Shoshana Zuboff drew this distinction in 1988, watching what happened when paper mills and offices put computers into work that had been done by hand and eye. It is the most useful pair of words available for AI deployment decisions, and almost nobody making those decisions has the words.
The answer, in one line
Automating replaces human judgement with a machine. Informating generates information that deepens the worker's understanding of the work.
Definition#
Automating versus informating: automating replaces human judgement with a machine; informating generates information that deepens the worker's understanding. The same system can do either, and which one happens is a management choice rather than a property of the technology.
The same model, configured two ways#
A system that produces the finished document automates. A system that shows the person why the document should say what it says, what it rests on and where it is weakest, informates. These are frequently the same underlying model with different instructions and a different interface, which is what makes the choice so easy to leave unmade.
Zuboff's observation was that organisations tended to take the automating option by default, because it is the one that shows up in a budget. Informating requires someone to decide that the worker's understanding is an output worth paying for, and no line in the spreadsheet asks for it.
Why it is sharper than augmentation#
Augmentation names an outcome: the person does better with the tool than without it. Informating names a mechanism: the work becomes more visible to the person doing it. The two come apart, and the gap between them is where most of this estate's evidence lives. A tool can lift output while hiding the work completely, which is how an organisation gets better results and worse people at the same time, with nothing on any dashboard showing the second half. That accumulation is capability debt.
What happens when the automated part was the teaching part#
Bainbridge established the consequence forty years ago and it has not been overturned. Automating the routine portion of a task leaves the person the hardest residue, monitoring and handling exceptions, while removing the practice that built the competence to do it. Graded entry. Informating is the answer to that irony rather than a softer version of it: a system that keeps showing the reasoning is one where the practice does not disappear when the labour does.
The strongest recent illustration is Dell'Acqua's field experiment with 776 professionals, where AI erased the difference between what a technical specialist and a commercial specialist proposed. Everyone's output converged on the balanced version. That is an automating configuration behaving as advertised, and the bill is the thing the organisation hired two different kinds of expert for. Graded entry.
The word Zuboff has to do without here#
Nothing on this estate measures informating against automating directly. No study takes one organisation, configures a system both ways and follows the people. The distinction is carried here because it describes the choice accurately and because the evidence on each side of it is strong, not because the comparison has been run. Treat it as a way of asking a better question, and not as a finding with a number attached.
Zuboff's own work is a 1988 field study of a specific technological moment, computers arriving in industrial and clerical settings. The transfer to generative AI is an argument this estate is making, and she is not responsible for it.
The question to ask before a deployment#
For any system about to go in, ask what the person will understand afterwards that they do not understand now. If the answer is nothing, the deployment is automating, which may well be correct for that task and should be a decision rather than an accident. If the answer is something specific, the deployment is informating and is worth more than its time saving suggests.
That question is the operational form of the choice set out at drift versus design, and the place to record the answer is a delegation boundary map.
Key sources
- Zuboff, S. (1988). In the Age of the Smart Machine: The Future of Work and Power. Basic Books. The origin of the distinction. Named here as the source of the terms and not used as evidence for any claim on this page.
- Bainbridge, L. (1983). Ironies of Automation. Automatica, 19(6), 775-779. Graded entry.
- Dell'Acqua, F., Ayoubi, C., Lifshitz, H. et al. (2025). The Cybernetic Teammate. NBER Working Paper 33641; Organization Science, June 2026. Graded entry.
Related SuperSkills research#
The choice this names is set out at drift versus design, and the artefact for recording it at the delegation boundary map. On what accumulates when the choice goes unmade, capability debt. On the oversight consequence, the invisible work of oversight. On designing work so people still learn in it, how humans learn with AI.
About this research#
Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. Automating and informating are Shoshana Zuboff's terms from 1988 and are credited to her, not claimed. Drift versus design is his and is the later, looser cousin; naming the better-specified ancestor is more useful than not naming it.
Explainer · SS-2026-217 · Graded against the published rubric
Hirji, R. (2026). What is automating versus informating?. The SuperSkills evidence base, SS-2026-217. https://thesuperskills.com/research/what-is-automating-versus-informating. Last reviewed 11 September 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work