You're right to be skeptical of AI.
Good. You should be. AI has real problems: sloppy output, real energy use, real disruption to jobs, and murky ethics around the data it was trained on. This page is not a pitch. It is an honest accounting of what concerns us too, where we think a more careful, more local approach genuinely helps, and where it does not.
The concerns we hear most
Our homepage touches on these in a few sentences each. Here is the fuller version, the same four things, argued out properly instead of summarized.
AI slop
Most of what floods your feed right now is the same handful of AI tools running the same generic prompt with nothing behind it: no real voice, no real standards, no one checking the output before it goes out. That is not a reason to avoid AI. It is a reason to be picky about how it gets used. Anything we build gets set up with your actual voice, your actual FAQ, and your actual standards loaded in first, and someone on our team reads the output before it reaches a client. That review step is the difference between slop and something that sounds like you wrote it.
Energy and water use
Running these models at scale uses real electricity and water. That is true of the frontier cloud models specifically, the big general-purpose ones running in a data center somewhere. We are not going to attach a specific percentage or a specific number of gallons to that sentence, because we cannot point to where a defensible number would come from, and we are not going to make one up to sound more precise than we actually are. If that concern matters to you, it is a real one, and it is part of why we build toward local options wherever a task allows it. More on what that actually looks like below.
Job displacement
We are not going to pretend this one is small. AI is already changing which tasks a business needs a person for, and that will keep happening faster than most industries can retrain around it. We do not think automating the repetitive parts of a job erases the need for the person doing it, but we are also not going to promise nobody's role changes. If part of what worries you is watching a role disappear entirely, that is a fair worry, and no consultant selling AI tools is a neutral voice on it, including us.
The pace of change
Something new ships every few weeks, and most of it is noise. We do not expect you to track any of it, and we do not think you should have to rebuild your systems every time a new model comes out. Our job is to filter that noise, adopt the changes that actually earn a place in your business, and leave the rest alone until it has proven itself.
Two things the homepage does not get into
These do not come up as often in a sales conversation, but they are part of doing this honestly.
Where the training data came from
The large models we build on were trained on enormous amounts of text, images, and code, much of it pulled from the open internet without the original creators' permission or payment. That is a real, unresolved ethical question, and it is being argued out in courtrooms right now. We are not the company doing that training. We are a downstream user of tools built by Anthropic, OpenAI, and others, the way a contractor uses lumber without owning the sawmill. That does not make the question go away, and we are not going to pretend it does. It also means we are not the right people to resolve it. What we can do is be honest that it exists and let you weigh it for yourself.
Is this all a bubble?
Maybe some of it. Money is pouring into AI companies at a pace that has not always matched the revenue coming back out, and some of that will correct hard when it does. We do not think the underlying capability disappears if a few overvalued companies come back to earth, the same way the internet did not stop being useful after 2001. But we are not going to tell you there is zero risk here. Anyone claiming total certainty about where this goes, in either direction, is selling you something.
What actually happens to your data
Not a hypothetical. This is the plain-language version of what actually happens when you type something into an AI tool.
A cloud AI tool
Type into ChatGPT, Claude, or any other cloud AI tool, and what you type does not stay on your computer. It travels to that company's servers to be processed, the same way a web form does the moment you submit it. Depending on the provider and the plan you are on, that conversation may be stored, may be used to help train a future version of the model, and may be kept for some period of time after you are done. The exact policy is different for every provider and changes with their terms of service, so the honest answer is: read the terms for whatever tool you are actually using, because there is no single universal rule here.
If it runs locally instead
Running it locally changes that picture, because there is no server on the other end to send anything to. What you type stays on the machine you typed it on. Nobody's terms of service apply, because nobody is collecting it. Nothing gets stored on a company's servers and nothing gets used to train a future model, because that company never receives it in the first place.
Where local models genuinely help, and where they do not
This is where local models earn a real place in this conversation, not a marketing one. A model running on your own machine instead of a company's servers is a genuine, practical answer to two of the concerns above: it keeps your data out of someone else's data center, and it draws a lot less power doing it. That is real progress, specifically on privacy and on energy use.
It is not a fix for everything else on this page. A local model still came from the same training process as the cloud version, so it does not resolve the copyright question. It does not change what AI automates, so it does not touch job displacement either. And honestly, it is slower and less capable than the frontier cloud models right now, a good fit for drafting an email in your voice or summarizing a document, not the right tool for the hardest reasoning work. We use it for the first kind of task and a cloud model for the second, and we will tell you honestly which is which for whatever you are trying to do.
We are not going to pretend AI is the right tool for everything, because it is not. If the honest answer for your business is a better process, a plugin, or simply hiring someone instead of building an AI system, we will tell you that instead of selling you something you do not need. That is not a sales tactic. It is the same reason we build custom systems instead of handing you a generic login: the tool has to actually fit the problem, or it is not worth your money.
If you want to talk, we're here
If anything here made you more curious than skeptical, the door is open. If it did not, that is a fair outcome too. Either way, you now know more about how we actually think about this than most AI pitches will tell you.