Will they? Do we think any of the executives making this decision have any idea of the problems of adopting AI? Most companies today can barely create a functioning website that allows you to see and pay your bills.
I've been dealing with a fibre upgrade issue for three months now. Four times in total their support has offered to call me with an update. Twice they scheduled a specific date and time for the call. I've yet to receive a single phone call or email from them.
Their checkout flow can break. In some browsers it'll get stuck in an infinite redirect loop. If you're using edge you'll get redirected to a page that shows the ~10 year old BT Infinity branding, with a totally different style, packages, and prices.
Their chat bot is of the kind that decides whether or not you're allowed to speak to a human based on what it thinks your question is.
They're doing a massive build out of fibre to the premises (replacing old copper) and VOIP at the moment so I guess it makes sense that they'd need a lot fewer people once that's complete. They're assuming the new fibre will take much less maintainace than old copper. And better chatbots will replace a lot of CS staff.
42% fewer employees, 7 years in the future? Including hopes for things like this:
"Jansen also said that around 10,000 fewer people would be needed to service and repair digital networks, which “go wrong less often” and can be fixed more easily than older networks."
That feels optimistic. Modernization does help with that sort of thing in some ways. But it also often adds complexity, higher dependency counts, etc.
Modernization does help with that sort of thing in some ways. But it also often adds complexity, higher dependency counts, etc.
The complexity is inside opaque boxes that BT deploys though. Rather than fixing things like they did in the past, sitting at the end of a street patching cables in a roadside cabinet, BT's engineers of the future will just unclip a box that's returning an error code and plug in a new one. Or someone in a NOC will reset it from 300 miles away. Maybe some AI will do that. Or a cron job. With fibre networks once the cabinet is installed and plugged into the network there's very little physical work to do. That's what engineers do now - manually splicing and connecting copper wire. That's the work that's going away.
Not only that, but it's cheaper and easier now to monitor these boxes via software.
So you don't need phone operators to wait for calls about interrupted service, the software will just detect some unreachable part of the network and dispatch a team to investigate.
The technology was improving even on the copper stuff. The last time my broadband went down, I clicked a button in the app and they acknowledged a fault. An engineer came out the next day, plugged his iPad in and could see that there was a break in the cable approximately 16 metres away. And it turned out to be correct.
To say it’s all based on copper cables, it was pretty impressive.
You could do that 20 years ago by calling the lineman’s number and choosing the relevant function - it’s all on tablets now, I’m sure, but around 2000 it was all done using a phone call and a robotic voice, and there were engineer pins floating around on the internet.
I work in this space at an American telco and you are correct. The only thing I haven’t seen is AI doing it. However there are AI/ML features built into the monitor tools we use.
A copper line isn't a continuous thing from the exchange to the customer. It's made of a number of sections that are joined together.
Now someone who once worked for a large British telecoms company once told me that they'd done some modelling, maybe using something like system dynamics if I was guessing, to gauge the effect of work on the network.
Any interaction with the equipment, whether to set up lines for a customer or even fix faults leads to more faults. So it ends up being more profitable to just leave equipment in place than to reuse it.
Based on that I would say that having all the pairs terminating at the cab instead of the exchange and using remotely configurable equipment would be a greater improvement again.
Just guessing you understand, I have no inside knowledge. ;)
My team is responsible for many services running over openreach and other BT point to point fibres, both local connections like EADs, and longer distance stuff. My company has thousands.
The number of RFOs I see which report outages due to either fibre disturbances in an exchange or is phenomenal. I don't really have any idea how that compares with copper, but my home DSL runs over copper to the exchange and doesn't go down anywhere near as much.
> which “go wrong less often” and can be fixed more easily than older networks.
What they mean is that consumers will be stuck with slow internet connections. As a result of customers switching to other providers they’ll have less equipment breaking down. Since you know it’s no longer in use.
The BT group run Openreach who most other UK ISPs will resell as their own connection. Customers switching to other providers won't harm them as much as you'd think.
“Adopting AI” before there is any definitive proof that replacing workers with GPT tools is actually effective, is just cover for layoffs they were already planning on doing.
If I needed to do a layoff now I’d just say “AI LLM Chatbot!!” woo-woo as cover.
I don't think it really classes as woo-woo - replacing live chat operatives with an LLM is a pretty straight forward and obvious use case.
Besides, the full 55k are about all automation, restructuring and digitisation efforts. Doesn't sound entirely unreasonable.
They probably are also looking to justify investment and say that they are investing into making the business more profitable for the next 7 years, so don't expect too much profit until after this investment is done.
When will IBM finally give up the ghost? Do they add any sort of real value for customers anymore? Their cloud products are awful (I've used them), and I've only heard bad things about their consulting (and that's been the story for ages).
The last good thing about IBM that I remember is the Thinkpad, but Lenovo's made them for longer than IBM did at this point.
Now I know that employers will ultimately replace many employees with automation but this plan as stated seems rather ambitious if not completely unrealistic.
Based on the last few lines of the article I think this CEO is trying to hype up a company that is facing serious struggles and decline of their core business:
> BT said revenue fell 1% to £20.7 billion ($25.8 billion) for the year to March, with growth in Openreach, its fibre broadband network, “more than offset” by declines in other businesses. Its adjusted earnings rose 5% to £7.9 billion ($9.8 billion).
> Shares in the company tumbled 8% in London, as higher expenditure leading to a drop in free cash flow disappointed investors.
This other quote is hilarious to me as well:
> “Our chatbot Amy deals with lots of customer queries already”
Any time you see a company that has a chatbot with a friendly human name you know it’s going to be a completely useless hunk of garbage. I am highly skeptical that “Amy” does anything besides delaying customers from talking to the human who can actually fix their problem (or just throwing them into a frustration loop that leads them toward canceling their service).
It's the constant yin-yang of MBA types running companies. They proclaim some modernization effort or another will let them downsize the expensive employees (or they just do a layoff to bump the stock price).
Then when the project fails to deliver, the sales pipeline dries up, projects stop getting done, etc they'll announce a hiring spree with great fanfare. Look at me, I'm a great CEO and leader! I'm going to re-hire the X positions we laid off or eliminated over the past 5 years under my prior leadership to solve the problems I created!
In the UK BT is a captive utility so shuffling the deck chairs is as good as doing real work I guess.
Around 2026, look for massive hiring as these large scale AI Automation projects colossally fail. It is quite simple: the AI surface is moving to fast for large scale automation because the "must have latest candy" executives will not be satisfied with a solution using the SOTA when they started the project because by the time that project is up and operating the SOTA will be a milestone ahead - so the executives will insist/demand on an update. Rinse and repeat, the project will never complete before expectations and financing die for the vast majority of these AI ambitions.
> “Our chatbot Amy deals with lots of customer queries already,” he added. BT was beginning to explore new products and services that might come from “generative AI and large language model AIs.”
Im quite sure the chat bot is an annoying veneer over the help pages, and a hindrance to actually speaking to a human, because youve already checked the help pages and need to talk to a human to sort the problem.
the games industry moved from guess the verb to point and click adventure games over 30 years ago. why is customer service going the other way?
I'm working on an AI agent that functions as a Project Manager. It will pull Jira task statuses, ask for updates, lead standup, and measure velocity - all with a bit of sass.
Once in a while, you should swap out the PM’s personality with a new one and your robot HR manager can make the firing announcement. Come to think of it, every company could use a few AI-based middle managers for entertainment purposes.
As somebody who went through the MBA education for defensive reasons (i.e. acquiring skills to protect me from managerial BS), I would say you can extend it to most business jobs excluding accountants and quants.
Come to think of it, having a fake management layer that endlessly has meetings and makes “decisions” could be an effective anti-phishing mechanism. Try your luck finding the 1% of my employees that are living, breathing humans with any access to real resources!
woah. As I mentioned, this thing is going to have sass. Do your push notifications ask you why your task too way too long, implying you've been awol? I don't think so.
It's buried in the article, but they were doing a massive modernisation project so hired to deliver it. Now that it's done they'll downsize.
Obviously everyone is focussing on the AI part.
EDIT: Sorry scratch that. I've confused the URL. I didn't see this is a CNN article. This one https://www.bbc.co.uk/news/business-65631168
from bbc is much better (which I had read earlier).
It shows the breakdown of where the jobs lost come from.
> but they were doing a massive modernisation project so hired to deliver it. Now that it's done they'll downsize.
Broadly right, but your tense is wrong. Openreach's[1] FTTP rollout program extends to the end of December 2026. Realistically, it may be delayed, because the scale is rather huge.
[1] A subsidiary of BT who own the infrastructure, for competition grounds
I was rather surprised when Openreach came round and fitted our fairly rural area in Scotland with FTTP. We went from 20 down and 1 up to 1000 down and up.
Might not be 1000 up but certainly seems like it compared to what we had before!
Like usual the UK have already fitted substandard infrastructure.
The BT "FTTP" network is in fact shared fibre with up to 63 of your neighbours. It's based on old GPON technology.
This is why you may get 1Gb/s down but you will be limited to approximately 110Mb/s up.
The reason being that your downstream is broadcasted to all terminals, but AES encrypted (so your traffic is also sent to your neighbours) but the upstream traffic is sent using time division, so you get 1/64th of a unit to upload.
Then everyone is over provisioned, and over sold, so they give you a number that is higher than 1/64 because of the assumption that you and your neighbours won't all trying to use your maximum bandwidth at once.
Compare this to HyperOptic, or CityFibre who will plumb a slither of glass from your home, directly to the ISP. You get a full fat ethernet connection straight to the internet, 1Gb/s down, 1Gb/s up. BT could have done this too, but instead we get ADSL version of fibre.
Yes, GPON is overprovisioned (at 1:32 for Openreach). However, that isn't the full explanation for mediocre upload speeds on Openreach. GPON has a total shared capacity of 2.4 Gbps down and 1.2 Gbps. So you'd think if 1 Gbps down is possible, 500 Mbps up would be.
Simply, Openreach probably have mediocre upload speeds on FTTP, so they can protect their leased line profit margins.
CityFibre use GPON, the exact same technology as Openreach. Yes, CityFibre offer a nice symmetric service on top of that, because real world contention is usually much lower than the theoretical maximum. It's also worth noting CityFibre are starting to upgrade to XGS-PON.
> GPON has a total shared capacity of 2.4 Gbps down and 1.2 Gbps. So you'd think if 1 Gbps down is possible, 500 Mbps up would be.
The "download" direction is sent to you and all your neighbours and only you can access the information as it is AES encrypted to your terminal (CPE, etc.). So you could get 2.4Gbps down, as long as your neighbours are not using it.
1.2Gbps "upload" direction is time shared, so you will never get the full 1.2Gpbs unless you have literally zero neighbours / other terminals on the optical bundle.
BT has been downsizing for years, it is a shell of its former self already.
What strikes me as odd is they're viewing 'modernisation' efforts as one-time activities. Come 2030 are they saying they won't need to invest in their network as much or upgrade things? I think that's what they are saying, and I don't think they're going to be right.
I think once you have fiber to peoples homes, you don’t really need to dig it all up again for probably another 50 years. Replacing every copper cable with fiber in the country is a gargantuan effort.
That might be true if it was proper FTTP, as optical glass can last centuries, and is the perfect practical medium for information transfer. You could probably do it once and never have to replace it ever again.
But BT cheaped out and put down GPON. Maybe in 50 years time the average household will need more than 100Mb/s upload.
First, I seriously doubt AI will replaced 55K people by 2030. However, even if it turns out to be a mess and we look back and laugh at it like we do with BTC. The negative side of this type of corporate behavior will only push more and more people away from the traditional work force or just work force completely. I'm not a big fan of modern capitalism and its overworking culture, but societies in generally can't handle massive amount of able body people who simply don't work. Monetary issue is one thing, but mental health will significantly impact our society safety.
BTC isn't even a good comparison in this case. We can talk about its intentions endlessly, but the fact of the matter is that you can buy some actually useful stuff with it at the end of the day (pornography, VPN services, domain names, computer and electrical equipment from NewEgg, etc), pay state taxes with it in Colorado (and taxes in the Swiss canton of Zug), and it has even gained legal tender status in some countries.
Generative AI and stuff like that will be much easier to judge the success or failure of. If we are ripping our hair out when contacting customer support 15 years from now (roughly the amount of time Bitcoin has existed), generative AI (I assume that's what you're talking about) for the purposes of customer support is a failure. Beyond that, sophisticated generative AI is the digital equivalent of discovering nuclear fission: it destroys online human discourse and pollutes the areas it infects with artificial garbage (fallout). I know I'm not the only one who's sick of seeing DALL-E stuff on Instagram and DeviantArt, and seeing stuff obviously written by LLMs in internet comment sections.
The AI mention is interesting even if PR. There's a wide gap between no AI, and AI doing all the hard / boring stuff so people can be happy. And the way we (might) transition from the former to the latter will have a big impact on many people. If AI becomes a wealth only accessible to some economic classes leveraging it for profits, it will propel pain toward the other. Anybody trying to ensure this evolves in a smooth way for everybody ?
> “Our chatbot Amy deals with lots of customer queries already,” he added.
> Jansen also said that around 10,000 fewer people would be needed to service and repair digital networks
I believe this is what C. Doctorow would describe as “enshittening” customer support operations.
It doesn’t require AI or any other magic, just a juicy performance-based incentive and a big parachute. The next CEO will focus on “improving the customer experience”.
59 comments
[ 1032 ms ] story [ 3672 ms ] threadI haven’t dealt with them in recent years but based on previous dealings, I would put BT firmly in that category.
Their checkout flow can break. In some browsers it'll get stuck in an infinite redirect loop. If you're using edge you'll get redirected to a page that shows the ~10 year old BT Infinity branding, with a totally different style, packages, and prices.
Their chat bot is of the kind that decides whether or not you're allowed to speak to a human based on what it thinks your question is.
Sigh... it's depressing just reading this... you have my sympathy! :)
"Jansen also said that around 10,000 fewer people would be needed to service and repair digital networks, which “go wrong less often” and can be fixed more easily than older networks."
That feels optimistic. Modernization does help with that sort of thing in some ways. But it also often adds complexity, higher dependency counts, etc.
Sure, but nobody will remember someone said that seven years before. They might be hiring by that point, too.
But I'm with you in the sense that new technology that does more than what they replaced will absolutely be added and need fresh workforces.
The complexity is inside opaque boxes that BT deploys though. Rather than fixing things like they did in the past, sitting at the end of a street patching cables in a roadside cabinet, BT's engineers of the future will just unclip a box that's returning an error code and plug in a new one. Or someone in a NOC will reset it from 300 miles away. Maybe some AI will do that. Or a cron job. With fibre networks once the cabinet is installed and plugged into the network there's very little physical work to do. That's what engineers do now - manually splicing and connecting copper wire. That's the work that's going away.
So you don't need phone operators to wait for calls about interrupted service, the software will just detect some unreachable part of the network and dispatch a team to investigate.
To say it’s all based on copper cables, it was pretty impressive.
it's going to be more reliable even if they completely screw everything up
A copper line isn't a continuous thing from the exchange to the customer. It's made of a number of sections that are joined together.
Now someone who once worked for a large British telecoms company once told me that they'd done some modelling, maybe using something like system dynamics if I was guessing, to gauge the effect of work on the network.
Any interaction with the equipment, whether to set up lines for a customer or even fix faults leads to more faults. So it ends up being more profitable to just leave equipment in place than to reuse it.
Based on that I would say that having all the pairs terminating at the cab instead of the exchange and using remotely configurable equipment would be a greater improvement again.
Just guessing you understand, I have no inside knowledge. ;)
The number of RFOs I see which report outages due to either fibre disturbances in an exchange or is phenomenal. I don't really have any idea how that compares with copper, but my home DSL runs over copper to the exchange and doesn't go down anywhere near as much.
What they mean is that consumers will be stuck with slow internet connections. As a result of customers switching to other providers they’ll have less equipment breaking down. Since you know it’s no longer in use.
If I needed to do a layoff now I’d just say “AI LLM Chatbot!!” woo-woo as cover.
Besides, the full 55k are about all automation, restructuring and digitisation efforts. Doesn't sound entirely unreasonable.
They probably are also looking to justify investment and say that they are investing into making the business more profitable for the next 7 years, so don't expect too much profit until after this investment is done.
The last good thing about IBM that I remember is the Thinkpad, but Lenovo's made them for longer than IBM did at this point.
Based on the last few lines of the article I think this CEO is trying to hype up a company that is facing serious struggles and decline of their core business:
> BT said revenue fell 1% to £20.7 billion ($25.8 billion) for the year to March, with growth in Openreach, its fibre broadband network, “more than offset” by declines in other businesses. Its adjusted earnings rose 5% to £7.9 billion ($9.8 billion).
> Shares in the company tumbled 8% in London, as higher expenditure leading to a drop in free cash flow disappointed investors.
This other quote is hilarious to me as well:
> “Our chatbot Amy deals with lots of customer queries already”
Any time you see a company that has a chatbot with a friendly human name you know it’s going to be a completely useless hunk of garbage. I am highly skeptical that “Amy” does anything besides delaying customers from talking to the human who can actually fix their problem (or just throwing them into a frustration loop that leads them toward canceling their service).
Then when the project fails to deliver, the sales pipeline dries up, projects stop getting done, etc they'll announce a hiring spree with great fanfare. Look at me, I'm a great CEO and leader! I'm going to re-hire the X positions we laid off or eliminated over the past 5 years under my prior leadership to solve the problems I created!
In the UK BT is a captive utility so shuffling the deck chairs is as good as doing real work I guess.
Im quite sure the chat bot is an annoying veneer over the help pages, and a hindrance to actually speaking to a human, because youve already checked the help pages and need to talk to a human to sort the problem.
the games industry moved from guess the verb to point and click adventure games over 30 years ago. why is customer service going the other way?
Obviously everyone is focussing on the AI part.
EDIT: Sorry scratch that. I've confused the URL. I didn't see this is a CNN article. This one https://www.bbc.co.uk/news/business-65631168 from bbc is much better (which I had read earlier). It shows the breakdown of where the jobs lost come from.
Broadly right, but your tense is wrong. Openreach's[1] FTTP rollout program extends to the end of December 2026. Realistically, it may be delayed, because the scale is rather huge.
[1] A subsidiary of BT who own the infrastructure, for competition grounds
Might not be 1000 up but certainly seems like it compared to what we had before!
Like usual the UK have already fitted substandard infrastructure.
The BT "FTTP" network is in fact shared fibre with up to 63 of your neighbours. It's based on old GPON technology.
This is why you may get 1Gb/s down but you will be limited to approximately 110Mb/s up.
The reason being that your downstream is broadcasted to all terminals, but AES encrypted (so your traffic is also sent to your neighbours) but the upstream traffic is sent using time division, so you get 1/64th of a unit to upload.
Then everyone is over provisioned, and over sold, so they give you a number that is higher than 1/64 because of the assumption that you and your neighbours won't all trying to use your maximum bandwidth at once.
Compare this to HyperOptic, or CityFibre who will plumb a slither of glass from your home, directly to the ISP. You get a full fat ethernet connection straight to the internet, 1Gb/s down, 1Gb/s up. BT could have done this too, but instead we get ADSL version of fibre.
Simply, Openreach probably have mediocre upload speeds on FTTP, so they can protect their leased line profit margins.
CityFibre use GPON, the exact same technology as Openreach. Yes, CityFibre offer a nice symmetric service on top of that, because real world contention is usually much lower than the theoretical maximum. It's also worth noting CityFibre are starting to upgrade to XGS-PON.
The "download" direction is sent to you and all your neighbours and only you can access the information as it is AES encrypted to your terminal (CPE, etc.). So you could get 2.4Gbps down, as long as your neighbours are not using it.
1.2Gbps "upload" direction is time shared, so you will never get the full 1.2Gpbs unless you have literally zero neighbours / other terminals on the optical bundle.
Again, CityFibre is not Point to Point fibre. They offer 1Gbps upload on GPON, contended at 1:32. From the horse's mouth: https://cdn.cityfibre.com/Partner-page/Downloads/CF-FTTH-Loc...
Even Openreach have a 220Mbit/s upload tier, but it costs more than twice as much so nobody offers it for home users. https://www.openreach.co.uk/orpg/home/products/pricing/loadP...
What strikes me as odd is they're viewing 'modernisation' efforts as one-time activities. Come 2030 are they saying they won't need to invest in their network as much or upgrade things? I think that's what they are saying, and I don't think they're going to be right.
But BT cheaped out and put down GPON. Maybe in 50 years time the average household will need more than 100Mb/s upload.
Generative AI and stuff like that will be much easier to judge the success or failure of. If we are ripping our hair out when contacting customer support 15 years from now (roughly the amount of time Bitcoin has existed), generative AI (I assume that's what you're talking about) for the purposes of customer support is a failure. Beyond that, sophisticated generative AI is the digital equivalent of discovering nuclear fission: it destroys online human discourse and pollutes the areas it infects with artificial garbage (fallout). I know I'm not the only one who's sick of seeing DALL-E stuff on Instagram and DeviantArt, and seeing stuff obviously written by LLMs in internet comment sections.
What kind of culture and employee morale do they expect to have now that they've announced this?
> Jansen also said that around 10,000 fewer people would be needed to service and repair digital networks
I believe this is what C. Doctorow would describe as “enshittening” customer support operations.
It doesn’t require AI or any other magic, just a juicy performance-based incentive and a big parachute. The next CEO will focus on “improving the customer experience”.