Thank you everyone for the wonderful feedback on my previous post on AI.  It seems like everyone is curious about this powerful “new” technology and how to create value with it.  I’m glad I could demystify the topic for many of you.

Today, I’d like to clarify a common misunderstanding of AI: it cannot be bought off-the-shelf like other enterprise software applications (think ERP, CRM, HRM, etc).  Yet, I see so many companies out there treating AI as if it were, falling for vendor claims that distort reality.  This is an expensive and misguided path, and there are much better alternatives to an alleged off-the-shelf solution.

Now I am not talking about software that has a deliberate business purpose that “contains” some AI element—say a supply chain management system that uses AI in some particular way to enhance its pre-AI functionality.  That’s perfectly fine, and I’ve seen some solutions that work well in that regard.  Rather I am talking about a platform that you can use without writing code to build ground-up AI. Such a platform does not exist.

I’ve observed companies worrying about which LLM is “the best” among Microsoft, Gemini, ChatGPT, Claude, and others.  The fact of the matter is that there is a constant horse race among the LLM providers, and almost no one who works in the AI field (like us) commits to any one but actually uses several different platforms within one application.  

Enterprise software is more-or-less a “one and done” process.  You implement the software, configure it to your company’s specifications and requirements, then “go live”.  Of course there are periodic updates and fixes and so forth, but most of the work is done by then.  AI is not like that at all.  AI requires continuous oversight, coaching, and adjustment, just like a human worker.  In fact that is the best way to think about AI – not as a software tool but rather as an extension of your human staff.  

Say for example you wanted to build something that carefully managed your raw material inventories instead of having Mary do that task.  To completely automate a task like that, you MUST write some code.

The idea of custom code used to be  a rather scary notion—not anymore if it is done right.

First of all, the best AI tasks are those that are placed within a company’s value chain in small, surgical doses.  That means a very small amount of code, thereby significantly reducing the time to build it and the effort to maintain it.  

Another observation I’ve made is that companies fret over user adoption—if I build it, will my people use it?  That’s a leftover notion from again thinking about AI like it is enterprise software.  Frankly the best kind of AI is AI that is invisible—it works in the background without anyone actually realizing it is there.  Therefore user adoption is not an issue in most cases.

So if you buy my “surgical dose” argument, your next question to me is: “Where do I apply that surgical dose to get the greatest effect?”

Here’s a useful exercise.  Divide your company up into the chain of functions that make the business run every day.  So that’s business development, production, supply chain, product development, finance and accounting, etc.  Make a list of all of these functions and the costs associated with each.  As an extension, add all of the “friction” within each function, like: Jack has to convert all of the sales orders (CRM) into shop orders (Production) manually so that they can synchronize new sales with manufacturing and order fulfillment.  Think of friction as the bottlenecks to each function’s work.  Put all that in a vertical list.

Now imagine a world where all labor is absolutely free.  I can have an army of free labor any time I want with whatever skills that I want.  Where within that list can I apply all that free labor?  I have used this visual exercise with a number of my client companies and it has served to unlock better thinking about where to put AI at its highest and best use.

You cannot afford to wait around until AI comes gift wrapped at your doorstep.  By then your competitors (and disruptors) will be well ahead of you.  Get started today, even if your efforts aren’t perfect.  AI is rapidly becoming the line that separates the winners from the losers.

If you don’t want to miss one of my blog posts, make sure and subscribe to my monthly Newsletter. https://business-laboratory.kit.com/09d106d65b