Technology Psunami: Buyer beware, user be careful!
- Jun 25
- 3 min read

The social media feeds are suddenly filled with advisors and experts who want to sell you a technology that is not yet proven, and sometimes it's still a dream. The reality in construction is that accurate and contextual data simply does not exist in most companies.
The AI agent doesn't usually have video tape of your complete operation to analyze, nor does it generally have digital access to your records. Contractor's production records are generally vague and often non existent. Just a production quantity, some hours, and no context as to why the production was good or bad.
Project management software attempts to aggregate resources and quantities, but that data depends entirely on the foreman or engineer who is entering the data at the end of a long hot day.
This means that AI can only learn from the industry publications and any records you provide to it, but all of those records are based on "self reporting" by the contractors and employees. The context of weather, material delays, mechanical breakdowns, or absentee labor is rarely included along with the actual data. The incentive to include these factors is never present. The time must be reported so employees get paid and the equipment hours usually provide revenue to the equipment account. The production quantities are often estimated and they are necessary for the pay application to the owner. But nobody, except AI cares about all of the influences on those numbers.
AI is simply a pattern recognition program. Very powerful, but just as subject to error as we humans. Magicians are experts at taking advantage of our pattern recognition bias, so don't assume AI is better at catching deception than you are. It's just exponentially faster!
Can AI help you with estimating and bidding? Yes, "BUT", it only knows what you teach it about your company and what you think are adjustments that need to be made for certain conditions. If a crew has only one footing to form and pour, their unit productivity is going to be far less than if they have a hundred of them for build. Rhythm takes practice and productivity generally takes the shape of a bell curve.
Then there is data security. It's rumored that both Microsoft and Google are using data you think is safely stored in the cloud for their AI training. The reason that is plausible is because your data is encrypted on their side of the firewall. This means they hold the keys to your data and can unlock it at any time for any reason. Including legal discovery and of course AI training. I have moved my data to a platform that allows me to encrypt on my side and they never have access to the content, but that's just a personal preference. Any AI assistant being presented to you as an aid to your business needs to pass a smell test before you should trust your proprietary data to the model.
Believe me, I look forward to the day when a tethered drone can monitor your entire shift and provide real time analysis of the activity and production, but we are not there yet and we need to be extremely cautious about trusting this massive wave of technology and somehow filter out the snake oil flowing into the market.
If you look back at all the things that were promised to be right around the corner in the 1990's, you'll see many that never happened and some that took more than 20 years to mature. Be wise, be patient, and most of all be smart. There are consultants like myself out here who are willing to help you evaluate needs and vet the options.
BK

.png)


Comments