AI for waterproofing businesses.

Commercial waterproofing is a trade where the invoice depends on work done long before anyone gets on site. Tender invitations arrive every day. Each one needs plans read, areas marked up, a membrane matched to the spec, rates applied and a quote sent. Most of those quotes lose. The ones that win carry a liability that runs for years, because a leak found after handover can cost far more than the job was worth.

That makes a waterproofing contractor an unusually good fit for AI. The work is high volume, it follows rules that can be written down, and the expensive mistakes are the ones a tired person makes at the end of a long week of pricing. It also means the wrong AI project, aimed at the wrong problem, is an expensive distraction. The difference is in where you start.

Where the time actually goes.

When we sit with a commercial waterproofing business and walk the process from tender to final invoice, the same picture comes up.

• Tenders arrive by email and get typed in again. Invitations land in a shared inbox. Someone reads each one and re-keys it into the job system and a spreadsheet. The job system has a leads section, but it wants a full account and site address before it will save anything, so real leads live in a diary instead.
• Choosing what to price is intuition. Nobody can say which builders, job types or sizes the business actually wins. So estimators price whatever arrives, and the business pays senior salaries to prepare quotes it was never going to win.
• Estimating is careful, manual and hard to scale. Plans go into a mark-up tool, the estimator traces what needs waterproofing, the tool measures the area, and rates from a cost spreadsheet turn it into a price. The spec names one of dozens of membranes across many brands, and matching it correctly depends on experience.
• Scopes of works get marked up on paper. A long scope is printed, marked up by hand by the most experienced person in the business, and typed up again by someone else.
• Quotes go out with no relationship behind them. Large quotes are sent to a generic tender address with no call to the contract administrator, whose name and number were on the invitation all along.
• Handover to the site team is thin. What the estimator knew about the job does not reliably reach the project manager or the crew.
• Payroll is checked by hand. Crews on an enterprise agreement care, rightly, about every allowance and every dollar. So someone senior reads every payslip every week.
• The numbers arrive late. The accounts produce a profit and loss weeks after month end. The job system already holds gross margin for every job, and nobody reports on it.
• The IT is holding the business back. Files on an on-premises server that only one person can open at a time, so versions drift. Passwords in a spreadsheet. One person who holds the logins and checks everything.

None of this is unusual. Walk through any light industrial estate full of trade contractors and the pattern repeats. What is unusual is how much of it owners have never looked at directly. Ask how a membrane order gets placed, or how an estimator picks which tender to price next, and the honest answer is often "good question". Those questions are where the value is.

Where AI earns its keep in a waterproofing business.

The strongest uses share a shape: frequent, rule-bound work, with a person who signs off before anything reaches a builder or an employee.

• Bid or no-bid. This is the single biggest lever. An agent scores every invitation against what the business has won and lost before, by builder, job type, size and location, and recommends which to price. When most quotes lose, pricing fewer, better-chosen jobs does more for revenue than pricing faster.
• Tender intake with no re-keying. Invitations flow from the tender portal into a lightweight CRM built around how the business actually sells, and one click creates the quote in the job system. Once intake costs nothing, subscribing to more tender portals becomes cheap.
• Capture from the car. A voice note ("remind me to call the builder about the data centre job next week") or a forwarded email to a dedicated sales address becomes a CRM task. The diary and the envelope retire.
• An estimating assistant trained on your own history. Past plans paired with the final quote that went out are the training set. The assistant prepares the take-off, proposes the membrane from the spec and drafts the price. The estimator checks and adjusts. The aim is each estimator pricing more work, not fewer estimators.
• Scope-of-works review. An agent trained on before-and-after pairs of marked-up scopes flags the clauses your most experienced person would have flagged, and produces the clean version without the retyping.
• Builder profiles. Some builders buy on price. Others want quality, detail and long guarantees. A profile per builder tunes margin, inclusions and how the quote is presented.
• A proposal that looks like it came from a serious business. Most waterproofing quotes are a bare price on a letterhead. Presenting the same content well is low effort and a real point of difference in a trade that rarely thinks about brand.
• A relationship cadence during tender. The CRM prompts the estimator to call the contract administrator before and after a large quote goes in, so a human voice sits behind the number.
• Payslip checking. An agent that knows the enterprise agreement checks every payslip before it goes out and flags anything that looks wrong. A person reviews the flags, not every line.
• An AI receptionist. Calls answered and triaged during hours, and covered when the receptionist is away, with messages landing in the CRM.
• Visibility that rewards people. Dashboards of quotes prepared, win rate and job margin per person, used for recognition and bonuses rather than policing. Measured output is also what makes flexible and remote work possible for office staff.

What has to come first.

Every use above depends on the data underneath it, and in a trade contractor that data is spread across a tender portal, a job management system, an accounting package, a QA and defects platform, a plan mark-up tool and a set of spreadsheets. The order that works is consistent.

• Keep the systems you have. The job system, accounts, tender portal and QA platform all stay. They are not the problem. The gaps between them are. Replacing them is a separate decision that can wait.
• Connect, land, then look. Connect each system, land its data nightly in one governed place, and build a basic dashboard of tenders, quotes, wins and job margin. That alone answers questions the business has never been able to ask.
• Expect the data to be messy, and fix it at the source. Client names and site addresses will be inconsistent. The useful response is a list of dirty records sent back to the people who own them, not a silent clean-up that breaks again next month.
• One new capability at a time. A trade business cannot absorb five changes at once. Plan, build, review, sign off and show the team one capability, then start the next.
• A person approves everything that leaves the building. Quotes, proposals, payslips and replies to builders all go through a named human. That keeps liability where it belongs and keeps the team in charge of the tools.

The two risks that decide whether it works.

Cash flow. On a large commercial job, most of the cost goes out on labour and materials well before the builder pays. A contractor that uses AI to win more work, faster, without funding in place can grow straight into a cash crisis. Overtrading is the classic way businesses like this fail. Model working capital alongside the sales pipeline before turning up the volume, because it limits the value of everything else you build.

How the team hears it. Crews and office staff will hear "AI" as "redundancy" unless they are told otherwise, clearly and early. The honest message is usually the right one: nobody is being replaced, the same people will do more of the work they are good at and less of the retyping, and the tools will be used to recognise good work. Say it before the first build starts, and show it at every monthly review.

Two smaller risks sit underneath. Key-person dependency, where one person holds every login and checks every quote, needs a business password manager and a recovery plan. And an on-premises file server needs to move to the cloud before an AI system can reliably read the documents on it.

What to demand from the firm that builds it.

• Will they walk the floor? The owner's view is one of several. A good partner interviews estimators, project managers, admin and the person who orders membrane, because the pain points owners do not know about are where most of the value sits.
• Will they work with your systems? A partner who opens with a platform replacement is selling software. Ask how they will connect what you already run.
• Where do your plans and quotes go? Your pricing history is your competitive advantage. It should stay in Australian regions, stay yours, and never train anyone else's model. Our piece on where your AI data can live covers what to check.
• Who approves what the AI produces? Every quote and payslip should have a named human approver, and the system should record who approved what.
• Will they start with the data? A partner who quotes an estimating bot before seeing your quote history is guessing. An AI readiness assessment tells you which of the uses above your data can support today.

Why RUBIX.

RUBIX has been an Australian data and AI consultancy since 2011. That is 15 years, 450+ projects and 115+ customers building governed data platforms and putting AI to work on them. We bring the same discipline to a trade contractor, at a scale and pace that fits one.

• We find the problems owners have not seen. Discovery is not a questionnaire. We walk the process end to end, from tender invitation to final payment, and ask the questions that surface the cost nobody is measuring.
• Data first, then agents. We build the data foundation that joins your tender, job, accounts and QA systems, then put AI agents to work on it, starting with revenue.
• Built for your business, not bought off a shelf. A lightweight CRM shaped around how you win work often beats a generic one nobody uses. We build what fits and keep what works.
• A steady monthly rhythm. One new capability a month, each one planned, built, reviewed, signed off and shown to the team, so the business improves without being overloaded.
• Senior people do the work. Our forward deployed engineers sit with your estimators and admin team, so the people who scoped the work are the people who deliver it.

If you run a waterproofing or specialist trade business and want to know where AI would pay back first, start with a conversation about how a tender becomes a job, not a demo of a tool.

General information only, not financial, legal or industrial relations advice. Current as at September 2026.