Some thoughts about using ManagerBot (my first real project/app)

[TheRealChipA]

Before I drop these thoughts, you should know that I started using the original Square AI very reluctantly.  I'm a Boomer, thought I am very tech-oriented as my first career was as a programmer and analyst.  This career began, however, back in the days of punch cards and programing languages that are either dead or should be dead, anymore.  LOL. I joke about the fact that Square had to drag me kicking and screaming into the brave new world, which is not too far off the mark.  But, a couple of years later, I have to admit that I'm a convert and can no longer imagine running my business without ManagerBot in my toolbox.

This post is not about the merits or lack of merits of AI in general.  We all know that any good business tool will have pros and cons, and we have to decide for ourselves whether the pros outweigh the cons.  If ManagerBot doesn’t work for you or fit your world view, then definitely don’t use it.  I can say that for my business, at least, Square has created a tool that checks all of the boxes.  It gives me a way to do down-and-dirty analysis quickly, that would have taken me hours if not days before.  It’s eliminated more than a few spreadsheets and CSVs.  Best of all, it allows me to create ad-hoc reports without having to wait for Square engineers to do them.  BUT…. it is a tool, and only a tool.  At least for the moment, MB can not run my business, nor would I ever want it to.  

All of that being said, I recently used MB to generate an “app” that I use daily.  I run a soft-serve shop.  Each morning, one of my big tasks is determining how much soft serve base product I need to order, looking forward 7 days.  There are a number of factors at play here — sales trends for the past ten days, what my theoretical inventory is versus what I actually have on hand after my morning clean-and-refill routine, etc.

One thing I learned about MB (and AI in general) a long time ago is that the acronym KISS is the best starting point.  I also learned that my background in IT actually can get in my way and make things much harder. Some of my earliest aborted attempts did not work at all — mostly due to the fact that I was throwing too much information to MB out of the gate.  That turns out to be one of the worst things someone using AI can do.

Here’s how I started building my app.  I gave MB this statement — “I want to run a daily theoretical report of soft serve base usage and inventory levels that I can use to assist me in ordering. I'll provide you with the current inventory for each base, as well as a CSV that defines how much base is used for items that require soft serve base. Each day I will rerun this report to update theoretical inventory, as we as provide you with daily exceptions like waste/shortage. How should we begin?”

From here, I just followed its lead.  I was pleasantly surprised at how much knowledge it has about what I was trying to do.  It asked pertinent questions.  A few of its questions surprised me, but shouldn’t have surprised me, I saw later.  I won’t bore you with the details of the process, but if you’re interested, I’m attaching a PDF of our initial chat.  Note that to view the entire PDF, you'll need to click the square with the arrow sticking out of the top.

You’ll see that I backtracked a time or two, changed how I wanted MB to structure reports and projections.  I had to do this mostly because I realized that I was still over-complicating things and MB — being a very literal tool — allowed me to over-complicate them, much to my detriment.  Regardless, in the end, the ending tool was shaping up nicely.

Over the past few days, I’ve refined it.  Today I started exploring adding functionality for MB to start a cart for my soft serve vendor (GFS) and add orders to it.  GFS is not my biggest vendor, but I do have a small order guide in addition to soft serve that I procure from them.  We’ll see how this newest addition to the app works, but I’m hopeful it consolidates a little more of my morning routine into one place.

Some of the advantages of doing this process this way over continuing my rather complicated spreadsheet?  First, MB remembers everything we’ve chatted about here.  It remembers assumptions, how things convert, answers to questions I’ve forgotten it asked and got answers to.  Spreadsheets, as we all know, are error prone — we forget what links to what, why we set up formulas the way that we did, etc. And don’t get me started on editing complex spreadsheets — introducing unintended errors is just way to easy to do.  When MB detects inconsistencies in my answers and instructions, it points those out and asks what I actually mean and how to proceed.  That makes me more confident in the app it is creating for me.

And, on that note, I’ll stop this stupidly long post.  If you decide to dive into the attached PDF to see the flow of the app build process, beware it is 18 pages (and will be more tomorrow!).  This is a work in progress.  (Note that to view the entire PDF, you have to click the box with the arrow pointing out of it at the top.)

But I’m realizing that if I understand that fact, then the sky is the limit for having MB help me run my business faster and more fact-driven than I’ve ever done before.  My next task when this one is complete is to build a full ordering and inventory system.  My inventory needs aren’t for the faint of heart — I have 75-ish toppings that are depleted differently depending on variations/sizes, etc.  All of these toppings and flavors and more can be depleted by item/variation sold as well as by modifier chosen which is further change by the variation sizes, etc.  Trust me when I say that I’ve never found an off-the-shelf app that does exactly what I need.  Given that MB knows everything abut the structure of my data, I’m hopeful that I can finally to replace the spreadsheet I created that I absolutely hate!