Our daughter Katie told us that planning a career feels pointless when AI might change the job before she has built a life around it.
I wanted to give her the optimistic founder answer. I could not honestly do it.
Katie works on reception at a GP surgery. Her day includes anxious patients, phone calls, documents, appointments and the messy judgement needed when a neat process meets a real person.
Helen and I run a business shaped around AI. Katie has watched us rebuild it. That makes the conversation harder, not easier.
I made the fear worse with fake precision
The first version of this article said that 99 per cent of Katie's work would be gone within two to four years.
I cannot support that number or timetable. It treated a father's fear as a forecast.
Research does show that clerical occupations are among those most exposed to generative AI. The International Labour Organization's task-level study also makes an important distinction: exposure means tasks may change. It does not mean every exposed job disappears.
The ILO's 2025 task-level study concluded that transformation is more likely overall than replacement because most occupations still contain work requiring human input.
That is less dramatic than "the job is gone". It is also what the evidence says.
A job is not one task
Some reception work is clearly easier to automate than other parts:
- finding and summarising information;
- classifying routine messages;
- collecting standard details;
- routing requests;
- preparing a response for approval.
The harder work sits around exceptions, distressed callers, incomplete information, safeguarding, responsibility and local knowledge.
Technology may change the balance between those tasks. Employers may redesign roles, reduce some work, add oversight or simply expect more from the same team. Nobody can honestly tell Katie which combination her employer will choose several years from now.
What can a parent say?
"Do not worry" is useless.
"Become an AI expert" is not much better. It turns a broad economic change into another homework assignment for the person already worried about it.
The conversation I want to have now is smaller:
- Which parts of your day are routine and which depend on trust or judgement?
- What new tools are actually appearing in your workplace?
- Who remains accountable when software gets something wrong?
- Which adjacent skills would give you more options?
- What support would make experimenting feel safe rather than threatening?
Katie may still decide she wants nothing to do with AI. That is her choice.
My job as her dad is not to sell her a cheerful future I cannot prove. It is to help her separate a changing task from a doomed life, keep her options open and admit when I do not know what comes next.