Will ChatGPT Replace me? It’s a question keeping a lot of people up at night. If you worked in the 1800’s , then you would likely be woken up each morning my a man banging on your bedroom window with a big stick! Because alarm clocks hadn’t been invented yet, people had to rely on knocker-uppers (Yes they were actually called Knocker-Uppers) to make sure people got up for work. However, now we have alarm clocks and phones and the noble profession of the knocker-upper has faded into obscurity, having been replaced by a piece of technology.
Over the years as technology has evolved, many professions have become irrelevant and if we look to the future, it looks like many of the jobs we see people do today, will soon become a thing of the past. Autonomous vehicles look set to change haulage and transport industries, increased automation in factories reduces the need for human labour, you can even buy a freshly cooked pizza from a vending machine these days!
As a health and safety consultant, I thought I was safe and to be fair, I cant see an android doing a workplace audit any time soon. However, as I have thought more about ChatGPT and similar upcoming systems like Google Bard, I am becoming increasingly concerned about my Job! I mean, a lot of what iSAFE do is producing policies and documentation clients, if there is an app that can do this, will my business relevant or is iSAFE destine to become another just knocker-upper?
So, I have decided to put myself out of a job by reviewing ChatGPT’s work. Will ChatGPT Replace me? I’ll be asking whether it can write me some usable health and safety documentation. If its crap, great I’m safe and will remain in gainful employment and if its good I’ll be heading to the job centre.
I’ve put my headline findings below, however if you want to see me reviewing documents in real time, check out the video below (and don’t forget to subscribe to our YouTube channel).
Limitations of ChatGPT
When asking Will ChatGPT Replace me? I have to start with it’s limitations. The home screen for ChatGPT is clear about the limitations of the software. It states that information may be occasionally incorrect and may occasionally produce harmful instruction (not a great start for health and safety documents). In addition the ChatGPT home screen states that it has ‘limited knowledge of world and events after 2021’. As such it may not incorporate recent changes in legislation or emerging best practice as part of the documentation it produces. As such, users need to have an understanding of the legal requirements around health and safety and have sufficient knowledge to review and amend the documentation produced by the software to ensure it hits the legal marks and doesn’t lead to you as an employer giving dangerous instruction.
For each of the documents ChatGPT produced, it issued a disclaimer at the top of each document e.g. ‘Here is a basic health and safety policy that you can use as a starting point for your organisation’. This is great as no matter what is produced by the software it is never going to be specific to your organisation or your individual needs and requirements, given the limited amount of data you can provide in your question to Chat GPT.
With those limitations in mind lets ask ChatGPT to write us some documents!
Chat GPT Prompt: Write Me a Health and Safety Policy
At some point, every health and safety person in the land will have to write or review a health and safety policy, so this seems like a good starting point for our test.
ChatGPT produced a health and safety policy statement, not a health and safety policy. A health and safety policy statement is more like an executive summary or a quick overview of your actual policy. It will outline the scope of the policy as well as the headline responsibilities and arrangements for health and safety management. It does not however, go into any depth of responsibilities at each level of the organisation or give specific details of how health and safety will be managed within the business.
As policy statements go, Chat GPT developed a decent start / template for the document, it even had somewhere to sign and date the document. However, it would need to have some significant amendments to make it a document which would meet a businesses specific needs and requirements.
You can find generic policy templates all across the internet that do the same if not a better job. The HSE even has a free resource, so if you are doing a DIY approach you are probably better using one of those, just make sure you know what you are doing before you try and write your policy, otherwise you could land yourself in hot water with the law. If you want us to help you, check out our competent persons services here.
Chat GPT Prompt:
Write Me a Risk Assessment Policy.
Ok so we didn’t really get a health and safety policy there, so lets try and get some arrangements that would be in the risk assessment manual by asking it to write us a risk assessment policy.
The answer was actually much better on this occasion, giving us some arrangements and processes. ChatGPT outlined the 5 steps to risk assessment and highlighted the requirement for considering elimination measures first, which is a requirement of many pieces of health and safety legislation.
Although the document still needed a little beefing up and would need specific processes for completing risk assessment and defined roles and responsibilities for following the process, this was very close to the mark as a starting point for a business.
Chat GPT Prompt:
Write Me a Rescue Plan for a Cherry Picker.
OK so policies are all well and good but lets ask it to put us together some operational documentation. The work at height regulations put in places a requirement to have a rescue plan for all work at height operations, so we asked ChatGPT to put us together a rescue plan for a cherry picker.
The results on this one were lacklustre, to say the least.
The rescue plan didn’t take into account that a person may be incapacitated and may not be able to lower the cherry picker themselves, which to be honest is the main reasons to have a rescue plan.
It also talks extensively about a rescue team actually getting in the basket with the operator, which is never the first option and can be incredibly dangerous for the rescue team. It didn’t highlight how, the rescue team would get access to the basket (I like to hope the would gain access using a second MEWP), however at least it didn’t mention ladders as this would be even more dangerous. The rescue plan should have first focused on lowering the MEWP from ground level as this is a much safer means of rescue, prior to considering basket to basket or alternative rescue techniques.
On a positive note, the rescue plan did mention training and drills, which is a common mistake with rescue plans. It’s a great idea to practice rescue under controlled conditions, so rescue teams are aware of the requirements and processes for rescue in the event of an incident. It also didn’t mention emergency services, which again is good as you are not allowed to rely on the emergency services for your rescue.
Overall, this was a pretty shabby example of a rescue plan and gave some pretty dangerous advice, so don’t use ChatGPT for rescue plans. Instead, consider using IPAF’s template if you want to do it yourself or give us a ring and we will help you put one together.
Prompt: Make Me a Forklift Truck Checklist.
Checklist are super useful for many aspects of health and safety, especially when it comes to checking plant, so lets see if it can put me together a checklist for a forklift truck.
The checklist, ChatGPT produced seemed to hit all of the marks and there were very few flaws that I could find with the checks it prescribed. The only real issue was that it wasn’t in a checklist format, which I wasn’t expecting anyway and you could easily copy and paste the checks into whatever format you like.
Although the FLT pre use check which was produced was pretty good, I would still have a health and safety practitioner check over any checklists produced before you put it into use. ChatGPT seems to produce different answers, even if you ask it the same question, so although it was good on this occasion it may not be so good next time. Also just because it hit the mark for a forklift truck, it may not hit the mark for an excavator or another piece of workplace equipment.
Will ChatGPT Replace Safety Consultants?
Ok so the documents produced by ChatGPT were a mixed bag but the general theme here is that it’s good but not quite good enough.
At the start of this article, I may be slightly embellished, the importance of writing documentation to iSAFE. Don’t get me wrong it’s a chunk of the work, but for us, the important part is making sure people are working to the policies and processes and that they are working to keep people safe at work. Ultimately, words on a page don’t keep people safe and it is the implementation of these words and the actions they inspire that will stop accidents and legal issues for your business. That’s why iSAFE provides vastly more face-to-face services such as auditing, coaching and mentoring, as part of our competent person services than just providing the documentation. Take a look at our services in the link below because I feel it will be a while before the AI overlords can do that kind of stuff. This could actually be a good thing for safety consultants as it will speed up the boring documentation stuff and leave us more time for the face-to-face stuff that actually keeps people safe.
Where this will affect our industry is with the bigger health and safety consultant which I spoke about these in our how much do health and safety consultants charge blog. These businesses have a much bigger focus on the provision of documentation rather than the face to face support. To be honest, if these guys aren’t looking at Chat GPT and similar software through a very worried lens and making adaptations to their businesses I don’t think they will be around in a few years time.
At the end of the day, this software is here now and even though it needs a bit of polishing, this will get better to the point where writing and reviewing documents can easily be done through an app. Although ChatGPT’s answers still need a little intervention from someone who knows about health and safety, it won’t be long before the answers and documentation it produces is on a par with the knowledge of the experts. The pace at which technology is emerging and accelerating means if businesses like iSAFE and other consultancies don’t adapt to this progression, we will become irrelevant.
But for now, thanks for reading and stay safe out there!
You may also enjoy reading:
