Sustainable Business and AI

Call me a radical, but I don't think a handful of billionaires controlling the majority of a country's wealth is a good thing. I don't think the ultra-wealthy need more benefits at the expense of everyone else. I don't think we should have just one grocery chain, or 5 tech companies -- everyone is far better off when there's competition, choice, and hundreds or thousands of smaller players instead of a few giants.

The tech industry is dominated by a handful of giants, and a bunch of startups all trying to corner some market through anti-competitive means funded by venture capital. These businesses are only considered successful if they are able to gain millions of users and get valued at figures hundreds of times of what was invested. This entire approach to business devastates the economic landscape -- providing apps and services way below cost, destroying competition that attempts to grow through healthy means like selling services at a sustainable price, and ultimately shitting on their customers and suppliers when they eventually try to cash in and profit off the mess they created.

The open source alternative

All these venture-backed companies have an Achilles Heel - the software they are built on is entirely free. They have certainly built a lot of custom functionality on top of this free base -- but more and more free alternatives are created all the time. And while they may not have all the functionality you need out of the gate, they have one characteristic that none of the big players have: You own the solution.

When you're using web applications, or buying software from the giants, you don't own it -- you're renting access. You have very little control. You can vote with your dollars, but if you're not among their biggest customers, you're not likely to persuade Microsoft or Google or Apple or Amazon to keep some software you've relied upon for a decade alive when they decide to kill it.

Owning your software

Using free software doesn't save you money. You still need computers to run it, and we still use cloud hosts to run those computers. You also take on responsibility for keeping it secure and up to date. When you buy a building, you become responsible for all the costs of upkeep, as well as anything you want to do with it -- when you use free software, you're taking on similar responsibilities and costs. But businesses buy property all the time, because at a certain point it's much better to own than rent:

• unlimited ability to tailor the space to your needs
• lower costs in the long run
• capital value in your business, assets that can be sold

But wait, didn't I just say the software is free, how can it be sold? It's not the software that has direct value -- it's your website property. The software, the "code", is similar to the building code in your physical properties -- blueprints, designs, the architecture. Those things have value, but not necessarily cost. You might pay an architect to create plans for a new building, like you would pay software developers to create new software -- but once created the value comes more from what you put in it than the designs themselves.

So if you can create your own software platform tailored for your business, and have full control over its development, evolution, and lifespan, doesn't that sound better than renting something from a company that might go out of business -- or at best get acquired by a big player and then get shut down?

Maintenance

The big advantage of renting is that usually the landlord handles maintenance -- if you own your property, you're responsible for all of it. If you're running open source software, you're responsible for all maintenance and upkeep -- and if it's online, it's important to not neglect, or else your secrets (and possibly your customers' secrets) will be spread across the Internet.

That's a fundamental service we provide at Freelock -- maintaining web apps for our customers. Just like you can hire companies to maintain your physical properties, you can hire us to maintain your web apps.

If you have solid maintenance practices, it can be straightforward to expand this to maintain more than just a single application. Most of our maintenance is for Drupal and WordPress websites -- but we also maintain servers hosting them, and a wide range of other line-of-business tools we use ourselves to support our business.

Open Source Business Software alternatives

At Freelock, we self-host pretty much all of the systems we depend on for running our business. We wrote the book on using open source in business over two decades ago. I wouldn't necessarily advocate ditching all your proprietary software at once -- but if you want to have a sustainable platform for your business, you should consider using open source alternatives before investing heavily in a proprietary platform.

Here are some of the platforms we use, instead of the commercial alternatives:

That's just scratching the surface of what's available, and what we actively use. There are reasons to use some of these commercial services that the open source solutions don't provide -- from this list Hubspot might be the biggest example of a service that provides a lot of help and guidance for content marketing beyond what you'll find in an open source platform.

Why do you call this "sustainable"?

The one key difference between all the software in the middle column and all the software in the right column is how the software was funded. Most (not all) commercial/proprietary software services are startups that were funded by Venture Capital (VC). This is the business model that dominates the software industry -- startups get a bunch of investment from VC investors who are primarily looking for a return on their investment. It's a game where the successful players take the money they've gotten by selling a "successful" business, and look for other companies to invest in -- in other words, it's a lot like a big pyramid scheme.

The VC investors all believe they are bringing new companies and new technologies to life, and argue that without this model, we wouldn't have the internet of today, or all these great services.

But... that's not exactly true. The Internet did not come from Silicon Valley, and was not created by entrepreneurs -- it was created by the US Military and a bunch of universities. What has made it so transformational and ubiquitous is not VC money -- it's through open standards and shared code. The VC model does more to funnel more money up to the people who already have it, than bringing new capabilities to humanity -- most VCs are looking for companies that have an "unfair advantage" or some way to corner a particular market before they will invest -- this is not about solving problems like Climate Change or the division of wealth -- it's a way to exacerbate the division of wealth by recruiting a bunch of smart people with a vision to make those founders personally wealthy -- at the expense of everyone else.

The problem with the VC model is that it makes it hard for other businesses to compete. A VC-backed business takes investment money not necessarily to make the best product, as much as to corner the biggest market. VC funding goes towards massive growth, an attempt to "own" the market, dominate the marketplace, drive out any competition. A business trying to grow by selling services based on the value it provides can have a very tough time competing against a VC-backed business that gives that same service away for free or far below what might've otherwise been the market value.

Meanwhile, Open Source has been there all along. More and more open source projects have small businesses providing support and services. And the thing is, if a business doesn't take VC money, it has far more flexibility to grow within their sales. Here are a few examples from the list above:

Drupal

We are Drupal vendors -- this is the main platform we provide services around. There are over 700,000 active Drupal sites right now, and many of those provide custom line-of-business functionality, they are far more than "just a website". Drupal is completely open source and free of charge -- and there are thousands of independent developers and agencies available that can make your Drupal site do exactly what you need it to do.

Drupal is one of the best examples of open source -- it's a platform that has some huge companies behind it, but you can manage it with a single developer, and it's developed in the open with core developers from hundreds of different, competing companies. Nobody has a monopoly on Drupal.

NextCloud

NextCloud has a different business model. It is backed by a single company based in Germany -- but like Drupal, the software is entirely free for you to install, use, and modify. You never have to buy from NextCloud itself -- but you can certainly hire them for services related to the platform, and other companies (like Freelock) can provide services around NextCloud.

NextCloud itself demonstrates one of the crucial tenets of open source -- the ability to "fork". NextCloud "forked" from an earlier project called "OwnCloud", which still exists. The founder of both companies is the same person -- for OwnCloud he took investor money that ended up leading the company down a path he didn't like, so he left and forked the code into NextCloud, starting again with a much more progressive business model.

As a result, there's now a thriving ecosystem of add-ons for NextCloud that provide all sorts of useful things for business. Some of the surprising things we do with our NextCloud is:

• Collaborative editing -- we can share a doc with clients using just a link, no account signup necessary -- far easier than Google Docs or Office 365.
• Digital Signatures -- recently the LibreSign app has reached the point where it has become a compelling, entirely free alternative to Docusign. You can easily find a document you recently saved as a PDF in NextCloud, and kick off the signature request process. You add the people you want to sign, and draw a box where you want their signature. They can come and upload or draw or use a selection of fonts as their signature to sign the document, and when it's all done it adds a QR-code watermark that can take anyone to the digital copy to verify the exact version as it was signed, through a cryptographic signature.
• Appointment booking -- the Calendar app makes it easy to send a link to anyone to book an appointment on your calendar. It automatically detects events from multiple calendars and allows you to create all sorts of rules around how soon to book, how much time to leave between appointments, and much more.

LedgerSMB

LedgerSMB is a completely free and open source double-entry accounting system. We have been using it for our bookkeeping for more than 20 years, and I am one of the core developers on the project. The core team consists of people from several different companies, and we work together to provide the functionality we and our customers need. Compared to the other projects, this is a tiny one -- but its longevity and support is excellent.

Now, switching gears...

How does Artificial Intelligence fit into Sustainable Business?

Ok, not to give you whiplash or anything, but we're in the midst of a big hype cycle around AI and Large Language Models (LLMs). Don't those go against the sustainable business aims I just described?

It's true -- AI businesses are all the rage right now. VCs are throwing money at all sorts of businesses that tell a story around AI, and so many commercial vendors are touting their AI capabilities. So if you're trying to run a business on open source software, are you going to be left behind?

This is going to be an ongoing topic on this blog -- we are doing a lot of work with AI already, so stay tuned for some upcoming videos, examples, and offers. For today I want to bring up 3 different angles on this question, specifically to the topic of sustainable business.

AI Effectiveness

First of all, if you ignore the whole AI thing, will you get left behind?

Possibly. You certainly should be paying attention, and learning about it. A lot of it is entirely hype. Some of my favorite quotes:

• "ChatGPT is mansplaining as a service" - it's very confident in what it says, but it's often wrong
• "I'm not afraid of AI taking my [coding] job -- it's like the x-ray glasses on the back of a magazine" -- it's more like a parlor trick than anything actually intelligent.

I already have a bunch of stories around using AI -- most recently, I was trying to use Claude Code to help me set up a development environment for LedgerSMB for a new platform (Nix). It very quickly gave me a setup that looked good -- but it didn't work. After an hour I was super impressed by the platform -- but it took me another 6 days and lots of expert guidance before I got it actually working -- so now I'm a bit disillusioned about the whole thing.

That said, there are a bunch of things that AI does well:

• Translations. AI-based translations are much better than the previous generations of machine translation.
• Summarizing long articles. (But it can hallucinate, and inject things that aren't really there!).
• Generating descriptions of images.
• Data Entry. (But be sure to verify the results -- if you're going to load transactions into a bookkeeping system, you don't want it making up numbers!)
• Audio processing -- I have an interview I'll be publishing shortly, and the audio was horrible. After some processing with an AI, it's no longer horrible -- it's not great, but you can at least understand it now...
• Transcriptions -- send over an audio or video file and get back text that can be indexed, appear in search engines, etc.
• Content evaluation -- for things like SEO, readability, tone, and more.

In general, I would say that using an AI these days is like using a brand new intern who is really fast but who also doesn't like to admit when they don't know something. There are plenty of intern-level tasks that it can deliver quickly -- but you do need to check its work to make sure it's not making some huge mistake.

Other downsides to LLMs

So if we're talking sustainable business, LLMs seem a bit far from that. They take intensive processing -- lots of compute power that takes huge amounts of electricity -- which can have an impact on climate change. They can be gamed to inject bias, or possibly reveal data they shouldn't. And they've mostly been made by the large venture-backed companies.

What I think is most exciting is that we're reaching a point where there's starting to be more emphasis on small, purpose-specific models. There's a lot of science fiction that features "personal AI" -- who doesn't want some sort of companion app that might remind you of the name of somebody you run into that you haven't seen in years, or to keep track of earlier conversations and help you better connect with them?

The most recent attempts to do this are still falling flat. The thing is, it's more likely for these small, entirely useful models, to come from open source and small businesses building on any of the thousands of models freely available on sites like HuggingFace.

Using a big model like ChatGPT, Claude, or Gemini brings with it concerns around privacy, bias, security, cost, and environmental impact. They are powerful tools, and we are still using them extensively, but I think we'll see more and more specialization around smaller purpose-built models that you can self host and that use a fraction of the compute power as the large models.

A lot of people knowledgeable about security get concerned about "prompt injection" -- ways of tricking LLMs into giving you data they are not supposed to do, bringing "social engineering" attacks to a whole new level. There's also the privacy issues around sending information to the LLM service -- many are not that transparent about who might be able to see your chats, or whether the LLM might leak that info to other users.

Another consideration is what does this do to the job market? Many are concerned about LLMs displacing jobs. I think these are fears of people less informed about how they work -- but it's a big problem that many of those "less informed" people are CEOs who might think they can save a bunch of money on payroll by replacing people with LLMs... from what I've seen so far, LLMs need more experienced review and expert guidance to get a result that still may need to get fact-checked. My hope is that LLMs will give people trying to solve big problems more leverage -- and this is apparently already happening in fields like protein-folding in medicine -- while still growing the overall employment like every other technology advancement has done.

If it doesn't necessarily mean less employment, it does exacerbate a trend I've seen during my career -- as the easy stuff gets automated, it becomes much harder to develop expertise. I've been working with websites for 30 years, and I've seen how and why each advancement has come along -- which means I have a huge leg up in troubleshooting when something goes wrong, simply because I have context our junior developers never got. If all the intern-level work is done by an LLM, how does somebody starting out get to the expert stage, where we will need more people?

Open Source AI integrations

With all the hype around AI, you might be surprised to learn that some open source projects are ahead of the software-as-a-service offerings. Drupal, NextCloud, and HomeAssistant all already have AI integrations you can use today. NextCloud was the first place I actually tried using an LLM -- the setup was easy to just start using.

Drupal has been at the forefront of integrating AI, and there are lots of ways you can use it today:

• Content editing, dropdown prompt you can use to request AI edits or wholesale writing of text
• Summarization
• Search
• SEO review
• Image selection from a stock image source
• Image generation
• Tagging articles
• Writing descriptions of images
• Adding fields to your site
• Creating and filling taxonomies with content
• Chatbots
• Reverse engineering website components from a screenshot

The Drupal AI people are constantly sharing things like how Drupal will work with the new "MCP" protocol that is taking the AI world by storm.

But it's not just hopping on the hype train. One key tenet the Drupal community has placed high value on is "human in the loop" -- working out ways where the AI can't just go change everything without human review first.

And Drupal's AI integration also allows you to swap out the actual LLM you use for each item -- while I'm mostly using Claude today, you can spin up and self-host a model from HuggingFace using a tool called Ollama, and have your own entirely self-contained special purpose AI that won't leak data back to the billionaires.

Wrap-up

To wrap up this long, meandering post - we are strong proponents of self-hosting as much as you can, and using as much free software as you can -- but AI is bringing a new wrinkle to that equation. As much as AI is flawed and over-hyped, it does provide real value and no business should be ignoring it. We are still in a time of experimentation with AI, but there are some emerging guardrails and practices such as "human in the loop" and paying attention to the possible risks of "prompt injection" and other new attacks that you need to consider.

The exciting future of all of this might just turn out to be that Drupal is the perfect platform for running your own personal AI. And while we may have needed some big investments to reach this point, using small models may prove to be far more sustainable than a service that loses money on every pro account!

Contact us if we can help you with your sustainable business journey!