I was scrolling LinkedIn yesterday and my feed is effectively bombarded by posts that revolve around the idea that "AI can build an app in 10 minutes and allows everybody, especially non-technical people to not need technical people to build tools for their businesses", idea which I find on one side cringe, and on the other side completely misleading.

These claims usually come from people self-entitled "AI gurus" or "AI consultants" that need to sell this hype to draw in business owners who believe that.

After all, why wouldn't somebody want to automate their whole business, make their operations more efficient and increase their bottom line?

But here's the truth: as a highly technical person who has to deal with balancing technical and business decisions every day, I've seen what AI can do and I've seen its limitations. In order to build with AI you need to do the following things:

• be very precise with your requirements and you have to go in the deepest details. For example if you run an e-commerce business, you need to even specify the case of the product titles you want to get, or the resolution of the rendered image. If left to the AI, it will take some decisions that will most probably not be compatible with your full result.
• know what to look for: there will me minor details in the generated code that will not work as expected, and you have to know to look for them and eventually patch them.
• you have to know how the data looks like and how the interaction with other systems look like: these things are highly technical and if you're not a technical person yourself, unfortunately this is a blocker step and you'll need to hire some techie anyway to get this fixed.

But let's go back to the original idea of this post: building tech that actually works.

Most of the time, the tech you need isn't the tech that various companies or individuals try to sell you. And most of the time, even if the tech is right, the difference between where you are now and using that product is too big to be able to do the transition smoothly in one leap.

So you have to think in steps: figure out where you are, identify the operational bottlenecks and inefficiencies, and then upgrade your tech to tackle these directly.

In some cases, this change could be as easy as centralizing some data and allowing your employees to access it directly, so you get rid of the risk of employees working with outdated information and having to invest time in updating their work all the time.

In other cases, this change could be as complex as fully automating some data flows, so that when an action happens in a warehouse, your web store gets automatically updated with that information, removing the need of a human in the loop.

But in either case, starting from the solution is the wrong approach. You need to identify your bottlenecks and inefficiencies and then go from there. I know I repeated myself but this is the most crucial aspect of modernizing the operations of any business.

And building it is another story. Most of the time, you don't even need to spend thousands of dollars in development costs, because development is really expensive nowadays. Building a tailored internal solution for your need can run even into the tens of thousands if you are not technically experienced to stay on top of the agency that is handling it.

In many cases, the automations you need can be pretty simple and could be created in one afternoon by hooking the data sources, perform some simple transforms and then pushing the results in other systems. Maybe even using pre-made modules from workflow managers such as Zapier, n8n or Make.com

Conclusion

I'm not going to sell any solution without knowing the full problem. And because each business is unique, their problems are unique too.

Some may have an urgency in with marketing and their voice isn't propagating enough in this overly crowded internet space

Some may have issues scaling their physical products handling within warehouses because there always seem that products come in faster than you can ship to clients, even though the orders are placed.

Some may have issues with their employees who seem to make many mistakes when handling product data, because in one spreadsheet the products haven't been updated in two weeks and you're now losing money on them (hint: data can be fully automated and remove the people in the loop, and utilize their intellect in more high-value tasks).

But (and I'll repeat myself again), in order to get results, you need to know what you are measuring and how you are measuring it.

You need to identify the bottlenecks and inefficiencies, and then decide what to build and how. Starting from the root problem and finding the right solution for it will always yield better results than getting a solution in and then adapting your workflows to it, just because a salesperson was good at their job and may have oversold you the benefits.