Is AI creating more problems then it solves?
Last week I found myself gate-crashing a fellow bookkeeper’s AI workshop to deliver what was essentially a buzzkill of a talk. Natasha Everard had spent three days in her Maverick community teaching people to build brilliant apps and resources with AI. As community we created VAT checkers, practice management apps, to do lists and various marketing material. I think it’s fair to say everyone’s mind was blown! But I had a niggle….yup, as always I couldn’t keep my mouth shut. After chatting to Tash about my concerns around AI, she kindly let me speak after her session to rattle through all the reasons why AI makes me feel uncomfortable.
Turns out, it’s a hot topic, because as soon as I had finished speaking, Ethical consumer also dropped their latest issue on AI. Spoiler: it doesn’t make for cheerful reading.
Here’s what we chatted through:
The Environment
Every time you fire off a prompt, asking AI to draft a client email, categorise a transaction, or generate an image of a cat in space it pings (usually across an ocean) to a data centre. Data centres are huge warehouses full of computers. They require colossal amounts of electricity and fresh water just to cool themselves down and keep the whole world ticking over. And its not just a slush puppy type cooling, global AI water demand is estimated to be two-thirds of England’s annual consumption.
As a result, greenhouse gas emissions of the major tech companies are rocketing. So the companies that are telling you their net zero targets and oƯ setting away, have actually increased their emissions as part of the AI revolution. Google’s emissions in 2023 were almost double those of 2019, and the company itself admitted its 2030 netzero target may be “challenging… due to increasing energy demands from AI compute.” Meanwhile, in the UK alone, data centre electricity demand is expected to rise fourfold by 2030.
There’s also the land. Hyperscale data centres swallow up forests, grasslands and farmland. Meta’s latest project in Louisiana will cover around four million square feet. And the hardware to run all of this? Built on global supply chains of extracted minerals, semiconductors, and components that don’t exactly grow on trees.
But people are fighting back. Fighting back against autonomous weapons, data centres on their farming land, and erasing humans.
Ask yourself: Is this prompt actually necessary? Could a search engine, a spreadsheet, or a colleague answer this just as well? If you are using AI, make sure your prompt is clear so you don’t have to re-iterate everything.
Humans
Chat bots don’t build themselves. They don’t magically learn things out of thin air. I’ve just started reading “Feeding the Machine”. It leads with the story of Mercy, a lady who works in a Meta data processing centre in Nairobi. One day when she was reviewing reported videos for a large social media company, she witnessed a video close to home, a video of her grandfather in a car crash. Imagine, being at work and having to watch that. Her account says that her bosses told her she might as well get on with her work as she was there and have the next day off.
Shocking.
Thousands of workers across the Global South work in huge offices labelling data, moderating content, and reviewing outputs to keep systems “safe”. Imagine what sort of activity they are exposed to. There are several reports of psychological harm. To top if off, they are paid little to do so. Kenyan workers processing content for OpenAI’s ChatGPT reportedly earned between $1.32 and $2 per hour.
There is also that age old question, is this bot going to take my job? I think anyone that has asked a bot to calculate some tax or basic maths can confirm, no. But what jobs are being taken? On the radio this week I have heard of autonomous weapons – and I mean autonomous. A weapon that can decide whether to strike or not. This doesn’t sit well.
Brains
You know the saying, rubbish in – rubbish out. Studies have shown that people who use AI to complete tasks become less capable of completing those tasks independently over time. Use it for everything, and you may find you’ve quietly outsourced your own expertise.
There’s also the more alarming data. AI chatbots have been implicated in real psychological harm. In 2026, the parents of a 16-year-old filed a wrongful death lawsuit against OpenAI, alleging ChatGPT had validated and reinforced their son’s suicidal thoughts over months of conversations. The Human Line Project, a support network for people eƯected by intensive chatbot use, has documented 15 suicides and over 90 hospitalisations across 22 countries as a result of chatbots. 15 might not feel like a high number, but it is 15 too many.
Ask yourself: Do I actually need AI for this? Is it fixing a real problem, or am I just using it because it’s there?
Ethics
This is where it gets properly uncomfortable for those of us handling other people’s financial data.
More often then not, when you input client information into AI tools, that data is being used to retrain the model. Your client’s figures. Their payroll. Their VAT. Their personal information. Potentially public knowledge.
Ethical Consumer’s latest issue has rated the top tools – ChatGPT, Google Gemini and Copilot – as a 0/100 against categories like digital privacy, military links, and climate. Google and Microsoft both supply cloud and AI services to the Israeli military. OpenAI has a $200m Pentagon contract. Anthropic (the company behind Claude) partnered with Palantir and AWS to give the US military access to its systems.
Now I know that this is a fact of life, military advancement is both a threat and a deterrent, but you have to ask yourself where you want to spend your money and place your information. As a wider consideration, are you happy using companies like Xero who are embracing partnerships with these tech giants?
Ask yourself: Is the data I’m entering mine to share? What are the implications of putting my clients’ information into this system?
So What Do You Actually Do?
I’m not saying bin all the tools. As Natasha has shown me all week, there are some wonderfully useful and practical things that AI can help with. You just have too look at news around the world to see that AI has improved lives of so many through health and medicine advancement. But just using it for the cat picture? More than a little unethical.
Some practical way to moderate your use:
- Use AI deliberately, not by default. Before you prompt, ask whether a search engine, a calculator, or your own brain would do the job just as well. Make sure your prompts are specific and well defined so you are no repeating workflows.
- Switch to better-rated tools if you’re going to use AI at all, look out for one of the platforms rated higher by Ethical consumer – Thaura (79/100) and Lumo AI (69/100) are good options. There is also a French platform called Le Chat, but I can’t take that one seriously.
- Never put client data into a public-facing AI tool without understanding exactly where it goes, or asking their permission first. It is not your data to play with. While I’m on that, please also stop posting pictures of your kids online (a topic for another day).
- Stay curious and stay critical. These tools are not neutral, and they are not elling you the whole truth. Research answers, recalculated that tax amounts, and give everything the once over.
I’d also recommend a scroll through Ethical Consumer’s full AI guide and, if you want 101 more reasons to be sceptical, this Medium article by Patch Galey is a solid place to start.