The Grout Guy, an Australian tile and grout restoration company that started as one man in a van in Western Australia, has deployed a fleet of AI agents that process work orders in two minutes — down from one to two business days. The company, now approaching 200 employees across six states, won a 2026 Salesforce Customer Success Award for the work. And its founders are now packaging the entire system as a new platform called Servly, designed to bring enterprise-grade AI operations to every small trades business in Australia.
This isn't just another "company uses AI" press release. The Grout Guy's story stands out because its leadership is refreshingly honest about what went wrong — hallucinated weekdays, unexpected content filters, confusing token pricing, and voice technology that still isn't good enough for customer calls. For a business audience trying to separate AI hype from reality, that honesty is more valuable than the headline numbers.
What Walter actually does
The Grout Guy's AI deployment is built on Salesforce's Agentforce platform, and it's not a single agent — it's a crew. As CTO Anthony Messina explained at Dreamforce, the company runs multiple purpose-built agents: Walter processes incoming PDF and email work orders, Casie triages and prioritises customer cases, Penny handles invoice payment reminders by voice, Mason manages out-of-hours cancellation requests, and Groutie runs the public website.
The headline result is Walter. Property managers and corporate clients send work orders via PDF. Previously, staff had to manually process each one — reading the document, checking for duplicate records, creating opportunities and service appointments in the system. It took one to two business days before the team even got to it, followed by another 10–15 minutes of manual processing per order.
Walter now does this in two minutes, around the clock. According to Salesforce's case study, the agent reads the documents, performs duplicate checking on accounts and contacts, and creates the full chain of records — opportunity, work order, and service appointment — automatically. Messina said the system has saved over 20 hours per week of manual data entry.
The scheduling and dispatching gains are equally striking. Before the Salesforce overhaul, the company needed one dispatcher for every four field technicians. That ratio has improved to around 1:25 or 1:30. The company has grown from 30 field technicians to over 130 in three years, with 40% year-on-year revenue growth and a target of 60% growth in the year ahead. Lead conversion rates have also improved by more than 20%, from the low 30s to over 50%.
The honest part: what went wrong
This is where The Grout Guy's story earns its place beyond a vendor case study. Messina was candid about the problems.
Early scheduling work produced incorrect days of the week — the AI simply hallucinated weekdays. Content safety guardrails unexpectedly flagged requests that mentioned "Grout Gal" as opposed to "Grout Guy." Australian slang confused the models. "I expected hallucinations," Messina told Mi3, "and I still didn't know how to deal with the hallucinations."
Testing was a revelation. "You go into it thinking, 'Oh yeah, no worries. We'll just write test cases because that's how we've always done it with software,'" Messina said. "But with AI, it could do something 99 times, then one time it won't because it's interpreted it slightly differently."
Pricing was another headache. Changing token cost structures made ROI calculations difficult, forcing the team to estimate consumption and benefits per individual workflow rather than treating AI as a flat-rate investment. Messina's solution was practical: build agents by use case with predictable token volumes, and use conventional code where AI isn't necessary.
"That's why we build agents by use case, rather than having one big agent that does everything," he explained. This approach — scoping each agent to a bounded task with knowable costs — is a lesson worth noting for any business weighing up AI deployment.
Voice remains the most visible gap. The Grout Guy uses AirCall for telephony and has a voice agent, but Messina said the phone experience was still too choppy or laggy for customer-facing use. Salesforce's voice functionality in Australia lags behind what's available in the US. Founder Brad Young summed it up at Dreamforce: "We want the pre-recorded demo to actually be real."
From internal tool to trades platform: Servly
Having battle-tested the system on their own business, Messina and Young are now launching Servly, a platform that packages Salesforce Field Service, Data Cloud, and Agentforce into a product designed specifically for small and medium trades businesses.
Unveiled at Agentforce World Tour Sydney in February 2026, Servly targets businesses employing two to 50 workers — the kind of operators still running on spreadsheets and phone calls. It unifies siloed data into a single customer view and deploys the same AI agents The Grout Guy uses: Mason, Walter, Casie, and Groutie.
Servly has secured a partnership with Salesforce that fewer than five other operators in Australia and New Zealand hold. The startup went through Spacecubed's AI Founder Sprint in Perth and is now progressing pipeline deals with several paid discovery projects underway.
"We've spent the time to make this work for our own business, and now we've built it so others can just switch it on," Messina told SmartCompany. "It's about being able to run a massive operation with a tiny office team because the agents are doing the heavy lifting in the background."
A market ready for disruption
The addressable market is enormous. There are 462,939 construction businesses in Australia, according to ABS data compiled by Master Builders Australia — more than any other industry — and 98.6% of them employ fewer than 20 people. Most are still managing operations through a combination of phone calls, spreadsheets, and standalone tools like ServiceM8, Tradify, or AroFlo.
Servly isn't alone in spotting the opportunity. Brisbane-founded simPRO launched Lightning in May 2026, embedding AI agents into field service software used by 24,000 trades businesses. Newer entrants like TurnkeyAI and Zatersio are also targeting Australian tradies with AI-powered quoting, scheduling, and invoicing. Meanwhile, platforms like Xero and MYOB are embedding AI directly into the accounting software these businesses already use.
The pattern mirrors what CSIRO found earlier this year: AI-adopting Australian firms aren't replacing workers — they're scaling faster with them. The Grout Guy didn't cut dispatchers. It made each one dramatically more productive.
What to watch
The real test for Servly is whether The Grout Guy's playbook transfers. Building AI agents for your own well-understood business is fundamentally different from packaging them for thousands of businesses with different processes, tools, and expectations. Messina has acknowledged as much — the Spacecubed Sprint shifted his thinking toward the end-user experience of working with agents rather than just building features.
Voice will be the next frontier. If Salesforce delivers the telephony capabilities in Australia that Young has seen demoed, the opportunity expands from processing paperwork to handling inbound customer calls — which, for a trades business, is where the money is. Every missed call is a missed job.
For business owners watching this space, The Grout Guy offers a useful framework: start with bounded, high-volume tasks where AI costs are predictable, use conventional code where it's cheaper, and don't deploy anything customer-facing until it's genuinely ready. The AI agents that cleared the work-order backlog are real. The pre-recorded demo isn't — yet.
Sources
- The Grout Guy's growth bet: Agents process work orders in two minutes — Mi3
- The Grout Guy grows 40% year-on-year with Agentforce Field Service — Salesforce
- Empowering the "agentic tradie" — SmartCompany
- From The Grout Guy to Servly: Building AI for the Trades — Spacecubed
- The Grout Guy wins 2026 Salesforce Customer Success Award — LinkedIn
- Australian Trades Industry Statistics (2026) — Orkkid (citing ABS / Master Builders Australia)
- Simpro Launches AI Agents for 24,000 Trades Businesses — CoterieLabs
