX1 Pro it pays NZ$149 flat: every person has their own login and role, the pre-start gates the clock-in, the office sees timesheets roll up for approval, quotes go out branded and GST-correct, and the owner asks their AI which invoices are overdue on the drive home. The books leave as Xero-shaped exports, or sync straight into the connected Xero organisation. The software bill stopped being a variable cost the day they moved.

Software for a trade business should be priced like a utility and trusted like one.

Check the claims

The pricing and plan limits live at onyx1.co.nz/pricing, including the comparison with per-seat alternatives. The AI model and its guardrails are documented at onyx1.co.nz/ai, and the security architecture, from database row-level tenant isolation to the audit trail, at onyx1.co.nz/security. The Xero story is stated plainly at onyx1.co.nz/product/xero. Every screenshot on this site is the real product, and the loop itself runs real Auckland trade businesses every working day.