All Posts

What makes packhouse safety uniquely challenging A packhouse shares many features with a warehouse or manufacturing facility, but it operates under conditions that amplify the risk in ways specific to horticulture. The seasonal surge The defining challenge of packhouse safety is the speed at which operations scale up and down. A kiwifruit packhouse in the Bay of Plenty might run a skeleton crew for months, then ramp to hundreds of workers within weeks when harvest begins. Tha

One of the most common questions we hear from EHS professionals evaluating computer vision AI is: "How does this fit with the systems we already use?" It's a smart question. Most organisations have already invested in an EHS platform — tools like ecoPortal, SafetyCulture (iAuditor), Cority, VelocityEHS, Intelex, or Benchmark Gensuite — to manage their incident reports, audits, training records, and compliance workflows. The last thing anyone wants is another disconnected tool

When safety teams evaluate AI monitoring platforms, the pitch they usually hear sounds something like this: "Our system detects unsafe behaviours in real time and sends alerts so you can act before an incident occurs." That is not a bad pitch. Real-time detection is genuinely useful, and most of the platforms that deliver it do so well. But there is a question that pitch leaves unanswered: what happens after the alert? Because alerts, on their own, do not change behaviour. Th

Most organisations that face a WorkSafe prosecution, an OSHA citation, or a WHS enforcement action did not set out to cut corners. They had policies. Some had systems. A few had even run safety programmes. What they typically lacked was the ability to verify — in real time, across their operations — whether those systems were actually working. By the time enforcement becomes a reality, the commercial conversation has usually moved well beyond the fine itself. There are legal

Think about the last time safety genuinely improved at a facility you know. Not because a new system was installed, or a policy was updated, or leadership sent a company-wide email. But because something actually changed in how people worked. Chances are, there was a supervisor involved. Someone on the floor who started asking different questions, responding differently when something went wrong, treating near misses as useful information rather than things to be managed. Tha

If you have ever walked around a blind corner in a warehouse and felt your stomach drop as a forklift appeared out of nowhere, you already understand the problem. That near miss — the one that left you shaken but unharmed — is not a one-off. It is happening across your site, multiple times a day, and almost nobody is reporting it. Blind corners and intersections are the most predictable danger zones in any warehouse, distribution centre, or manufacturing facility. They are th

Why machine-on-person risk is so hard to eliminate Manufacturing environments are designed for production. Machines, vehicles, and people operate in the same space because that's what the process requires. The challenge isn't that machine-on-person risk exists; it's that it's built into the way manufacturing works. Vehicles sharing space with people on foot On most manufacturing floors, forklifts, tow tractors, and automated guided vehicles (AGVs) share the same space as work

In traditional safety management, the word "incident" usually means something bad happened — someone got injured, equipment was damaged, or work had to stop. But the most important events for preventing future injuries are the ones where nothing bad happened yet . The near misses. The close calls. The risky interactions that nobody noticed or nobody reported. In the world of computer vision AI, these moments have a name: safety events . Understanding what a safety event is —

Getting AI safety monitoring working well at one site is a genuine achievement. Getting it working consistently across five, ten, or fifty sites is a different problem entirely. Most organisations that have tried to scale a safety technology deployment know the pattern: the pilot facility runs beautifully, results are strong, leadership approves the rollout — and then everything gets complicated. Different facilities have different camera setups. Site managers have different

A WorkSafe inspector can knock on your door any time. Under section 168 of the Health and Safety at Work Act 2015 (HSWA) , inspectors have the legal right to enter any workplace at any reasonable time — no appointment necessary. Most businesses that have been through a WorkSafe assessment will tell you it was less adversarial than they expected. Inspectors are there to help you understand your obligations, not to catch you out. But that does not mean being unprepared is fine

There's a version of this that I've seen at dozens of sites. The safety manager announces a new incentive programme. If the team goes 90 days without a recordable injury, everyone gets a reward: a gift card, a team lunch, a bonus, a day off. The intention is good. The logic seems sound. Reward safe behaviour, and you'll get more of it. Three months later, the team hits the target. The gift cards go out. Everyone celebrates. And the safety manager quietly wonders why near-miss

Most warehouses have a traffic management plan. It lives in a folder somewhere, usually created when the site first opened or when someone last asked for it during an audit. The problem is that these plans are typically drawn from memory, based on assumptions about how forklifts move, and updated only when something goes wrong. That is a plan built on guesswork. And guesswork is not a strategy when forklifts are involved. The numbers make this clear. According to the National

What makes waste and recycling sites uniquely dangerous A materials recovery facility (MRF), transfer station, or recycling centre is a fundamentally different safety environment from a standard warehouse. The materials are unpredictable, the equipment is heavier, and the interaction between people and machines is more complex. The tipping floor: where most of the risk concentrates The tipping floor is the heart of any MRF or transfer station, and it's also the most dangerous

If you mention AI cameras in a room full of warehouse workers, the first question won't be about detection accuracy or dashboard features. It'll be: "Is this going to be used to watch us?" It's a fair question. And how you answer it will determine whether your computer vision AI deployment succeeds or fails — not technically, but culturally. Because even the most sophisticated safety platform won't deliver results if the people it's designed to protect don't trust it. Privacy

When we started building inviol, the idea of using existing CCTV cameras to detect safety events in real time felt like a stretch for most of the people we spoke to. Computer vision for workplace safety was a niche concept, the market was small, and most EHS teams were still managing safety through spreadsheets, periodic audits, and reactive investigation. That was only a few years ago. Today, computer vision AI for safety is a category. The global AI workplace safety market

Across inviol's customer base, vehicle-on-plant incidents reduce by an average of 61%. That's not a projection or a target. It's a measured result across real facilities, running real operations, with real forklifts and real pedestrians. When we share that number with prospects, the first response is usually some variation of "that sounds great, but how?" Which is the right question. A percentage on its own is meaningless without understanding what's being measured, why the r

If you run safety across multiple sites, you've almost certainly sat in a meeting where someone compared Lost Time Injury Frequency Rate (LTIFR) across locations and drew conclusions about which site is "doing well" and which one "needs attention." On the surface, it seems like a reasonable exercise. In practice, it's one of the most misleading things you can do with safety data. The problem isn't that benchmarking is a bad idea. Benchmarking is essential for multi-site organ

One of the most common questions I hear from operations and EHS managers evaluating computer vision AI for safety is: "How long is this going to take?" It's a fair question. Most people's experience with enterprise technology deployments involves months of scoping, lengthy integration projects, infrastructure upgrades, and a go-live date that keeps moving to the right. When you've been through that cycle a few times, scepticism about timeline promises is healthy. So here's t

If your organisation operates in more than one country (or if you're evaluating safety technology for sites across different jurisdictions), one of the first things you'll notice is that incident reporting requirements are not the same everywhere. The terminology is different, the thresholds are different, the timeframes are different, and the regulators you report to are structured differently. This can get confusing quickly, especially if you're an EHS manager trying to bui

Ask most safety managers how they measure safety culture and you'll get one of two answers: "We look at our injury rates" or "We do an annual perception survey." Neither answer is wrong, but both are incomplete. Injury rates tell you about outcomes, not culture. Perception surveys tell you about attitudes at a single point in time, not about what people actually do day to day. Safety culture is what happens when nobody's watching. Measuring it properly means looking at multip