Safe CEO: AI is an assistant, not a replacement
(2026/01/02)
- Reference: 1767344406
- News link: https://www.theregister.co.uk/2026/01/02/safe_ceo_interview/
- Source link:
Interview If AI can take on the role of a junior programmer, what happens when senior staff start retiring? Industry veteran and CEO of Safe Software, Don Murray, reckons the technology is becoming indispensable, but the human can never be removed from the loop.
"I think as an assistant," says Murray, "AI is indispensable. But it is not an authority."
Murray's company, Safe, specializes in data collation. Be it real-time or static, data flows through Safe's data pipe and can be output to any visualization tool the customer is using. It is therefore ripe for AI applications, both in mining the flow of data and in pursuing an agentic AI path to take action. Murray, having seen industry fads come and go over a multi-decade career, is therefore well placed to comment on the latest tech industry obsession: AI.
[1]
"If there's one thing AI has a never-ending thirst for," he says, "it's data."
[2]
[3]
Workers might also worry that AI – or its proponents – have a thirst for something else: their jobs.
Murray, while an enthusiast for the technology, is also a realist about its application. Yes, it can certainly appear magical when it comes to accessibility and user interaction (Murray uses the example of the technology being used to build systems that work the way users do, instead of forcing users into a box defined by technology), but he is also a realist about the service and what it can do, cannot do and should never do.
[4]
Murray cites the example of a company working on a major engineering project in the UK (Safe has many such customers, including airports and power networks) that asked a vendor: "How accurate is your AI?" The answer, "oh, you can get to 80 or 90 percent," was nowhere near the requirement, which is more like 99.999 percent, certainly when it comes to engineering.
"And so one of the things you're going to see in agentic systems," he says, "is this 'human in the loop'.
"Basically, the AI will do its magic, and then it will throw it to a human who has the credentials to sign off on the design.
[5]
"AI can help you build a bridge. If you're an engineer, it can go through lots of different designs and help you get there, but there's no organization that's going to allow it to sign off."
Instead, Murray reckons AI will be an assistant. Able to throw up some suggestions to, for example, a medical professional, and come up with ideas that might never have occurred to a Doctor. But it can't be the final authority on a subject.
The problem comes with companies that see AI as more than an assistant. "I talk to utility companies," says Murray, "and they're saying that the experts who have been in the business forever use AI. The company is saying, 'We don't have to hire junior people.'
"So then I'm like, 'When all those people retire, who's going to be the expert now? You're not giving these people the experience for 10, 20, 30 years so they can step into the role. There's only one way to get experience, and that's time and actually experiencing things."
Murray is a veteran of the tech industry and practices what he preaches when it comes to AI adoption for his own staff. "We're using it to bring everybody up to speed," he explains, with the assistant having an impact at all levels.
[6]One real reason AI isn't delivering: Meatbags in manglement
[7]Microsoft wants to replace its entire C and C++ codebase, perhaps by 2030
[8]SoftBank scrambling to come up with $22.5B in OpenAI funding before New Year
[9]AI is rewriting how power flows through the datacenter
"I have this guy who is amazing and has been around forever, and AI for him is very valuable because he can see very quickly what is correct and what isn't correct. Whereas the junior people, it is a benefit that can help them write some code quickly, but they still have to go through the testing."
Murray reckons that, instead, AI is useful for suggesting how a given system might be supported or which libraries to use.
"We're hiring juniors and seniors," he says, "and then we're using it to bring everybody up to speed."
As far as Murray is concerned, the genie is out of the bottle. "People say," he says, "it's just a bubble. Maybe it is just a bubble. But AI is here for good; it's sort of like the Internet. It was a bubble, but it didn't go away [when the bubble burst]."
Regardless of what happens, "AI is here to stay." ®
Get our [10]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/12/24/reason_ai_isnt_delivering/
[7] https://www.theregister.com/2025/12/24/microsoft_rust_codebase_migration/
[8] https://www.theregister.com/2025/12/22/softbank_funding_openai/
[9] https://www.theregister.com/2025/12/22/ai_power_datacenter/
[10] https://whitepapers.theregister.com/
"I think as an assistant," says Murray, "AI is indispensable. But it is not an authority."
Murray's company, Safe, specializes in data collation. Be it real-time or static, data flows through Safe's data pipe and can be output to any visualization tool the customer is using. It is therefore ripe for AI applications, both in mining the flow of data and in pursuing an agentic AI path to take action. Murray, having seen industry fads come and go over a multi-decade career, is therefore well placed to comment on the latest tech industry obsession: AI.
[1]
"If there's one thing AI has a never-ending thirst for," he says, "it's data."
[2]
[3]
Workers might also worry that AI – or its proponents – have a thirst for something else: their jobs.
Murray, while an enthusiast for the technology, is also a realist about its application. Yes, it can certainly appear magical when it comes to accessibility and user interaction (Murray uses the example of the technology being used to build systems that work the way users do, instead of forcing users into a box defined by technology), but he is also a realist about the service and what it can do, cannot do and should never do.
[4]
Murray cites the example of a company working on a major engineering project in the UK (Safe has many such customers, including airports and power networks) that asked a vendor: "How accurate is your AI?" The answer, "oh, you can get to 80 or 90 percent," was nowhere near the requirement, which is more like 99.999 percent, certainly when it comes to engineering.
"And so one of the things you're going to see in agentic systems," he says, "is this 'human in the loop'.
"Basically, the AI will do its magic, and then it will throw it to a human who has the credentials to sign off on the design.
[5]
"AI can help you build a bridge. If you're an engineer, it can go through lots of different designs and help you get there, but there's no organization that's going to allow it to sign off."
Instead, Murray reckons AI will be an assistant. Able to throw up some suggestions to, for example, a medical professional, and come up with ideas that might never have occurred to a Doctor. But it can't be the final authority on a subject.
The problem comes with companies that see AI as more than an assistant. "I talk to utility companies," says Murray, "and they're saying that the experts who have been in the business forever use AI. The company is saying, 'We don't have to hire junior people.'
"So then I'm like, 'When all those people retire, who's going to be the expert now? You're not giving these people the experience for 10, 20, 30 years so they can step into the role. There's only one way to get experience, and that's time and actually experiencing things."
Murray is a veteran of the tech industry and practices what he preaches when it comes to AI adoption for his own staff. "We're using it to bring everybody up to speed," he explains, with the assistant having an impact at all levels.
[6]One real reason AI isn't delivering: Meatbags in manglement
[7]Microsoft wants to replace its entire C and C++ codebase, perhaps by 2030
[8]SoftBank scrambling to come up with $22.5B in OpenAI funding before New Year
[9]AI is rewriting how power flows through the datacenter
"I have this guy who is amazing and has been around forever, and AI for him is very valuable because he can see very quickly what is correct and what isn't correct. Whereas the junior people, it is a benefit that can help them write some code quickly, but they still have to go through the testing."
Murray reckons that, instead, AI is useful for suggesting how a given system might be supported or which libraries to use.
"We're hiring juniors and seniors," he says, "and then we're using it to bring everybody up to speed."
As far as Murray is concerned, the genie is out of the bottle. "People say," he says, "it's just a bubble. Maybe it is just a bubble. But AI is here for good; it's sort of like the Internet. It was a bubble, but it didn't go away [when the bubble burst]."
Regardless of what happens, "AI is here to stay." ®
Get our [10]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aVelT1ep7AKPD7pP5gci1gAAAAY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/12/24/reason_ai_isnt_delivering/
[7] https://www.theregister.com/2025/12/24/microsoft_rust_codebase_migration/
[8] https://www.theregister.com/2025/12/22/softbank_funding_openai/
[9] https://www.theregister.com/2025/12/22/ai_power_datacenter/
[10] https://whitepapers.theregister.com/
druck
In terms of software engineers, its very hard to argue that AI hasn't outpaced its entire cohort of junior devs in terms of quality of output.
Only if you think a junior dev is one that is complete shit and incapable of learning anything.
"Experience" is only one dimension
Bebu sa Ware
I was thinking about the senior people in skilled roles.
While years of experience is one characteristic, another is the diversity of strengths and by implication weaknesses of particular individuals in identical roles.
One might be a devil for detail and irreplaceable when troubleshooting the most subtle of problems, but have the imagination of an extinguished candle; another might have a double helping of imagination and creativity but blind to a small but critical detail; another blessed with the penchant of asking the right questions - most of which they themselves are incapable of answering; and so it goes.
The intellectual life of people, or at least some, is exceedingly complex (natural languages alone make that quite clear) and undoubtedly requires far more than the trumped up finite state automata that passes for AI, to replicate.
In the wash·up after the damage has mostly been remedied, I imagine AI/LLM technology will be just slightly more useful than expert systems were after the somewhat fainter bloom of AI 1.0 in the 1980·90s. That is to say there, in the background, doing useful work in very narrow well defined domains, but largely invisible.
Not convinced
I'd like things to play out the way the author suggests, but unfortunately, I think he's wrong while I want him to be right.
AI does do a lot of what juniors do in many professions, including making lots of mistakes. The real debate is by the time today's juniors become tomorrows seniors (actual seniors not juniors with 5 years experience and a wholly inappropriate job title), what will AI have become? For some, perhaps many, roles its not inconceivable that it'll be capable of doing senior level work and so the need for seniors will have been reduced or absorbed.
Translators, for example, will never be needed in anything like the same quantity as before AI, if at all by the time a junior translator hopes to become a senior translator in say 15 to 20 years.
In terms of software engineers, its very hard to argue that AI hasn't outpaced its entire cohort of junior devs in terms of quality of output. There's a credible argument that it won't continue to do so, but there's an equally plausible argument that it does. While personally I think there will still be a need for senior engineers, there is a limit of how much money I'd want to bet on that. What if, instead of investing in todays juniors, a business saves that cash (literally) and uses it to fund hiring tomorrows scarcer senior staffer? There's no reason to think that those investing in juniors will reap any reward when they become seniors because they'll go for the money, same as everyone else.
The hype around AI is well overblown, for sure, and its definitely not going to be all things to all people, but there are a lot of careers never coming back, and there's a lot more still that won't require the same volume of people to do the same level of work. Marketing, basic accounts clerks, etc won't be needed in future at anything like the level of employment they've enjoyed previously, and for sure the same is likely to be true of some tech jobs also.