What does a tool change?
When I look at a new AI tool, I’m interested in what it can do. Soon, other questions follow: how will it change the work of the person using it, what will they need to learn, and what happens to the time saved? There may be room for better research and more thoughtful design. There may also be more tasks to complete each day.
MIT’s The Work of the Future: Building Better Jobs in an Age of Intelligent Machines helps organise this second level of discussion. David Autor, David Mindell and Elisabeth Reynolds published it on 17 November 2020. It predates the current wave of generative AI, so I read it as an analysis of the relationship between technology, work organisation and the distribution of gains, grounded in the conditions of its time. [1, pp. 2–6] (publication date)
A task is part of a larger job
The authors show how automation can take over individual tasks, increase productivity in others and create demand for new kinds of work. Their examples include designers using digital tools and doctors supported by developing diagnostic capabilities. [1, pp. 8–11]
This is a useful distinction in project work. Producing a graphic variant, gathering material and organising a document are tasks. A whole assignment also includes understanding needs, choosing a direction, reaching agreement and taking responsibility for the result. It is worth examining how a new tool changes the relationships between those activities.
The historical emergence of new occupations does not guarantee that everyone losing their existing tasks will find a better position. The report itself notes that no economic law guarantees a balance between job creation and destruction. [1, p. 11] For an individual, what matters is the availability of opportunities, location, time and conditions for acquiring new skills.
Productivity and pay are separate questions
A central concern of the report is the divergence between productivity growth and the wages of many workers in the United States. The authors connect these outcomes with labour-market institutions and political decisions. They note that other industrialised countries facing similar technological pressures and global competition achieved better results for workers. [1, pp. 3–6]
The US context matters. I would not transfer individual diagnoses or solutions directly to Poland. What remains relevant to me is the question of sharing gains: who benefits from more efficient work, and which rules determine that?
Within an organisation, the questions can be concrete. Will the new technology reduce repetitive work? Will employees have time to learn? Will they gain autonomy? How will expectations, performance assessment and pay change? The answers shape the value of implementation as much as the feature list.
Job quality also depends on how work is organised and how gains are shared.
Workers hold knowledge that implementation needs
A vision system
A small manufacturing example in the report illustrates this relationship well. After a vision system was introduced at one of the studied plants, the number of reported defects rose sharply. Using their experience and process-control training, workers recognised many alerts as false alarms. They worked with engineers to reposition the system and improve its performance. [1, p. 48]
To me, this is a good picture of collaboration around new technology. The person doing the task can use their knowledge to assess a result, spot a flawed assumption and improve the solution. That is why the ability to challenge a system’s output belongs in the process.
Health Hubs
The report also describes Kaiser Permanente’s Health Hubs. Staff were involved in designing and testing new facilities; simulations of technology, patient flows and work organisation were run before opening. Participants later trained their colleagues. [1, p. 74]
In this example, learning and agreeing on changes were part of implementation. Technology met people’s experience while there was still room to improve how the whole process worked.
Learning needs time and organisation
MIT devotes one area of recommendations to developing skills throughout life, a second to improving job quality and a third to supporting and directing innovation. The authors see these as connected: new qualifications alone do not resolve employment conditions, just as buying technology alone does not create a well-functioning organisation. [1, pp. 5–6]
Applied to a smaller project, this suggests some simple agreements. Which task are we trying to improve? Who knows it from the inside? How will we compare the result with the current way of working? What will we do if a tool speeds up one step but creates more corrections in the next?
Time saved should be assessed alongside quality, workload and responsibility. A faster first version can be valuable if it leaves room for checking. Otherwise, effort can simply move elsewhere in the process.
What kind of work grows around technology?
What interests me most in this report is the possibility of shaping work deliberately. It highlights pay, security, development, workers’ voice and management practices. These are areas where organisational decisions affect how people experience technological progress. [1, pp. 5–6, 72–74]
When I think about future projects and teams, I want to be able to discuss both sides of implementation: what the technology can do, and the everyday work we create around it. I’m interested in organisations that can experiment, share knowledge and give people a say in how they do their work.
The 2020 report remains a useful reference point. Tools’ capabilities need continual reassessment, but questions about job quality are worth asking from the start of every implementation.