Evidence & ideas / Research archive
Alex's Imperial MBA research on junior development and professional judgement. Early findings, shared in October 2026.
The question
The study explores UK mid-market professional services and creative agencies. Junior work has traditionally produced client output and developed judgement at the same time. Automation separates those two things.
Early findings
Contributors and firms remain anonymous. These are practitioner accounts, not proof of what happens in every firm.
AI can remove the work that gave juniors a feel for what good looks like.
Repeated exposure helps people spot patterns, recognise differences and learn from mistakes. The answer isn't to keep every old task - it's to replace the learning that disappears when useful practice is automated. This isn't inevitable: some firms report no clear loss of junior exposure.
A polished answer no longer tells you what someone actually understands.
AI can produce convincing work before the person can explain, defend or adapt it. Managers need to ask what someone tried, what they rejected and what would change their mind - not just review the finished document.
Firms have clearer rules for using AI than for learning alongside it.
Rules on data, confidentiality and approved tools are taking shape. What replaces the apprenticeship is less settled. Leaders are experimenting with independent first attempts, AI as a critic, closer client exposure and increasing responsibility - not one established model.
Time saved isn't automatically time spent developing people.
Two pressures keep coming up: billable work and coaching capacity. Freed-up hours can become more client work, while senior people need both the time and the ability to coach. Development has to be planned and resourced, not squeezed into whatever is left.
What's been said
“Some of the boring junior work is incredibly valuable.”
“You can give someone a junior-level task and AI can make them look incredibly capable.”
“We have a phrase internally: proximity is development.”
“If I save my junior ten hours a week but use those ten hours to give them ten more tasks, I've increased productivity. I haven't developed them.”
A proposed framework
Four decisions for leaders, grounded in interview accounts. The approach has not been tested as an intervention.
The condition: protect senior time.
Experienced people need time to explain, question and debrief. If every hour saved becomes more delivery, learning still loses out.
Move 01
Before automating, ask what capability the task develops. Keep or redesign the learning, not the busywork.
Move 02
Let juniors form a view first, then use AI to challenge or extend it. Match the support to their experience.
Move 03
Ask why they reached the recommendation, what they rejected and what would change their mind.
Move 04
Make room for client meetings, observation and debriefs, then increase what juniors are responsible for.
Limitations: this work reflects leaders' and practitioners' perspectives, not juniors' own experiences. It does not establish that AI is reducing junior hiring or that the proposed approach improves outcomes. Research with juniors and longer-term testing are still needed.
Discuss the evidenceIn your inbox from January 2027
Three developments. One idea. One practical takeaway.
The Positions Briefing