I was at a random event yesterday where the speaker got distracted by his excitement at how useful the Copilot integration in DynamicsCRM is. He couldn't stop raving about how it was saving him 20% of his time on a daily basis.
It was a mind-numbingly boring seminar though, so perhaps it's these boring business functions that where the most benefit is seen, and they are so boring that journalists ignore the stories.
20% is a very strong claim. Effectively someone is saying they got a whole day out of their working week back for free. I find that unbelievable: or at least, it prompts my inner Carl Sagan to ask for extraordinary evidence.
In the case of Dynamics, and indeed of other CRMs (which already have a long history of ML/AI tooling, like Clari, Gong, etc) no amount of AI can solve the fundamental problem that if humans dont keep their records in the CRM clean and accurate, (which, by the way, they hate doing and absolutely suck at) then the AI is presenting insights based on garbage.
This task was mostly about summarizing old email threads when being passed between employees. It's the kind of thing that is quite time consuming but AI is very good at, and was a fairly significant part of his workflow so I found it reasonably credible.
Since it was literally email threads he was working from the "junk data" thing wasn't a huge problem.
Interesting how the narrative being pushed by companies spending a lot of money on AI always seems to be gravitate to around 20% of time saved. I've not seen anywhere how this figure of 20% is determined.
It seems to be a number that the masses are willing to just accept without any reasoning.
If they said 30% people might start to ask how, if they said 5% people might just shrug and think it's not significant.
> I've not seen anywhere how this figure of 20% is determined.
A lot of people are probably just repeating a number that is big enough to impress management but not so big it sounds far fetched. Otherwise evaluating Copilot should be no different from evaluating any other tool claiming to improve productivity. How would you evaluate Excel vs. pen and paper? Self reporting, estimating time saved per action or common scenario, counting how many actions are taken per day, etc.
This wasn't someone from MS, it was his estimate. Based on this workflow I'd believe it - he did lots of digging through emails to work out what was going on, but the CoPilot summaries were giving him that stuff straight away on the front screen.
I don't see 20% in any of MS' marketing material. The claim 6 hours saved weekly for a team of 5, so 2.5%. They claim 29% quicker on "completing everyday tasks", but specifics range from 74% faster (summarizing a meeting) down.
Plot twist: the number was originally generated by an AI, and that AI was selected for redistribution because it produced numbers that humans can easily get behind.
I used to (got sick of it) tally up the number of times "AI, CoPilot, ChatGPT" got said during the weekly local MS 30 min meetup.
Not including the number of times it was written on the presentation.. just the number of times it was said.
the average was 80 times in 30 mins.
utterly ridiculous. Like being in that film "being John Malkovich".
I no longer want to attend any MS "catch up" meetings... its all they want to talk about.
I use Copilot on a daily basis and I think it's a fantastic product. It isn't particularly exciting, but I don't think it's supposed to be exciting - it just takes a lot of the tedium out of routine business tasks. It writes boring reports and boring memos to a perfectly acceptable standard. It saves me from having to skim-read through a long document to find one salient fact. It does for office workers what a belt sander or a table saw does for carpenters.
It's integrated incredibly well into the rest of the 365 suite and Microsoft provide really good onboarding and training resources. It Just Works, even for non-technical users who don't have the first idea about LLMs.
Zvi Mowshowitz spends a lot of time talking about the mundane utility of AI and I think Copilot is currently the best example. It isn't something that promises to change the world, it's something that makes life marginally easier for a big cohort of people. Saving the average office worker an hour or so a week isn't revolutionary, but it delivers a very respectable ROI.
>It isn't something that promises to change the world, it's something that makes life marginally easier for a big cohort of people.
....and life miserable for a far larger cohort.
The number of jobs replaced, the number of jobs where the only parts capable of bringing joy will be automated away, and the number of people subject to ever-finer micromanagement as streams of more and more data are fed into authoritarian matrix multipliers will skyrocket.
The amount of human interaction, personal connections, and actors capable of sympathizing and working within the intent of the rules rather than the letter will plummet.
The ability to trust anything, to enjoy most things, and to believe that something is a real thing someone created to bring joy instead of a hypertargeted psychological adworm will disappear entirely.
A few people will get stunningly rich; a small cohort will make things easier. Many, MANY will be far worse off.
> the number of jobs where the only parts capable of bringing joy will be automated away
I've seen this claimed several times, but it goes against my personal experience using it in business and software development where AI tools remove a lot of boring rote work.
Can you give some examples of the parts of these or other jobs that bring joy which AI will replace?
All of content creation jobs are gradually evolving from replacing an artistic task with writing prompts.
Even programming - oh, good, LLMs are approximately as good at a fresh college graduate as spewing out a bunch of code that isn't quite right and I have to go debug and review. Except at least I can get joy teaching the new grad and seeing them grow.
They're allowing monthly payments for the first time, which are slightly more expensive than paying yearly like every company on the planet does. Way less of a deal than the title makes it sound.
I was at a random event yesterday where the speaker got distracted by his excitement at how useful the Copilot integration in DynamicsCRM is. He couldn't stop raving about how it was saving him 20% of his time on a daily basis.
It was a mind-numbingly boring seminar though, so perhaps it's these boring business functions that where the most benefit is seen, and they are so boring that journalists ignore the stories.
20% is a very strong claim. Effectively someone is saying they got a whole day out of their working week back for free. I find that unbelievable: or at least, it prompts my inner Carl Sagan to ask for extraordinary evidence.
In the case of Dynamics, and indeed of other CRMs (which already have a long history of ML/AI tooling, like Clari, Gong, etc) no amount of AI can solve the fundamental problem that if humans dont keep their records in the CRM clean and accurate, (which, by the way, they hate doing and absolutely suck at) then the AI is presenting insights based on garbage.
This task was mostly about summarizing old email threads when being passed between employees. It's the kind of thing that is quite time consuming but AI is very good at, and was a fairly significant part of his workflow so I found it reasonably credible.
Since it was literally email threads he was working from the "junk data" thing wasn't a huge problem.
AI to inflate outbox, AI to deflate inbox. Clever!
Interesting how the narrative being pushed by companies spending a lot of money on AI always seems to be gravitate to around 20% of time saved. I've not seen anywhere how this figure of 20% is determined.
It seems to be a number that the masses are willing to just accept without any reasoning.
If they said 30% people might start to ask how, if they said 5% people might just shrug and think it's not significant.
> I've not seen anywhere how this figure of 20% is determined.
A lot of people are probably just repeating a number that is big enough to impress management but not so big it sounds far fetched. Otherwise evaluating Copilot should be no different from evaluating any other tool claiming to improve productivity. How would you evaluate Excel vs. pen and paper? Self reporting, estimating time saved per action or common scenario, counting how many actions are taken per day, etc.
This wasn't someone from MS, it was his estimate. Based on this workflow I'd believe it - he did lots of digging through emails to work out what was going on, but the CoPilot summaries were giving him that stuff straight away on the front screen.
I don't see 20% in any of MS' marketing material. The claim 6 hours saved weekly for a team of 5, so 2.5%. They claim 29% quicker on "completing everyday tasks", but specifics range from 74% faster (summarizing a meeting) down.
https://www.microsoft.com/en-us/microsoft-copilot/meet-copil...
Big enough to be worthwhile, but not too big as to be unbelievable.
If you want to convince someone of something, you can’t go wrong with 20%.
Plot twist: the number was originally generated by an AI, and that AI was selected for redistribution because it produced numbers that humans can easily get behind.
> I've not seen anywhere how this figure of 20% is determined.
maybe laying off 20% of the teams/ICs?
We have seen them happening while the profits are growing.
I used to (got sick of it) tally up the number of times "AI, CoPilot, ChatGPT" got said during the weekly local MS 30 min meetup. Not including the number of times it was written on the presentation.. just the number of times it was said.
the average was 80 times in 30 mins.
utterly ridiculous. Like being in that film "being John Malkovich".
I no longer want to attend any MS "catch up" meetings... its all they want to talk about.
I use Copilot on a daily basis and I think it's a fantastic product. It isn't particularly exciting, but I don't think it's supposed to be exciting - it just takes a lot of the tedium out of routine business tasks. It writes boring reports and boring memos to a perfectly acceptable standard. It saves me from having to skim-read through a long document to find one salient fact. It does for office workers what a belt sander or a table saw does for carpenters.
It's integrated incredibly well into the rest of the 365 suite and Microsoft provide really good onboarding and training resources. It Just Works, even for non-technical users who don't have the first idea about LLMs.
Zvi Mowshowitz spends a lot of time talking about the mundane utility of AI and I think Copilot is currently the best example. It isn't something that promises to change the world, it's something that makes life marginally easier for a big cohort of people. Saving the average office worker an hour or so a week isn't revolutionary, but it delivers a very respectable ROI.
>It isn't something that promises to change the world, it's something that makes life marginally easier for a big cohort of people.
....and life miserable for a far larger cohort.
The number of jobs replaced, the number of jobs where the only parts capable of bringing joy will be automated away, and the number of people subject to ever-finer micromanagement as streams of more and more data are fed into authoritarian matrix multipliers will skyrocket.
The amount of human interaction, personal connections, and actors capable of sympathizing and working within the intent of the rules rather than the letter will plummet.
The ability to trust anything, to enjoy most things, and to believe that something is a real thing someone created to bring joy instead of a hypertargeted psychological adworm will disappear entirely.
A few people will get stunningly rich; a small cohort will make things easier. Many, MANY will be far worse off.
> the number of jobs where the only parts capable of bringing joy will be automated away
I've seen this claimed several times, but it goes against my personal experience using it in business and software development where AI tools remove a lot of boring rote work.
Can you give some examples of the parts of these or other jobs that bring joy which AI will replace?
All of content creation jobs are gradually evolving from replacing an artistic task with writing prompts.
Even programming - oh, good, LLMs are approximately as good at a fresh college graduate as spewing out a bunch of code that isn't quite right and I have to go debug and review. Except at least I can get joy teaching the new grad and seeing them grow.
> It does for office workers what a belt sander or a table saw does for carpenters.
Jobsite table saws are a thing as are portable belt sanders, but those aren't the most common carpentry tools.
Or is that what you meant?
Whichever AI wrote this article used way too many star trek puns
There seems at times a bit of a knee jerk assumption that articles must be written by AI (as opposed to AI-assisted), especially when they concern AI.
I don't think that's the case here, or if it was, it's good enough not to matter in terms of readability.
I find it’s now fairly common to call any article one doesn’t agree with or enjoy “ai generated” without any proof.
The literal last place I can see publishing AI generated slop is el reg.
Come on, they were so corny they got good again!
Who doesn't want a Photon Torpedo of Productivity?
Certainly they are disruptors in this space.
Honestly I got about a quarter through the article before I decided it wasn't written by a human.
They're allowing monthly payments for the first time, which are slightly more expensive than paying yearly like every company on the planet does. Way less of a deal than the title makes it sound.
It's to entice companies (esp. MS shops) to jump on the bandwagon - try it for less - and switch to annual if you like it...
It's really annoying that Microsoft-owned Github and Microsoft proper both have AI-based systems called "Copilot" which are unrelated.
> vile poison
Author could do with some grass touching. Yes there are real privacy concerns but it’s not the end of the world