AI AI Sir!

Artificial Intelligence has been around for some time, though it has gained prominence in the last 18 months and seems to have made incredible leaps.

It's difficult to ignore and any CTO will be keen to explore ways to make it work, if they're not using it already, whether that's generative or agentic in nature.

I'm no exception and I'm particularly keen on potential productivity and efficiency gains that can be made in our operational areas - from underwriting to customer servicing to claims. There are more manual process at work than we'd like to see and we're keen to relieve the burden on our colleagues, alleviate problems caused by data quality and speed up our processes with the aim of improving the experience of our advisors and policy holders.

The weight of evidence, though, is not overly favourable just now -with Gartner showing that around 74% of generative productivity improvement products are failing to deliver the slated benefits.

One of the difficulties, methinks, lies in the fact that these are repetitive tasks - by their very nature making them good candidates - but also, by their very nature, being difficult to quantify (without creating a cottage MI industry) in terms of benefit. Simplifying, speeding up and reducing rework can't fail to improve efficiency....right? But, precisely because they're small pockets of time - it becomes difficult to see what you could be doing instead.

The reality is that you're unlikely to use the new found time to do new, exciting things - you're actually using the time to do the things you don't have time to do right now. Maybe we should automate the things we're not actually doing? But then, we're not doing them because they're a lower priority, right? A bit of a dilemma.

When you get excited about the endless possibilities, sometimes you miss the endless practicalities.

It's important to remember the things AI can do and the things it can't do - for anyone worried about being replaced by a robot.

• It can collate, assimilate, and present information - but can't distinguish fact from misinformation
• it can calculate probability but can't assess risk
• it can provide useful summaries but can't make the innovative leap
• it can tell you what the law says but can't guide your morality
• it can list your options but can't decide for you
• it can retrieve accounts of what happened but can't remember
• it can answer questions ad infinitum, but can't ask any itself
• it is reactive, but can't be proactive
• it stops when you stop asking it questions - it can't reflect, consider, re-consider or change it's mind
• it can create beautiful images or poetry but it can't feel
• it learns but does not understand
• it's as much use to organised crime as it is to legitimate business

It cannot tell right from wrong, it doesn't know anything and it can't think.

Most importantly, it can't absolve you of your responsibility.

It is not human and can never replace you - though it probably can take away some of the more mundane things you spend your time on.

The trick then is to figure out what we can do with the infinite possibilities our collective brain power provides for us.

Exciting, eh?