It’s one thing to say you’ve got this amazing idea that’s going to set the world on it’s ear. It’s another thing to build it yourself. It’s still another thing to be the only one who gets to use it.
Having never owned a bathroom scale in my life, I figured I’d better get cracking on that before I hit 60. My recent doctor checkups gave me a preview of how delightful it will be to wake up every day finding out what a fat-ass I am. Why not get a scale of my own? One that hooks up to my home wi-fi, comes with an iPhone app, can be given a name and hogs yet another IP address on my router? One that somehow determines my BMI, muscle mass %, heart rate, weight and pronouns – through my freaking feet – before telling me the weather and giving an inspirational quote. Something along the lines of “We made the scale so big so you won’t try to eat it too.” Yeah, I went a little overboard when shopping for this thing. But hey – it’s 2026 – scales have pretty much been like washing machines and toilets for the past 100 years. They work the exact same, with the same internal mechanisms, even though the outside shells might get a sleek, shiny or curvy new look or a random LED panel. Well not for me, I said! Give me something that makes me feel like Tony Stark when I step on it.
And with that inspiration, I did what most people do; I hit the internet. ChatGPT confirmed the thing I was looking for was indeed a “thing”. It’s called a Withings Body Smart home scale. It does the same stuff that regular scales do – only this one requires AAA batteries, connects to wi-fi and talks to my phone over Bluetooth. All for a cool $130. For a freaking SCALE. As expected, I did what most people do; I said “F that!” and slammed my laptop shut. After fuming a while, I sidled slyly back up to ChatGPT and – very subtly – mentioned that only suckers pay list price. Find this thing – new in box, because I sure don’t want a used one – for the cheapest price around. And off it went. I didn’t like the first results. It helpfully asked if I’d like to set up an agent to check all those stores again a couple times a day. Seemed a bit much, but I said OK and forgot about it. A couple days later I got an email from ChatGPT saying it had found a sale at Best Buy. Not shabby.
Also in my inbox was a notice from Amazon about Prime Day sales coming this week. ChatGPT explained that this scale historically has been featured in Prime Day sales, sometimes with even deeper discounts if it goes on a “lightning sale”. That’s a limited time sale for a super-low price that ends when the inventory – or the timer – runs out. Inventory usually goes fast so you have to jump on those when you see them. Problem is, you never know when these pop-up sales…pop up. Again ChatGPT read my mind and said we could modify the search to check every 5-10 minutes during Prime Day event. Ideas for a smart shopping app began to form. I went into brainstorming mode.
We shop for all kinds of things. Not all of them are on Amazon. Depending on the type and category of product or service, the stores selling them change. Prices go up and down, forming a historical reference and hi-low ranges. Patterns form – all different depending on what the desired thing is. TV’s have the best sale prices during Black Friday and the Super Bowl. Mattresses during President’s Day. Apple laptops during Back to School – iPhones in November just before Black Friday and Christmas shopping. Apple products also have cycles just like cars – they have new models coming out in [pick a month] so they need to clear shelves and blow out last year’s models just prior. Again, patterns form. Is a better price just around the corner?
I hear what you’re thinking: Amazon’s already got wish lists. Just put stuff in there and they will alert you to a sale. That’s well and good if Amazon’s the only store selling it, which often isn’t the case. Lots of people compete with Amazon and even match their prices. What about Honey? That browser extension that tries to literally snipe your purchase right out of a seller’s cart by suggesting alternate stores with cheaper prices. You can imagine how much delight and work that creates for the store who’s about to lose a sale to a third party plugin’s big mouth. As a result, Amazon is one of the internet’s hardest e-commerce shops to monitor by automated means. They block spiders, scrapers and other technologies trying to price check and essentially pull data from their site.
It is into this strange confluence of commerce, data, history, ontology, forecasts and behaviors that I decided to launch my little kayak of an experiment. The idea I have is called PAZARLi, after the Turkish pazarlik, which means to barter or haggle for an agreed upon price. I think that AI can help build a very intelligent shopping experience in which agents are programmed to look in the best places where a certain product is available for sale, take a snapshot of the current price, look at a host of other factors including history, but also new product announcements, the news – economic conditions that could help or hurt the supply line, seasonal patterns for sales and sale types themselves – then auto-tune each product being tracked and report in a dashboard which items are buy now, buy if you must (ie, laptop was stolen vs. thinking of upgrading) plus a learning shopping engine that is building a dataset record to draw upon for bargain hunters. All tied to the latest in alert and messaging standards so you’re notified the instant a “must buy” trips the alarm conditions. The interesting thing is this “shopping concierge” will not only be building a data set of a range of products and their associated traits – but also an interest graph of the shopper themself, including the all-important data point, “Target Price” that prompts an immediate buy recommendation when that condition is met.
The value of these two data sets is considerable. Both ends of the value chain benefit from this kind of service. Competing retailers want to sell stuff. Consumers want to buy stuff. There is (at least in the US) no real allocation for pazarlik in typical e-commerce situations. But imagine going into a price negotiation (for anything; a car, a house, a salary, a Playstation 5) knowing exactly what the number needs to be for the customer to yell, “I’ll take it!” Granted, lowball target prices aren’t going to trip many alarms, but if both parties are aware of what guarantees a sale, both have incentive to work together.
Because of the nature of the data this can generate, I’m keeping it on a private server and making myself the only user. After all this talk of innovation, the app might suck. It might not be as smart as I think it could be. Hell, it might COST me money rather than SAVE me money in some instances. That’s why I’m going to test the heck out of it. I want to see the data it generates. I want to measure, tweak, yank out big parts and replace with sleeker, smarter ones. All in the spirit of making a truly smart “shopping concierge”. In these days of AI agents and data centers hogging up all the powerful silicon chips and screwing with the environment, the engines they power should at least be smart enough to save me a few bucks on a freaking scale. Time to eat my own dog food and find out.
