Most of the writing on chatbots/ChatGPT/AI is focused on potential “big” changes: putting lawyers, programmers, and the “creative” class out of work.
A second category of examination by the media focuses on how it will help humans “create” and be a “copilot” and reduce tasks that no one likes, such as scheduling meetings.
For this discussion, I’ll suggest this model. Think of two axes. One is the effect on existing worker. I’ll call this Replace-Augment: On one end it completely replaces the lawyer and on the other, it augments their function. The second axis is Domain. On one end are old established industries like farming and doctoring and on the other are domains that have not been around long or might not even exist yet.
Replacing Lawyers is Replace-OldDomain. New forms of artistry might be Augment-NewDomain.
But there are two other quadrants. One is Replace-NewDomain. This one has little data in it because in new domains there’s not that many people to replace. Programming or protein folding might lie here.
But the quadrant really packed with opportunity is Augment-OldDomain. Sure it’s fun to think of putting lawyers on the curb, or giving each person the ability to create a feature film by typing in a few sentences. But the Old Domain is where the problems are and Augmenting is going to have far less resistance than Replacing.
Additionally, in the Old Domain there are problems we had long given up on. Problems that we just live with, believing that that’s just the way it is because they don’t seem solvable. Impossible. Much like the “impossible” feeling each of us got when interacting with their first good chatbot.
So what’s one problem we’ve given up on? Terribly composed emails.
We’ve pretty much accepted who amongst us is longwinded, who is overtly terse and cold, who goes off-topic, who is slow to respond, who doesn’t pay attention, etc.
Magnifying this problem is the fact that we spend a LOT of time on email.
The average professional spends 28% of the work day reading and answering email, according to a McKinsey analysis.
28%!!!! On average. Even cutting it down to 27% across the entire professional class could be a monumental savings.
So what if AI would proof your email before sending? I know there’s already some tools out there that do some of this, but let’s try to flesh out more features that I think are almost certainly on their way.
Here’s a few tests that could be run by AI when you try to send an email.
Does This Need Edited?
Your particularly long-winded colleague sends and email containing way more information than is necessary. (Sometimes as a CYA and sometimes because they just don’t know better or think people want to know loads of detail). When they hit “send”, the system reads the email. If it’s crap, it suggests improvements and will not let you send it until it is better. Or, perhaps it is sent out with two tabs: original and the AI version.
Is This On Topic?
We’ve worked with people who often hijack threads to completely unrelated purposes. The system will not let you send an email that is off topic, or, perhaps if it does it comes with a warning to recipients that you are attempting to hijack the conversation.
Is The Tone Acceptable?
There are those who write with a certain tone – whether intentional or not – that really just gets on other people’s nerves. AI can determine if you’re about to annoy people when you could have just stuck to the facts.
Is The Message / Ask Clear?
So many emails are sent that are clearly half-baked. If the AI senses that you are being pretty incoherent, it can ask you some questions to help clarify your thoughts before you waste other people’s time because you didn’t think it through.
Has This Already Been Covered?
If the topic has already been covered in the thread or the documentation exists online, there’s no need to bug coworkers. The AI can catch this and let you know your answer.
Who Are The Right People To Include?
Gmail and other systems to this to some extent, but suggestions appear more calculated based on previous groupings. The content of the email does not seem to effect who is cc’ed. Whether you missed someone or someone needs dropped off, AI can suggest it for you.
The AI Gatekeeper
This is more an inbound tactic than outbound tactic. You don’t get to check your emails. The AI does it for you. This way you can get blocks of time to do crucial uninterrupted work. If something comes in that is actually urgent (not that just sounds urgent in the mind of your email-writing colleague) the system will disturb you. You can work without feeling anxious that you are “missing something” when you are doing real work. Separate times of day are set for checking emails.
Could AI start grading your future emails? Certainly. But the real question is “when will AI start grading all your existing emails?” If you’ve been at a company for 10, 20 years, that’s a lot of emails. How concise do you think you’ve been? How annoying? How much productivity loss have you caused yourself and your co-workers? How good or poor of a listener are you? Are you responding when you should and not responding when you shouldn’t?