How Soon Might Humans Be Replaced At Work

As noted by Thomas Claburn in The Register, there seems to be a contradiction between two pieces of research relating to the development and use of AI in business organizations.

On the one hand, teams of researchers have developed benchmarks to study the effectiveness of AI, and have found success rates between 25% and 40%, depending on the situation.

On the other hand, Gartner reports that business executives are expecting a success rate nearer to 60% - if we interpret not-being-cancelled as a marker for success. More than 40 percent of agentic AI projects will be cancelled by the end of 2027 due to rising costs, unclear business value, or insufficient risk controls.

 

History tells us that the adoption of technology to perform work is only partially dependent on the quality of the work, and can often be driven more by cost. The original Luddites protested at the adoption of machines to replace textile workers, but their argument was largely based on the inferior quality of the textiles produced by the machines. It was only later that this label was attached to anyone who resisted technology on principle.

Around ten years ago, I attended a debate on artificial intelligence sponsored by the Chartered Institute of Patent Agents. In my commentary on this debate (How Soon Might Humans Be Replaced At Work?) I noted that decision-makers may easily be tempted by short-term cost savings from automation, even if the poor quality of the work results in higher costs and risks in the longer term.

In their look at the labour market potential of AI, Tyna Eloundou et al note that

A key determinant of their utility is the level of confidence humans place in them and how humans adapt their habits. For instance, in the legal profession, the models’ usefulness depends on whether legal professionals can trust model outputs without verifying original documents or conducting independent research. ... Consequently, a comprehensive understanding of the adoption and use of LLMs by workers and firms requires a more in-depth exploration of these intricacies.

However, while levels of confidence and trust can be assessed by surveying people's opinions, such surveys cannot assess whether these levels of confidence and trust are justified. Graham Neubig told The Register that this was what prompted the development of a more objective benchmark for AI effectiveness.

---

Thomas Claburn, AI agents get office tasks wrong around 70% of the time, and a lot of them aren't AI at all (The Register, 29 June 2025)

Thomas Claburn, AI has had zero effect on jobs so far, says Yale study (The Register, 1 October 2025)

Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock, GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models (August 2023)

Wikipedia: Luddite 

Related Posts: How Soon Might Humans Be Replaced At Work? November 2015, RPA - Real Value or Painful Experimentation? (August 2019), Explaining Layoffs (October 2025)

Automation Ethics

Many people start their journey into the ethics of automation and robotics by looking at Asimov's Laws of Robotics.

A robot may not injure a human being or, through inaction, allow a human being to come to harm (etc. etc.)

As I've said before, I believe Asimov's Laws are problematic as a basis for ethical principles. Given that Asimov's stories demonstrate numerous ways in which the Laws don't actually work as intended. I have always regarded Asimov's work as being satirical rather than prescriptive.

While we usually don't want robots to harm people (although some people may argue for this principle to be partially suspended in the event of a "just war"), notions of harm are not straightforward. For example, a robot surgeon would have to cut the patient (minor harm) in order to perform an essential operation (major benefit). How essential or beneficial does the operation need to be, in order to justify it? Is the patient's consent sufficient?

Harm can be individual or collective. One potential harm from automation is that even if it creates wealth overall, it may shift wealth and employment opportunities away from some people, at least in the short term. But perhaps this can be justified in terms of the broader social benefit, or in terms of technological inevitability.

And besides the avoidance of (unnecessary) harm, there are some other principles to think about.

• Human-centred work - Humans should be supported by robots, not the other way around. 

• Whole system solutions - Design the whole system or process, don’t just optimize a robot as a single component.  

• Self-correcting - Ensure that the system is capable of detecting and learning from errors. 

• Open - Providing space for learning and future disruption. Don't just pave the cow-paths.

• Transparent - The internal state and decision-making processes of a robot are accessible to (some) users.  

Let's look at each of these in more detail.


Human-Centred Work

Humans should be supported by robots, not the other way around. So we don't just leave humans to handle the bits and pieces that can't be automated, but try to design coherent and meaningful jobs for humans, with robots to make them more powerful, efficient, and effective.

Organization theorists have identified a number of job characteristics associated with job satisfaction, including skill variety, task identity, task significance, autonomy and feedback. So we should be able to consider how a given automation project affects these characteristics.


Whole Systems

When we take an architectural approach to planning and designing new technology, we can look at the whole system rather than merely trying to optimize a single robotic component.

• Look across the business and technology domains (e.g. POLDAT).
• Look at the total impact of a collection of automated devices, not at each device separately.
• Look at this as a sociotechnical system, involving humans and robots collaborating on the business process.

Self-Correcting

Ensure that the (whole) system is capable of detecting and learning from errors (including near misses).

This typically requires a multi-loop learning process. The machines may handle the inner learning loops, but human intervention will be necessary for the outer loops.
 

Open

Okay, so do you improve the process first and then automate it, or do you automate first? If you search the Internet for "paving the cow-paths", you can find strong opinions on both sides of this argument. But the important point here is that automation shouldn't close down all possibility of future change. Paving the cow-paths may be okay, but not just paving the cow-paths and thinking that's the end of the matter.

In some contexts, this may mean leaving a small proportion of cases to be handled manually, so that human know-how is not completely lost. (Lewis Mumford argued that it is generally beneficial to retain some "craft" production alongside automated "factory" production, as a means to further insight, discovery and invention.)


Transparency

The internal state and decision-making processes of a robot are accessible to (some) users. Provide ways to monitor and explain what the robots are up to, or to provide an audit trail in the event of something going wrong.

---

Related posts

How Soon Might Humans Be Replaced At Work? (November 2015) Could we switch the algorithms off? (July 2017), How many ethical principles? (April 2019), Responsible Transparency (April 2019), Process Automation and Intelligence (August 2019), RPA - Real Value or Painful Experimentation? (August 2019)

Links

Jim Highsmith, Paving Cow Paths (21 June 2005)

Wikipedia

Job Characteristic Theory
Just War Theory

Fail Fast - Burger Robotics

As @jjvincent observes, integrating robots into human jobs is tougher than it looks. Four days after it was installed in a Pasadena CA burger joint, Flippy the robot has been taken out of service for an upgrade. Turns out it wasn't fast enough to handle the demand. Does this count as Fail Fast?

Flippy's human minders have put a positive spin on the failure, crediting the presence of the robot for an unexpected increase in demand. As Vincent wryly suggests, Flippy is primarily earning its keep as a visitor attraction.

If this is a failure at all, what kind of failure is it? Drawing on earlier work by James Reason, Phil Boxer distinguishes between errors of intention, planning and execution.

If the intention for the robot is to improve productivity and throughput at peak periods, then the designers have got more work to do. And the productivity-throughput problem may be broader than just burger flipping: making Flippy faster may simply expose a bottleneck somewhere else in the system. But if the intention for the robot is to attract customers, this is of greatest value at off-peak periods. In which case, perhaps the robot already works perfectly.

---

Philip Boxer, ‘Unintentional’ errors and unconscious valencies (Asymmetric Leadership, 1 May 2008)

John Donohue, Fail Fast, Fail Often, Fail Everywhere (New Yorker, 31 May 2015)

Lora Kolodny, Meet Flippy, a burger-grilling robot from Miso Robotics and CaliBurger (TechCrunch 7 Mar 2017)

Brian Heater, Flippy, the robot hamburger chef, goes to work (TechCrunch, 5 March 2018)

James Vincent, Burger-flipping robot takes four-day break immediately after landing new job (Verge, 8 March 2018)





Related post Fail Fast - Why did the chicken cross the road? (March 2018)

Working for the Machine

#orgintelligence The recent appointment of an algorithm to a Board of Directors raises the spectre of science fiction becoming fact. Although many commentators regarded the appointment as a publicity stunt, there has always been an undercurrent of fear about machine intelligence. Even the BBC (following Betteridge's Law of Headlines) succumbed to the alarmist headline Could a big data-crunching machine be your boss one day?

There are several useful ways that an algorithm might contribute to the collective intelligence of a Board of Directors. One is to provide an automated judgement on some topic, which can be put into the pot together with a number of human judgements. This is what seems to be planned by the company Deep Knowledge Ventures, whose Board of Directors is faced with a series of important investment decisions. Although each decision is unique, there are some basic similarities in the decision process that may be amenable to automation and machine learning.

Another possible contribution is to evaluate other board members. According to the BBC article, IBM Watson could be programmed to analyse the contributions made by each board member for usefulness and accuracy. There are several ways such a feedback loop could enhance the collective intelligence of the Board.

• Retrain individuals to improve their contributions in specific contexts.
• Identify and eliminate individuals whose contribution is weak.
• Identify and eliminate individuals whose contribution is similar to other members. In other words, promote greater diversity.
• Enable trial membership of individuals from a wider range of backgrounds, to see whether they can make a valuable contribution.

Organizational Intelligence is about an effective combination of human/social intelligence and machine intelligence. Remember this when people try to develop an either-us-or-them narrative.

---

#QTWTAIN

Jamie Bartlett, Will Artificial Intelligence put my job at risk? (Spectator 6 June 2014)

Adrian Chen, Can an Algorithm Solve Twitter’s Credibility Problem? (New Yorker 5 May 2014)

John Rentoul, Will Artificial Intelligence put my job at risk? (Independent 6 June 2014)

Richard Veryard, Does Cameron's Dashboard App Improve the OrgIntelligence of Government? (23 January 2013)

Matthew Wall, Could a big data-crunching machine be your boss one day? (BBC News 9 October 2014)

---

Other Sources

Algorithm appointed board director (BBC News 16 May 2014)

Bud Caddell, Your boss might be better as an algorithm (Quartz, 16 November 2017)


Related Posts

Intelligence and Governance (Feb 2013)


Link added 11 December 2017

Art and the Enterprise

Artistic Innovation

Sometimes artists produce amazing innovations, with painstaking labour. Subsequent technology makes this painstaking labour unnecessary. We can recognize two separate innovations - the product innovation (what the artist produces) and the process/production innovation (what the technology produces).

What is the linkage between these two innovations? To what extent has the artistic innovation stimulated the technological innovation, established a proof-of-concept which technologists can then implement. Great art changes the way we perceive the world, and this may include changing our understanding of the possible.

One example I have quoted a couple of times is Karlheinz Stockhausen and the synthesizer. In the 1950s, Stockhausen and other composers produced some innovative pieces of electronic music, for which every sound had to be hand-crafted. Once the synthesizer had been invented, similar music could be easily produced with a few quick knob-twiddles. As a result of the widespread use of synthesizers in popular music, as well as the many rock musicians (Beatles, Can, Zappa) who pay explicit tribute to Stockhausen, a piece like Kontakte sounds a lot less strange to the modern ear than it did to the contemporary ear.

On this point I disagree with Brian Eno, who once commented that "Stockhausen was an example of a charismatic theoretician who inspired a lot of people but whose own work is generally unlistenable." [source: MOJO April 1997, interview with Andy Gill]

Does this familiarity diminish the striking originality of Stockhausen's work? Or should we regard Stockhausen's achievement with greater respect, because of his lack of tools, and his (arguable) influence over later technology as well as (acknowledged and unacknowledged) influence over later music-making.

I should also mention the BBC Radiophonic Workshop here. The brilliant Delia Derbyshire produced the original theme music for Doctor Who using hand-made equipment - music that today still sounds utterly wild and futuristic. (Sadly, the BBC no longer uses this version, and now plays a tame remix recorded with modern equipment.)

Similar examples can be found in the visual arts before the invention of photography. Many artists developed styles of painting and perspective which predated the modern camera.

Enterprise Innovation

I think there are three important patterns here that may be relevant for enterprise innovation.

• Innovation before automation. Don't automate something until you have understood it, simplified it, improved it. (This principle is sometimes known by the slogan: Don't pave the cow-paths.)

• Retain the capability for further innovation. Automation should not eliminate the possibility of hands-on creativity. Lewis Mumford (in Technics and Civilization) argues that it is generally beneficial to retain some 'craft' production alongside automated 'factory' production, as a source of 'education, recreation and experiment' and 'as a means to further insight, discovery and invention'.

• Invent in order to innovate. A composer such as Thomas Dolby, who sets out to invent new tools for producing music (including building new hardware and software), may thereby be able to produce music that is different to what everyone else is producing.

---

For further discussion of Stockhausen, see my posts on Thomas Dolby's keynote speech at the 2005 Rational Conference (May 2005), Lightweight Enterprise (March 2006), and Grandpa's SOA (Oct 2007).

Chloe Glover, Manchester honours the woman behind the pioneering music of Doctor Who (Guardian 10 Jan 2013)

Updated 12 Jan 2013 (Delia Derbyshire Day)
Updated 12 March 2015 (removed link to wrong Andy Gill)

Automation and Skill

New technologies often present an interesting choice.

• The old way of doing something presents us with a low level of difficulty all the time;
  
  
• A more sophisticated/automated way of doing it spares us the difficulty most of the time, only to present us with occasional bursts of extreme difficulty.

The latter may deliver much higher productivity when you have gotten up the learning curve, but it's a much steeper learning curve. You can't progress from simple tasks to hard tasks, if the technology has already captured all the simple tasks. New technologies call for a concentration of skills, but interfere with our ability to learn these skills effectively.



Software example: code generation tools.