Feature Proposal: A Free Household Purchasing Optimizer Built Into ChatGPT
I would like to propose a free, open-source Household Purchasing Assistant for ChatGPT.
The idea came from using ChatGPT to stretch a real $100 weekly grocery budget. I found that ChatGPT’s reasoning was much more useful than simply comparing grocery prices because it could consider the household’s actual needs, existing food, required purchases, coupons, quantities, substitutions, and budget together.
I think this could become an extremely useful public-interest feature, especially for families and individuals who need to make limited household budgets go as far as possible.
The basic idea
A user could tell ChatGPT something as simple as:
“I have $100 for groceries for two adults and two children for the week. I need coffee and baby wipes, already have chicken, eggs, rice and peanut butter, and I’m willing to visit two stores.”
ChatGPT would then work with a purchasing-optimization tool to determine not merely where the cheapest products are, but what the household should actually purchase.
The optimization could consider:
-
Target budget and absolute spending limit
-
Mandatory versus optional purchases
-
Existing pantry, refrigerator, freezer and household inventory
-
Household size and number of days being covered
-
Required breakfasts, lunches, dinners and snacks
-
Package sizes and unit pricing
-
Product substitutions
-
Current sales and coupons
-
Promotion thresholds such as “$30 off a $75 order”
-
Nutrition and meal coverage
-
Household preferences and picky eaters
-
Waste and spoilage risk
-
Loyalty programs
-
Pickup, delivery or in-store shopping
-
User preference for personally selecting fresh produce
-
Number of stores the user is willing to visit
-
Travel distance
-
Fuel or EV charging cost
-
Extra shopping time
-
Existing errands that make a store stop essentially “on the way”
-
Whether the savings from another store are actually worth the extra trip
The goal would not simply be:
“Find the cheapest grocery store.”
It would be:
“Given this household’s available money, needs, constraints and nearby purchasing options, maximize the real-world value they receive.”
Route and in-store optimization
The assistant could also optimize the trip itself.
For example, saving $4 at another store should not be recommended if reaching that store costs $3 in fuel and 30 minutes of additional time.
Once the stores are selected, ChatGPT could order them into an efficient driving route.
Inside each store, the shopping list could also be ordered according to aisle or department location so that the shopper follows a natural path instead of repeatedly walking back and forth across the store.
Where retailer aisle data exists, the tool could use it directly.
Where it does not, users could optionally help improve the system by reporting aisle locations or uploading pictures of aisle directories/signs. With explicit permission, the useful store-layout information could be anonymized and shared with other users while avoiding retention of unnecessary photographs or personal information.
Over time this could create community-improved store maps.
Human preferences should matter
I think the assistant should ask questions in normal language rather than forcing users to assign dollar values to everything.
For example, instead of asking:
“What is one hour of your time worth?”
ChatGPT could ask:
-
I’ll go out of my way if it saves money.
-
Balance savings and convenience.
-
Only add another store for meaningful savings.
-
Keep the trip as fast as possible.
Similarly, users could specify that they want to personally select fresh produce while allowing curbside pickup for packaged products.
ChatGPT could translate these normal human preferences into optimization constraints internally.
Why ChatGPT is a good interface for this
I have looked at grocery comparison tools such as GroceryChop and similar products.
They can perform useful price comparisons, but this problem is larger than price comparison.
A household rarely begins with a perfectly defined shopping list.
People say things like:
“My kids won’t eat that.”
“I already have four cans of soup.”
“I need school lunches but don’t care what brand.”
“I want to spend about $100 but absolutely cannot exceed $115.”
“I’ll visit another store if it actually saves enough money.”
“Don’t let store employees select my avocados.”
ChatGPT already understands this kind of messy natural-language context exceptionally well.
Rather than requiring people to leave ChatGPT and operate another application, the purchasing engine could provide structured price, inventory, routing and optimization information while ChatGPT handles the human conversation.
Keep it free
I would strongly prefer that this remain free to consumers.
The households that could benefit the most are often exactly the households that should not have to pay another monthly subscription in order to save money.
I am not proposing this because I need to personally monetize it.
I would be happy for the underlying project to be open source under an appropriate permissive license so that developers, retailers, researchers, food-assistance organizations and the community could improve it.
Retailer sponsorship should never determine which product or store the optimizer recommends. The purchasing recommendation should work for the user.
I am willing to help build and test it
I would be interested in helping prototype this rather than simply suggesting the idea.
I have already spent time working through the decision model, including household intake, budget constraints, promotions, marginal-value purchasing, route economics, fuel costs, fulfillment methods, produce-selection preferences, uncertainty scoring and in-store routing.
I also have a real household shopping workflow available for continual testing against actual receipts and store experiences.
My primary constraint is access to enough AI development capacity while I am working on another project that currently needs to generate income.
If OpenAI believes this concept is useful and could provide upgraded ChatGPT/Work access, development credits, or another appropriate level of development access, I would be happy to contribute time toward building an open-source proof of concept and share the results with the OpenAI developer community.
I am not looking to restrict ownership of the idea. I would rather see a useful version of this become available to everyone.
The basic mission would be:
Help people use the money they already have as intelligently as possible.