Do you do AI?

Sure. I mean, who doesn't? But I take a specific approach to it: the responsible and sustainable use of AI.

What does that mean? I don't use AI just because I can. I work out exactly which part of a process really needs it, put the right safeguards in place to protect the data it's given, and build in a way to confirm the output is correct. Then I use exactly as much as I need and no more. A lot of problems can be solved with traditional programming, with a sprinkling of AI on top.

Side note: technically, what most of the market calls "AI" is a large language model - one class of transformer. But let's not split hairs.

It's like a table saw

Used properly, a table saw lets you build quality furniture, or a house for your family, that lasts for generations. Used improperly, you can hurt yourself and the people around you. AI is much the same: a powerful tool that has to be used safely, responsibly, and sustainably.

And when the only tool you have is a hammer, everything looks like a nail. AI is one tool in a bigger toolbox. Every problem gets looked at on its own, and I pick what fits: AI, traditional code, or a bit of both.

Where it earns its keep

AI is good at the low-level, non-creative work that's boring and not the best use of your team's time. Things like:

Reading paper forms and typing them in, with a human checking each field. I've built this - it's faster and less error-prone than manual typing, without removing the human from the loop.
Triaging emails and support tickets - working out what's urgent and who it should go to. Doing the legwork, not writing the reply.
Finding specific information buried in large amounts of text - the research grunt work, always with a human checking the answer.
Taking human data that's slightly different between each human and interpreting it consistently.

Use it to free your people up for the work that needs human judgement and empathy.

Where it doesn't

Blind trust. It's a probability machine. It generates output based on what's likely, not what's certain, and those probabilities aren't always right. Everything gets checked.
Medical, legal, or safety related questions. It can point you in a direction, but it lacks the nuance and accountability of the right professional.
Creative work that's meant to move people. AI can generate art, music, or prose, but it lacks the lived experience that makes those things land.

Privacy

If you type customer data into a public AI chatbot, assume it's being used to train their models. Don't do that.

Used properly - through the paid APIs, with the provider's data agreements read and understood - you can use these tools with confidence, provided you have the right safeguards and processes in place. And smaller models are now good enough to run on your own hardware or in a secure cloud instance, so sensitive data never leaves your control. That's how I built the pathology form reader: a local model, because patient data can't go anywhere.

At a previous employer I wrote the company's AI usage policy - how the tools could be used while protecting customer privacy and company IP, and making sure every output was reviewed. I can help you work through the same questions for your business.

Sustainability

AI has a real environmental cost, so I'm deliberate about how much of it I use:

Pay per use, not flat subscriptions. Paying per token shows you the true cost of what you're using, and naturally pushes you to use only what's necessary.
The smallest model that does the job. Less power for the same result.
Specific, efficient prompts. Less wasted compute.
Local models where practical, so you have some say over where the electricity comes from.

Models improve at an incredible pace, so part of the job is keeping an eye on new ones and evaluating whether they do the same job cheaper, faster, or with less.

How I use it myself

Programming. I still design every system myself. AI helps write the low-level code faster. I review the whole solution, and I'm responsible for the result - not the model.
Research. AI sifts through a lot of information quickly. I verify and make sense of it myself against real references.
First drafts. Documentation and emails get a starting point down quickly. I review and edit everything for tone, accuracy, and the human touch.

In short

AI is here to stay. Like every disruptive technology before it, we have to work out how to use it safely, responsibly, and sustainably. Should you use it in your business? Yes - if it's the right tool for the job, and you understand the risks before you do. I'm happy to help you work through that. Let's chat.