Professional Judgement in the Age of Generative AI
Generative AI can produce an answer in seconds. Professional judgement is slower, more situated and harder to describe.
It appears when the brief is incomplete, the evidence is ambiguous, different stakeholders need different things, or the consequences of getting something wrong are not evenly shared. It shapes what people notice, what they question, what they accept and what they decide to leave alone.
I am leading a founder-led research experiment to understand how that judgement is changing as generative AI becomes part of professional work.
Corporate communications is the first research domain.
Communications offers a rich setting for studying judgement because it brings together uncertainty, competing priorities, stakeholder consequences and accountability. This study is limited to communications; it does not claim evidence about other professions.
Over the past twenty-five years, I’ve noticed something. Most communications problems aren’t really communications problems.
They’re decision problems.
Someone hasn’t been clear about what they’re trying to achieve. Leaders aren’t aligned. The organisation isn’t ready. The behaviour people want to change hasn’t been defined. Sometimes communication isn’t even the right answer.
By the time Communications gets involved, we’re often trying to make sense of decisions that should have been challenged much earlier.
That’s the question that’s fascinated me throughout my career.
And now AI has made it even more interesting.
My research and writing explore the intersection of generative AI and workplace communication, with a particular focus on professional judgement: what AI can contribute, what still requires human interpretation, and how organisations can adopt these tools without losing the experience and judgement on which good communication depends.
Taking Part
The research is intended for people involved in communications, organisational change, transformation or related leadership work.
You do not need to use AI or have specialist knowledge of it.
The interview contains 16 fixed questions and usually takes around 15–25 minutes. Depending on what you say, it may ask up to five brief follow-up questions drawn from an approved set, with no more than two in any section.
Questions appear one at a time. You can speak or type your answers, and both are equally valuable. You control when recording starts and stops, and you can pause for as long as you need.
Participation is voluntary. The interview is research, not an assessment, product demonstration or sales conversation.
This communications study follows the published methodology; before you begin, please read the Participant Information.
Why communications comes first
Corporate communications is a useful first domain because important work often begins with incomplete information, competing priorities and consequences for people and organisations. It makes the quality of professional judgement visible before, during and after writing.
We’ve spent the last couple of years talking about how AI can help us write. It can draft announcements, rewrite emails, summarise meetings and generate content in seconds.
That’s impressive.
But I’m much more interested in something else.
Can AI help us think better before anyone starts writing?
Can it help us ask better questions?
Challenge weak assumptions?
Spot gaps in thinking?
Help communications teams become more strategic?
Or is today’s general-purpose AI already good enough?
I genuinely don’t know. That’s why I’m doing this research.
The current study
The study asks:
How do communications professionals exercise judgement when planning and reviewing important work, and how does generative AI affect that judgement?
It explores real experiences rather than asking only for general attitudes towards AI.
The interview begins with a little about the participant’s work, then asks them to reconstruct a recent situation in which the right way forward was not completely straightforward. It explores uncertainty, competing considerations, AI use or deliberate non-use, human interpretation, quality, accountability and consequences.
AI scepticism, limited access, deliberate constraint, negative experiences and unknown outcomes are all valuable evidence. The research is not trying to prove that AI is beneficial, harmful or inevitable.
Given Time™
The study is being conducted through Given Time™, an AI-assisted qualitative research experiment I conceived and developed.
Given Time™ is designed to create better conditions for thoughtful reflection. It presents one question at a time, keeps the technology in a supporting role and gives participants space to answer in their own words.
The sequence is fixed and versioned. AI cannot invent or rewrite interview questions. It may only decide whether to move on or select a relevant follow-up from a predefined, approved set. Machine-generated summaries and suggested research codes are treated as proposals for human review, not as findings.
This is a design intention, not a claim that the experience has been scientifically shown to improve reflection or produce better evidence. Its quality and usefulness must be evaluated through the research itself.
Become a Research Partner
I’m also looking for a small number of people who’d be willing to contribute more deeply. That might involve an interview, reviewing planning approaches or taking part in future research.
No sales pitch. No expectation that you’ll agree with me. Just honest conversations about how communications is changing.
Where the research is today
This is now an active study. I’ll keep this record updated as the work progresses.
✓ Interview platform live
✓ Protocol v1.1 frozen
✓ Participant Information published
✓ Participant recruitment underway
⬜ Working Paper 1 published
⬜ Interim findings
⬜ Final report
Research Library
Supporting documents and papers are published as the project develops, so the basis of the research can be examined as well as its conclusions.
Working Paper 1
Interim Findings — forthcoming
Final Report — forthcoming
How the research will be used
Responses may be analysed to identify themes, examples, differences and contradictions. AI may suggest summaries and research codes, but a human researcher reviews the material before it is used in findings or publications.
Findings may be published in reports, articles, presentations or other research materials. They will normally be presented in aggregated or anonymised form. Short anonymous quotations may be used after human review to reduce the risk that a person or organisation could be identified.
The research does not produce participant scores, competency assessments, organisational ratings, commercial lead scores or automated decisions about participants.
A Promise
If the research supports my ideas, I’ll publish it. If it doesn’t, I’ll publish that too. The point of this project isn’t to prove a predetermined idea is right. It’s to better understand how AI is changing professional judgement in communications — and whether there’s actually a better way of working.
Research Report
If you’d like to receive the findings without taking part and when the research is complete, you can register here.