In the final instalment of this three-part series about AI, Damian Davies, Head of Engagement at The Timebank, takes a look at how even by AI amateurs can start to use the tools out there without feeling that dreaded sense of overwhelm. This article explores three key AI tools—search engines, Microsoft Copilot, and meeting note takers—highlighting their transformative potential and energy implications.
AI will not be Cyberdyne Systems and kill us all off – it’s too thick . Well, it is until it hooks up with quantum computing, then we are all dust.
In the meantime, however, AI is more likely to kill us all off through the amount of energy it will consume.
No-one knows how much energy AI uses. A few years ago, firms like OpenAI would give out details like how much energy was being consumed to train their tools, but there is nothing like that for recent models like ChatGPT and GPT-4.
Now, the AI companies are getting more secretive as they become commercially competitive.
So, some people (a guy called Alex de Vries actually) made calculations to work this out (probably not using AI as it’s not great at doing calculations – it’s weirdly a bit of a weak area still).
The answer is by 2027 AI will use somewhere between 85 and 134 terawatt hours each year.
I know, didn’t mean much to me either, but in relative terms it’s about the same as whole of The Netherlands uses annually (or about 0.5% of all global energy consumption).
This is terrifying when you consider that for the last 10 years or so, global data centres already account for 1% or 2% of global energy use each year.
That’s basically everything that makes up ‘the internet’, so every server, every shared file, every dancing cat video, every app on every phone.
So, soon AI will account for half as much energy consumption as everything on the internet.
Whilst it feels like everything has AI in it at the moment, 2027 will still really be only scratching the surface, so the energy consumption of AI will increase massively.
That is because of the scale of what AI will become.
In this article, I will explore three basic tools that you might want to explore. Just from these three, you will see why the scale of use will be astonishing, but also the scale of opportunity.
The tools basically fit into one of the two categories I outlined in the last article: Data and language.
I would say at this point, avoid image or video tools. These still need a bit longer in the oven, as you get that ‘uncanny valley’ effect from pretty much everything it generates.
The First AI tool I suggest you explore is a search engine.
There are a few dedicated AI search engines, like ChatGPT Search.
In reality, AI is starting to land in pretty much all search engines.
Searching on the internet has changed massively recently. For the past 20 years, searches have been quite static.
‘Restaurants near me’ or ‘good financial advisers’
Now, people are increasingly searching with questions to find things out. This is a result of AI being built into the engines.
Structuring searches as questions makes sense when you consider the four AI prompt pillars; the goal, the context, the expectations and the source.
‘What restaurants are near me that have a 4.5-star minimum Google rating and welcome dogs?’
Or
‘Which financial advisers near me are female and specialise in advice on divorce?’
It moves the search up the food chain and will start to challenge some businesses that already do these kinds of comparative searches.
It can also shape how a business like yours, which is ultimately experiential and serviced based, can present itself online.
The second AI tool I recommend you start exploring is Microsoft Copilot.
Copilot is an AI overlay in pretty much any Microsoft product.
- If you are drafting an email, it will write something for you.
- If you are looking at a spreadsheet of data, it will cut through the numbers and help you build formulas.
- If you are building a presentation, it will draw the structure together.
As with any AI, it will only be as good as the prompt you give it. Prompting is the art of AI, and as already stated, focuses on 4 areas: the goal, the context, the expectations and the source.
Consider Copilot a useful assistant to draw up a first draft. You will pretty much always have to tweak it, but it might save you some time and thinking effort!
Where Copilot will basically take over the world is in the ability to build an ‘agent’.
Basically, an agent can be that person everyone in the office turns to for information.
That person might always have an answer, and be really useful, but that is probably not what they’re really paid to do. Building an agent will save their time and keep everyone on a consistent source of information.
By way of example, an HR agent might be built to sift through all the contract and HR docs to answer things like:
“Hey, how do I claim an expense for travel?”
“My pet has to go to the vet. Am I allowed to take time off?”
An advice agent might be built to sit over your advice manual or CIP and answer things like:
“I have a client with £250,000 to invest for growth over 7 years and they are risk 6. Which platform should I use?”
Before long, every business will be building a whole bunch of these agents to do a whole raft of jobs. At the moment, you have to buy expensive SAAS tools that never quite do what you want them to do. An agent can be uniquely tailored. It’s pretty radical.
The final AI tool I recommend you start exploring is a meeting note taker.
The idea is the tool joins your online meeting and listens in.
It will then transcribe the meeting – big deal. What it will also do, however, is collate the content of the meeting into important points or lift actions that are agreed as you chat.
This is incredible for meetings with clients, as it enables you to dig really deeply into the meat of what is motivating a client to seek your advice without having to take notes yourself. Typing away on a keyboard just interrupts the magic of an incredible conversation.
Even more importantly, you can share the points with the client to ensure they agree with what was said and agreed, reducing the risk of a complaint arising from each having different interpretations.
The absolute Daddy of these, IMHO, is Fathom. Honestly, without wanting to sound like the Nick O’Teen Character, once you try it, you’ll be hooked.
So, these are a great starting point for any AI noobs to start exploring what the tech can do for you. Just remember, do your due diligence before you use any of it, make sure you have your DPIA complete and only use tools that benefit your business.
About Damian Davies
Damian’s career in financial services started back in 1997 as a young IFA before then finding his calling as a Paraplanner.
In 2002, Damian started an outsourced paraplanning business called The Timebank.
The Timebank has developed to be one of the primary sources of support available to advisers and planners and looks after clients in the UK and overseas.
