I’m rethinking software. I’d value your input.

[Frans VH]

I’ve been in software companies all my life. But the economics of software is changing fast. Custom software is now much cheaper and easier to build. Probably this shift will happen first in SMEs. Large corporations will keep buying SAP and Salesforce, because they move slow and commit to big systems.

SMEs are different. You can decide tomorrow to change direction. And instead of paying for SaaS tools that only partly fit, you can build simple internal tools and let AI handle repetitive work.

That idea excites me. But it also forces me to rethink where real value will be. So I’m trying to learn from founders outside the software world. I’m curious about businesses that:

• don’t require heavy capital or warehouses
• rely mainly on people and processes
• have a lot of coordination, admin, reporting or repetitive decisions

If you run an SME:

• Where is your business still very manual?
• What work consumes time but adds little value?
• if automation was simple and affordable, what would you tackle first?

This is not a pitch. I’m genuinely trying to understand where software can quietly remove friction inside real businesses and it what sectors that would work best.

 

[Chris Ashdown]

Custom software are fine until the shit hits the fan and your on your own to sort it out, and SME don't have the knowledge or time to sort it out, far safer to use a mega brand software if possible

 

[Frans VH]

The GenAI tools will evolve and like companies ran on MSExcel and MSAccess, there will be many companies on vibe coded tools. Many will stay on standaard saas products but will they be able to compete?

 

[gpietersz]

I think you have a good point in that lots of business do make use of custom apps written in Excel, Access or similar, often written by people who are not professional programmers. In fact Excel formulae by themselves are a Turing complete programming language and you have scripting on top of so you CAN do a lot.

The problem is that these apps are often problematic. Buggy, weird behaviours, incomplete, insecure or unreliable.

LLMs are a multiplier of what you can do, but also of what you can get wrong.

Another way of looking at it are the many, many efforts, staring with Cobol and Fortran in the 1950s, to eliminate programmers with higher level languages. It reduced the need for some skills, but exposed the need for other skills and just changed the skill set a bit. Using LLMs correctly requires the ability to review what it does, fix what it gets wrong, specifying requirements precisely....

My experience of LLMs so far is that they can improve productivity so they will bring down the cost of custom software, and therefore it will make it viable in many cases where it was not before. This is particularly true with things like writing a script to do a one off task rather than doing in manually, or small niche apps. However, if its business critical or security matters (e.g. anything that accepts incoming network connections) then you are still going to need to pay someone with the skills (albeit for fewer hours).

 

[GLAbusiness]

My experience of LLMs is generally good. BUT review of the results by a skilled user is essential.

I recently asked for help diagnosing a problem with my central heating. The results were overly complex but it missed the simple solution - It was turning off because I had set the thermostat to turn off if it did not detect movement. As I was not in the same same room as the thermostat it turned off. LLMM did not identify this possibility

 

[louis292]

Manual pain points I see a lot: quoting, chasing invoices, onboarding clients, moving data between systems, and reporting. None of it is sexy. All of it eats hours.

 

[Frans VH]

Manual pain points I see a lot: quoting, chasing invoices, onboarding clients, moving data between systems, and reporting. None of it is sexy. All of it eats hours.

The more boring it is, the better the result of automating it: fewer mistakes and more motivation for employees. What are some “boring businesses” you think would benefit most?

 

[fisicx]

The more boring it is, the better the result of automating it: fewer mistakes and more motivation for employees. What are some “boring businesses” you think would benefit most?

But does it need AI? Better processes can simplify and automate without using the resources drain that AI has become.

 

[Mark T Jones]

SME is sufficiently broad as to be meaningless, I'm guessing your target market is owner-managed businesses (the majority of which will fall firmly at the micro-end of small?)

From where I sit,dealing with a lot of them, the current issue is trying to shoehorn AI in as some kind of universal panacea- one lender, whose MD is a friend of mine is very much stuck here, and is looking at ways (which may or may not include AI) to sharpen up process, particularly underwriting and documentation whilst remaining firmly in the 'friendly/personal' zone.

It's a big challenge - if you can solve it, you may have struck gold

 

[fisicx]

I had a client a while back who wanted a plugin to do something. He suggested AI and all sorts of complications. I pointed out there was a simple API that could it all at far lower costs.

Process management can often solve many issues.

 

[Frans VH]

Manual pain points I see a lot: quoting, chasing invoices, onboarding clients, moving data between systems, and reporting. None of it is sexy. All of it eats hours.

I agree that moving data can be very timexonaing. But quoting and chasing invoices are part of the integrated systems like Odoo, Zoho and others. What in your experience are these systems missing that would make the process more efficient?