In July 2026, South Australia’s WHS (High Risk Construction Work) Amendment Regulations 2025 come into effect — and if you’re an electrical contractor working on roofs, you need to understand what’s changing.

What’s Changing

The threshold for “working at heights” drops from three metres to two metres.

That single metre makes a significant difference. Any task performed at or above two metres will now be classified as High-Risk Construction Work (HRCW) under the amended regulations. This aligns South Australia with the national model WHS laws that most other states already follow.

For electrical contractors, this isn’t just a regulatory footnote. It’s a volume multiplier.

Who’s Affected

If your work involves any of the following, you’re in scope:

• Solar panel installations on residential rooftops

• Rooftop HVAC electrical connections

• Antenna and communications equipment mounting

• Gutter and fascia lighting installations

• Switchboard upgrades in elevated plant rooms

• Temporary builder’s supply connections on multi-storey sites

The two-metre threshold captures the majority of single-storey residential roof work. A standard Australian home has eave heights between 2.4 and 2.7 metres — meaning nearly every rooftop electrical task now requires formal HRCW documentation.

The SWMS Volume Problem

Under HRCW requirements, a Safe Work Method Statement (SWMS) must be prepared before work begins. The SWMS must be specific to the task and the site — generic templates that worked at three metres won’t pass scrutiny at two metres.

For electrical contractors, this creates a compounding documentation burden:

1. More tasks qualify — Jobs that previously fell below the threshold now require SWMS

2. Site specificity matters more — Roof pitch, access points, overhead lines, and fall arrest anchor locations must be documented

3. Review cycles increase — Principal contractors (especially Tier 1) will scrutinise HRCW documentation more closely given the regulatory change

If you’re submitting the same template you used last year, expect rejections.

What to Do Now

1. Audit your current SWMS library

Review every template that mentions rooftop work, solar installations, or elevated tasks. Check whether the documented height threshold matches the incoming two-metre standard. Update references accordingly.

2. Map your common job types

List the 10 most frequent tasks you perform that involve elevated work. For each, identify:

• Typical working height

• Standard fall protection method

• Site-specific variables (roof material, pitch, anchor points)

This exercise surfaces which templates need the most work before July 2026.

3. Build site-specific detail into your process

Generic hazard lists won’t cut it. Your SWMS needs to specify:

• The actual fall distance at that particular site

• The exact isolation points if working near switchboards

• The fall arrest system configuration for that roof type

If your current process involves selecting from dropdown menus and clicking “generate,” you’ll need to add a site assessment step.

4. Brief your team

Anyone preparing or reviewing SWMS needs to understand the threshold change. Subcontractors especially — they’ll be the ones submitting documentation that you’ll be accountable for.

The Bigger Picture

South Australia’s amendment is part of a broader national harmonisation effort. The two-metre threshold already applies in most other jurisdictions under model WHS laws. If you work across state lines, this change simplifies your compliance baseline — but only if your documentation is genuinely site-specific.

Regulators aren’t looking for more paperwork. They’re looking for evidence that you’ve actually assessed the risks at each site and implemented appropriate controls. The fall height change increases the scope of what counts as high-risk. Your documentation needs to match that reality.

July 2026 will be here before you know it.

Start the audit now.

Sources:

• SafeWork SA — WHS (High Risk Construction Work) Amendment Regulations 2025

• Safe Work Australia — Duties Tool: Construction Hazards

On 11 February 2019, two labourers were removing steel handrails from the roof of a warehouse in Moorebank when the metal touched high-voltage power lines. One man, 25, fell to his death. His workmate suffered severe burns trying to help.

The two contractors involved, were fined $600,000 and $450,000 respectively for failing to comply with workplace safety obligations.

This was not an isolated event.

Every year, Sydney’s construction sector records around 12,600 workers’ compensation claims per year — roughly 35 serious incidents every workday.
The human and financial costs are staggering:

• 188 workers were killed from traumatic injuries in 2024.

• 146,700 serious claims nationwide required at least one week off work in 2023–24.

• That’s 400 serious claims every single day across Australia.

• Safe Work Australia estimates workplace injuries remove $28.6 billion from Australia’s economy each year.

Just imagine what could be built instead with that level of investment?

That’s equal to the budget of the entire Sydney West Metro line.

The Hidden Cost: Outdated Safety Workflows

Direct claims are only part of the burden. Lost time, insurance hikes, shutdown risks, and administrative drag all compound the problem.

Even though Australia is a leader in WHS, construction remains one of Australia’s highest-risk sectors, and falls from height are still the one of the 2 leading causes of serious injury and death in NSW. Yet the safety processes meant to prevent these incidents often lag behind the pace of work.

• Site managers report losing hours each week chasing signatures and digging up old SWMS files buried in inboxes.

• Manual SWMS workflows are error-prone, inconsistent, and difficult to audit.

• Compliance gaps — even small ones — can trigger fines, delays, or site shutdowns, and more critically, leave crews exposed to preventable hazards.

In short: the industry’s most essential safety documents still rely on slow, manual, fragmented systems.

A Critical Gap, Closed

RapidSafeSystems was built to address this exact pain point.

It’s an AI-driven construction safety assistant designed to generate accurate, site-ready Safe Work Method Statements in minutes, aligned with NSW codes and industry standards.

In this short demo, you can see how one can complete the crucial task of an electrical SWMS in a matter of minutes.

By digitising and standardising SWMS creation:

• Contractors eliminate paperwork bottlenecks

• Compliance becomes consistent and auditable

• Crews get safer, faster

• And organisations reduce exposure to the incidents that cost lives — and billions in productivity

For an industry under immense safety, regulatory, and cost pressure, RapidSafeSystems helps teams modernise the workflow and puts safety in the forefront using the Australian government’s laws and regulations.

Today’s guest, Wes Anderson, VP of Agronomy at SWAT Maps, believes this needs to change.

After 25 years in agronomy — from operating combines at 13 in his roots of Saskatchewan to mapping soil variability across four countries — Wes has a visionary take.

Precision agronomy should eventually just be called agronomy.

The Great Fertilizer Misconception

One would think high-yielding areas consistently need more nutrients.

But reality is often the complete opposite.

The biggest misconception Wes sees in the field?

“Fertilizer response is dictated by yield.”

represents a common misconception in precision agronomy, where high-yielding zones may thrive due to inherent soil potential rather than needing more inputs, while low-yield areas often require targeted nutrients.

Non-affiliated studies in Sub-Sahara confirm this statement.

For example, up to 25% of fields show non-responsiveness to fertilizers due to site-specific soil constraints like low micronutrients (Mn, Cu, B), disconnecting yield from fertilizer needs and emphasizing independent assessment of soil versus crop potential. Early reliance on yield mapping propagated this error, but addressing soil variability independently improves fertilizer efficiency.

The real ROI comes from understanding what the soil can provide versus what the crop actually needs, then managing that gap independently across every zone in the field.

Water: The Ultimate Differentiator

Fastest ROI often starts with water management over complex nutrient models, identifying zones that are too wet, too dry, or flood-prone to optimize conditions.

Precision water tools like smart irrigation reduce usage by up to 30% without yield loss, enhancing nutrient efficiency as proper water balance unlocks yield potential even in dry climates. The highest-yielding areas typically occur where water is "just right" minimizing losses from excess or deficit that cause nutrient inefficiencies like nitrogen volatilization. This presentation by Wes Anderson, goes into greater detail on this topic.

The practical applications are immediate:

• Acidic soils: Variable rate lime provides instant savings before yield gains

• Saline areas: Cut inputs where salt-affected zones can’t utilize nutrients

• Drainage issues: Address water logging that causes nitrogen loss and crop stress

The Logistical Reality Check

Here’s where precision ag theory meets farm reality. Wes showed me a 40-acre field with three distinct management zones, each requiring different nutrient strategies. The limiting factor? The farm was applying everything as a single blended product.

“Even though nitrogen rates might vary differently than phosphorus or potassium, when it’s all blended together, you can’t separate those,” he notes. This constraint forced uniform application across zones that needed completely different management.

But here’s the evolution: farms that start getting this soil variability data begin to see the value of separating products. “Yeah, it might add extra expense for application, but I’ll more than make that up by variable rating it on its own.”

Boots Beat Algorithms

The most refreshing part of our conversation was Wes’s emphasis on field time.

In an era of remote sensing and AI-driven recommendations, he’s adamant:

“This isn’t the type of stuff where you can just sit in an office and learn all you need to know. It does require some boots in the ground.”

Temporal variability alone demands field presence.

When spring is dry and crops are young, phosphorus deficiencies emerge if topsoil moisture is low — creating opportunities for deep nutrient placement that only becomes clear through tissue testing and field observation.

“You constantly measure and manage and relearn,” Wes explains. “There’s little nuances that occur with different seasons and different weather patterns. The way to learn those over time is just by physically being in the field.”

From Precision to Standard Practice

So what would it take for precision agronomy to just become agronomy?

“Wide adoption where it just becomes normal,” Wes says.

But more specifically: “We just need all agronomists to stop ignoring variability and actually acknowledging it, measuring it, taking time and effort to start diving into that.”

It doesn’t require every farm to adopt precision ag. It requires every agronomist to stop treating field variability as noise instead of signal. The technology exists. The ROI is proven. What’s missing is the fundamental shift in how we approach crop management.

After 25 years of watching precision ag evolve from yield mapping to comprehensive soil-water-topography analysis, Wes has learned more in the past decade than the previous 15. The learning curve is steep, but so are the returns for those willing to measure what matters.

Whether you’re growing wheat in Saskatchewan, cotton in Australia, or citrus in California, the fundamentals remain the same:

1. understand your soil,

2. manage your water,

3. never underestimate the value of walking your fields

Key Topics

• Getting into agronomy: From Saskatchewan farm kid to agronomy expert

• Core lessons learned: 25 years of precision ag evolution

• Precision becoming standard: When precision ag just becomes agronomy

• First common misconception: Fertilizer response vs yield correlation

• Biggest ROI gains: Area-specific opportunities from lime to drainage

• Data quality gaps: Farm logistics limiting precision implementation

• Blending models with fieldwork: Why boots on ground still matter

• ROI improvement advice: Starting with low-hanging fruit and water management

• Accessible expertise areas: Broad acre grains, cotton, rice, and beyond

Getting Into Agronomy

Aris: Wes, I want to ask you first, how did you get into agronomy? What led you to this path that you’ve taken? Can you walk us through from as far back as you want to to now?

Wes: Yeah, well, it started pretty early I guess. I grew up on a farm in East Central Saskatchewan near Yorkton. So it’s in Canada, mixed farm, but predominantly grain focused I guess. And so many years riding in a tractor combine with my dad and I mean as of the age of 13 started operating combine myself harvesting crops and starting to understand influence of management on crops and that always really interested me.

So yeah, I guess just as I grew up and got my degree in agriculture university, that field of essentially the science of growing a crop, which is what agronomy basically is really intrigued me more and more and got into it after university. So, I’ve really been into it for about 25 years now and yeah, it’s been good at always learning new things.

Core Lessons Learned

Aris: So if you were again to look back at those years, is there like a core set of lessons that has shaped how you approach nutrient management in general and just soil variability today?

Wes: Yeah, I don’t know if I could really say that exactly, but it’s been just a progression through my career. So certainly early in my career, just after I finished university, the precision ag tools really almost didn’t exist, right. Like it was very, very early. There were a few progressive farms just starting to yield map and which gave us the ability to truly understand just how much yield varied.

And that progressed into then some satellite imagery which I started using about 10 years ago or so in my career actually in rice and cotton systems when I first worked in Australia. But you know, it really hasn’t been until the last 10 years or so that these technologies have really picked up. And I think the learning curve has been steep and I think since starting with SWAT Maps in particular where we’re focused on mapping soil and water variability, you know, I’ve learned more in the past 10 years than I did the previous 15 for sure. Like it’s just been a massive learning experience, which is fantastic.

Precision Becoming Standard

Aris: Then you know, I want to focus a bit on your bio in a second. Take a quick look at that because you mentioned something that caught my attention and I reached out to you, which is that you said that precision agronomy should eventually just be referred to as agronomy. So what would it take on like a high level to get to that point?

Wes: That’s a good question. I think just, I guess you know, wide adoption for one thing, right, where it just becomes normal. I think that’s probably the simple answer. The real answer or more in depth answer would probably be just a lot more agronomists getting into precision ag and actually measuring variability and using those tools more regularly, which still falls under adoption, I suppose, but it doesn’t even necessarily take all farms to be using precision ag. We just need all agronomists to stop ignoring variability and actually acknowledging it, measuring it, taking the time and effort to start diving into that a little more and understanding it.

First Common Misconception

Aris: It makes sense from what I’ve seen because I have in the past worked on farms myself. I want to kind of zoom on today’s topic, you know, it’s again maximizing ROI. We just want to have a practical take going forward with these series in general, and I want the audience to just really see that we’re not just talking fluff. We actually have something of value to add. And what is that first misconception you see from growers or agronomists in general?

Wes: Yeah. Well misconceptions you know in the precision ag space I guess is that one big one would be fertilizer response is dictated by yield, right. So if areas of the field that yield more should need more fertilizer or more nutrients and that’s definitely not always the case. In fact, sometimes it’s completely opposite.

So that is a common mistake that unfortunately, I think early on, in the early days, again, going back many, many years when really maybe one of the only tools we had was yield mapping. You know, that was one of the things that people started doing. But the reality is, you have to disconnect yield potential from what we often broadly or I often broadly just call soil potential. In other words, what can the soil provide the crop versus what is the crop need? And we have to look at those two things independently to figure out what something like fertilizer response really is in all parts of the field.

Because the reality is that sometimes high yielding areas of the field are high yielding because the soil can provide everything that the crop needs. Sometimes areas are low yielding because they need more nutrients. So the ROI is related to that yield increase we get from added inputs, whether that’s lime, fertilizer, even seed.

Biggest ROI Gains

Aris: Since we’re going a bit more into the nitty gritty, I wanted in practical terms to see maybe where actually the biggest ROI gains are coming from right now. I know there’s some data that you prepared, you know, where we can talk a bit about nitrogen additions, phosphorus allocation and or fixing, you know, drainage or pH issues.

Wes: To answer your question, the correct answer would be this is a very area specific thing, right? I see data and work with partners, you know, doing precision agronomy in four different countries around the world, right? So it can vary massively in areas with pH problems like acidic soil problems, lime is definitely a big ROI. It’s a very instant, often savings product. So even before we get yield increases, you’re almost instantly saving on just product, right? You’re applying less because you’re only treating areas that have acid soils for example.

In areas like Western Canada where I’m from where soil salinity is a problem, again, you can get very instant and quick savings by cutting back inputs on saline affected areas because those areas just don’t need more fertilizer. I’m in Australia currently and variable rate gypsum is a really good one too. There’s certain areas of the field need gypsum.

So that is a very, you know, it depends where we’re talking about, but the example here on the screen would be a lime example that’s fairly typical, I suppose where you know, most of the field maybe needs a pH amendment. In that prescription example to the top there, it’s essentially every management zone is getting accept zone 1. But you know that rate can vary quite substantially, right. So those rates are in pounds per acre. So we’re going as high as 4300 lbs per acre, couple tons per acre basically and that’s dictated by the buffer pH in the soil test through that second column. And then in this example, this field had some fairly high exchangeable sodium in zone sort of seven through 10. And in that case we’re just, we’re almost just adding calcium to displace sodium more than anything, but it’s also going to correct pH as well. So that’s just one example of many I could give of variable rate lime.

Data Quality Gaps

Aris: Going into my next question, I want to talk about, you know, the emphasis I guess of quality soil water, but also topography data. You know, you definitely take into consideration like zone mapping in your assessments and your recommendations. What specific gaps do you see today that limit ROI and how do you think those can be bridged with your knowledge?

Wes: Yeah, I would say common one would be just farm logistical constraints. Like to give an example like this is just a fairly small field. I believe it was only 40 acres or so if I recall. And just an example, I mean, there’s three soil tests there from three different zones sort of showing all the extremes in this particular field. And I’m not going to dive in all the details. The point is, I’ve outlined a few sort of key, you know, production limitations, we’ll call them in the three different zones and they vary a lot, right?

And so there’s all these different things that we could try and manage better. But then it often comes down to, well, what can the farm do? Is there equipment capable of handling this sort of resolution, a map, you know, what’s realistic? Can we vary a product that much in that, you know, short of space, all that sort of thing. So yeah, it’s often the farm limitations.

I mean, true story, this farm was really applying all their nutrients pretty much as one single blend, right? All products blended together. So even though maybe nitrogen rates might vary differently than phosphorus or potassium or even sulfur or something, you know, when it’s all blended together as one product, you can’t separate those, right? They all have to go together. So those are real problems.

A lot of the time when maybe we start with farms down this path, that is a common scenario. But eventually they start to by getting this sort of data, they go, well, yeah, you know what, it would make sense to do this and separate out sulfur from the other nutrients and apply it differently because yes, it might add a little extra expense for application, but I’ll more than make that up by variable rating it on its own and managing it on its own. So that’s just one example.

Blending Models with Fieldwork

Aris: How do you blend the model outputs with what you learn from walking and mapping the fields? And like with any additional testing you might be doing, like tissue testing and how does year to year variability, you know, change things? Can you kind of like walk us through, I guess overall through like the process and you know how it changes your approach?

Wes: Yeah, well, I always say like you start by collecting data and then you constantly measure and manage and relearn, right. And yeah, you touched on that sort of temporal variability from season to season. You can get totally different weather that actually affects things, but by continually, you know, walking in fields and checking fields and going back to specific spots that maybe are problematic and maybe taking something like tissue test, for example, right, we start to learn what drives maybe a deficiency.

So for example, phosphorus. Well, we’ve learned that when spring is dry and the crop is young and that younger phase, if the topsoil is dry where a lot of the phosphorus is, we often see a lot of deficiencies or have the potential to get fairly big responses to that nutrient if it’s applied deep. So there’s little nuances that you know, occur with different seasons and different weather patterns. And the way to really learn those overtime is just by physically being in the field. This isn’t the type of stuff where you can just sit in an office and learn all you need to know. It does require some boots in the ground.

ROI Improvement Advice

Aris: And that’s a breath of fresh air to mention here because you often see, especially on like professional social networks like LinkedIn, people advocate for this silver bullet, you know, mapping, yeah, that’s all you need. And I am, you know, probably fed up and probably you’re more fed up than me because it’s just, it takes really knowing the land.

I want to kind of segue from this because you did mention variable rate practices are front and center really, at least with your philosophy and how agronomy should be approached. So what is one example I guess we can lean in on from what it would take for you to improve your ROI? If I’m a farmer coming to you and I want to start investing, how would you approach it? And if I want to see ROI next year, will I see it? What’s the advice you kind of give most often? And maybe can you walk us also through in this layered question, what do people expect to see as well?

Wes: It’s a good question and again, very farm specific, but we like to start with the low hanging fruit. And really, really before that, like you have to understand like you don’t know what you don’t know, right? And that’s in order to sort of know what the low hanging fruit is, we have to start with doing some fundamental data collection, right? For us, that’s using some proximal soil data and topography and elevation and understanding where water flows across the landscape, right? Those are the things we start with, then layer in soil tests and we start from there.

So yeah, that low hanging fruit can certainly vary a lot. But I think the thing that surprises a lot of farms is my first discussion is often just about water. Where is it too wet? Where is it too dry? Where does it frequently flood out, things like that. So a lot of time the initial discussion is just about water management, potentially drainage, whether that’s like subsurface tile drainage or just some very basic surface drainage.

You know, we’ve got a map up here as an example, the SWAT map on the top just showing some drainage lines of water. And then the bottom one is more what we kind of refer to as a drainage plan. So those rainbow colored polygons in the depressions that you can kind of see, they sort of tell us how deep that depression is or how deep the water would potentially be should it flood. And so, you know, when we look at this, some of those depressions are in an otherwise really nice productive soil. They’re just at risk of flooding out in a really wet year. And some of the other ones on the other hand, are in a different SWAT zone and they actually have some legitimate soil problems.

So even if we were to drain them, they might not be all that productive. So it’s understanding things like that helps us guide that discussion about water. And you know, do we need to first work in drainage because I will go on the record saying that the highest yielding areas of a field in any given year are where the water is just right or the best. So it’s neither too wet nor too dry even in dry climates. And I’ve worked in a lot of dry climates like both in Australia and Canada. Even in dry climates, you can often run into situations where there’s at least temporarily excess water and that water logging causes problems, whether that’s nitrogen loss, stress in the crop, you name it, and it causes yield loss.

So it’s yeah, it’s not uncommon for us to see areas of the field that are water limited and losing yield due to that limitation, but also actually, you know, getting too much water at times and causing yield loss due to that. So efficiency of nutrient use and yield all starts with water.

Accessible Expertise Areas

Aris: My next question will be also the last question. I understand very well now that if you know and you made it very clear that water is a huge component in successful, you know crop growth, it has to be just right. So, you know, we definitely have other variables with, you’ve made them very clear throughout this presentation. So my goal is to make you more accessible to the audience. Like I want people to know, you know, where your expertise lies, you know, what kind of crops you work with, what kind of climates you work with such that they can start looking at agronomy as not an add on, but rather a necessity for them to have like a positive growth plan with their farm, you know, 5-10 years down the line and even further. So are there any, let’s say particular farms or crops or climates you want to welcome to your practice?

Wes: Yeah. Well, I’m, I guess I’m most familiar and most trained and experienced in a lot of the what we would call broad acre small grains. So wheat, barley, oats, canola, you know, peas, lentils, all those sorts of small grain legumes that are growing. But I also have a decent amount of experience in irrigated cotton and rice production in Australia. So, and really this sort of technology that we’re talking about is applicable in all of them.

I’ve even since being in Australia have even mapped some, you know, citrus orchards, vineyards, soil. At the end of the day, it doesn’t really matter what you’re growing in it. All those fundamental properties that we need to understand to grow a successful crop, whatever that might be, it’s all those fundamentals are the same right. Management obviously can be a lot different in something like citrus versus wheat, but yeah, we can work with it all. It’s, we’re just trying to understand soil and water.

Closing Remarks

Aris: Well, I definitely see the value of precision agronomy and I hope for its adoption further. I just want the audience to keep note of this advice and to reach out to you. I will make sure that people get this message. If there’s anything else you want to share, or if you feel like there’s something that people should know, feel free to mention it now.

Wes: Yeah, I’m happy to chat with people if they’re early in their precision ag path, certainly. I mean, I like to share that experience and mentor people and they’re early in their career if they want or if they’re even mid or late career, that’s fine too.

Aris: Absolutely. Thank you, Wes. I really enjoyed the conversation and I hope to talk to you again very soon.

Wes: Thank you.

The question isn’t whether the technology works.

It usually does. The sensors detect. The models predict. The dashboards display. Pilots pass. Demos impress.

The question is whether the system — the full stack of tools, workflows, people, and pressures — produces an outcome anyone would call favorable. Not in a controlled environment. On a job site running behind. On a farm where the margin doesn’t forgive experimentation. In the hands of someone responsible for fifteen things who has to decide, in the moment, whether this is worth the friction.

That’s a systems question. And systems don’t fail in demos. They fail under load.

Construction

In Sydney’s construction sector, sites log roughly thirty-five serious injury claims every workday. Each one a person who went to work and didn’t come home the same.

Safe Work Australia estimates workplace injuries cost the New South Wales’ economy $28.6 billion annually — the equivalent of an entire metro line, gone. Money gone to compensate workers and grieving families.

Tools everywhere

Compliance platforms. Documentation systems. Digital checklists.

Technically adequate. Often ignored.

Not because crews don’t care.

Because a foreman stretched across fifteen problems, on a site three days behind, under cost pressure, doesn’t have the cognitive bandwidth for a system that adds steps. The paperwork gets filed. The hazard doesn’t get seen.

As a result, safety standards plateau. The nationwide fatality rates tells the story:

Nine years. The line hasn’t moved.

This isn’t a tool failure. It’s a system failure. The tools were never the constraint. The constraint is the conditions under which someone has to use them.

Agriculture

Model meet Reality

John Deere is a $130B company. Its See & Spray system — trained to identify weeds in early-season fields — required retraining after failing mid-summer in the same US farms. The crops had grown. The model hadn’t accounted for it.

Fahad Khan, head of product management for Blue River, John Deere’s platform handling data put it plainly:

“Sometimes it’s impossible to identify a weed from a crop,” Khan said. “If you ask my 50-plus ML engineers, they wouldn’t know.”

At a recent agtech conference, asked the question that silenced the room: “Are farmers actually making more money?” Speaker after speaker had celebrated efficiency gains — water savings, yield improvements, smart pest detection.

But yield isn’t income.

A farmer with 35% more output who gets undercut at the market gate is still struggling.

We keep building for performance metrics.

Metrics that matter

The numbers that get celebrated in pitch decks — accuracy, efficiency, adoption in pilots — aren’t the numbers operators live by.

Even the most capital-rich technology category is hitting this wall: MIT reported that 95% of generative AI pilots failed to reach production in the first half of 2025.

Not because the models didn’t perform.

Because nobody had (fully) designed what satisfies the end user.

Context as architecture

The assumption underneath most deployment is that knowledge generalizes. That a model trained in Iowa transfers to Pakistan. It doesn’t.

That a workflow designed in an office will succeed, let alone be adopted on a construction site. It rarely does.

Physical systems operate under constraints that don’t appear in demos.

Weather.

Regulation.

Labor availability.

Generational knowledge that doesn’t fit in a dataset.

Time pressure that makes “optimal” irrelevant.

The technology that survives isn’t the most sophisticated. It’s the technology built for how work actually happens — compressing decisions rather than adding steps, fitting existing rhythms before asking anyone to change.

Operational excellence isn’t a feature.

It’s a design constraint.

This is what I’ll be exploring: conversations with operators building and deploying technology in physical systems — agriculture, construction, infrastructure. The patterns that separate tools that ship from tools that stick.

Not success stories. Observations from the field.

First up: precision agronomy in Australia.

Subscribe now