In this video excerpt, Speaker A presents a newly launched, purpose‑built AI platform for an energy building compliance firm that conducts NatHERS 7‑star energy ratings for new residential constructions. The platform consolidates the entire job lifecycle—intake, pricing, quoting, task management, document handling, messaging, and invoicing—while integrating a custom AI engine that reads building plan images (floor plans, elevations, window schedules, insulation specs, construction details) to automatically pre‑populate assessment fields. The AI also provides confidence scores for each extracted item, flagging uncertain data for human review, thereby shifting assessors from manual data entry to a rapid verification role. This reduces assessment time from roughly 1.5 hours to 3–10 minutes per job. The system was built for $75,000 over under three months and is projected to generate about $750,000 in the first year, a ten‑fold ROI. Speaker B expands the discussion, arguing that enterprises must adopt AI‑native operating systems with self‑optimizing workflow agents, suggesting that such automation will render traditional human‑to‑human processes obsolete. The excerpt serves both as a concrete case study of industry‑specific AI implementation and a broader commentary on the strategic imperative of AI‑centric enterprise transformation.
Chapters
So I've recorded this little snippet because I want to show you something that we've just launched, and I'm pretty proud of it, and I think it's worth seeing. So this client runs an energy building compliance business. So they do the NatHERS 7-star energy ratings for new residential building, which is the paperwork that every builder has to get the house approved before they can go ahead and start building it. The client was doing about 15 of these per day, and almost all of them were manual. Every job, someone had to go through and check the emails, read the plans, get all the information off the building plans, key it into one piece of software, then analyze it. Then if the building failed the energy assessment, they had to go back and forth with the client by email to work out what the changes were going to be. And every building company has different defaults that they have as their changes that they're willing to make. And then storing stuff in spreadsheets and then manually entering it into the CSIRO system to produce the— the stamp and then stamp the plans. So we basically replaced all that with AI. So we built them their own platform, purpose-built just for this business. It manages the whole job lifecycle. So intake, pricing, quotes, tasks, documents, messaging, invoicing, all in one place. We have, we can set defaults and different prices for each building company, default specifications as well as default pricing. But the part that I am most excited about, apart from the admin side, is the AI. So when the job comes in, the client uploads their building plans like I'm doing now. There's a whole lot of stuff at the back end too where the checklists are all built in for the assessors and they automatically work through the steps so that the AI eventually will be able to take over a number of those tasks as well. But we have built an AI that actually reads the plans. So it's not just scanning for text on the page, but it's genuinely importing the images. It looks at the floor plan, it works out the orientation of the building. Reads the elevations, picks up on the eave overhangs, the window schedules, the insulation specifications, the construction details, all the stuff that's drawn, not just the stuff that's typed. And then it pre-populates the fields in the job form automatically. And here's the bit that I really love: it doesn't just fill things in and hope for the best. It tells the assessor how confident it is in each result, and anything it's not sure about, it flags for a human to check. So the assessor's not starting from scratch, but there is still a human in the loop, and the human is just reviewing and confirming rather than reading and typing. So what used to take an assessor an hour and a half now takes between 3 and 10 minutes depending on how complex the plans are. And here we have the information that's been populated and the degree of confidence for each item. So to give you the numbers, they spent $75,000 on the build. It took just under 3 months. We started it after Christmas when we came back, and what are we now, end of March? They have forecast that in terms of cutting expenses and a couple of extra little revenue things that will happen as a result, it makes $750,000 difference to their bottom line in this first year. So 10 times return on investment, not too bad. Anyway, I just wanted to show you this because I think it's a really good example of what's possible when you build something properly for a specific industry rather than trying to use generic tools. I'm happy to walk anyone through it if there's any questions.
This will be the biggest thing to hit the enterprise world in decades. Every organization in the world now has to do one thing to survive, one only, which is at the edge of your organization, create an AI-native operating system that's centric, AI-centric, and start moving workflows over to, over to it. One of the things about OpenCLAW was that you have recursive self-improvement in business workflows. Once you have that, all human-to-human workflows essentially evaporate. It does get, you can't sustain, you can't compete because once you put a workflow into this set of agents and they're optimizing it by themselves, you want to get the humans out of the way as fast as possible. And I think we're going to see adoption by enterprises at that scale
