The Sarah question: what we say to the person being automated out of a job
A hundred demos in, here’s the answer I give now.
Every demo, eventually, someone asks it. Not always with that name. Usually it’s phrased more carefully. “What about my team.” Or “What happens to the person in this role?” Or sometimes just a long pause followed by an uncomfortable question that starts with: “So…”
But it’s always the same question underneath. Sarah is a composite. She’s a PA who’s been managing meeting coordination for six years. She’s an account coordinator who spent the last three years thinking this role was the path to account management. She’s a research analyst who’s good at finding signals in data and who’s watched that signal-finding task get automated every quarter. She’s whoever is doing work that Orca now does.
The question, carefully phrased, is: what happens to her?
I’ve been asked this about a hundred times. I’ve given different versions of the answer each time, because I was figuring it out. This is the version I give now, because I think it’s honest.
The first part: what doesn’t work
Let me start with what definitely doesn’t work, because I see it get suggested all the time and it’s important to be clear about why.
“Retrain into AI skills.” This is the response that everyone seems to reach for. She’ll learn prompt engineering. She’ll become an AI specialist. She’ll level up into a completely different career.
Here’s what happens in practice. Sarah has five years of pattern recognition from her coordination role. She knows how to catch the thing that’s slightly wrong in a dataset. She knows how to sense that a client is not saying what they mean. She knows the client environment and the political dynamics and why the CRM is organised the way it is. Those patterns don’t transfer to prompt engineering.
Prompt engineering at a valuable level is a completely different cognitive architecture from coordination work. It’s technical, iterative, requires real experimentation. By the time Sarah finishes an eight-week course, the prompt engineering itself is being automated. The tools are getting better at figuring out their own prompts faster than humans can learn to write them.
Sarah doesn’t need to become a prompt engineer. What might need more people is knowing what problems are worth solving with AI, checking whether solutions actually work, making sure output’s being checked properly. That’s the meta-layer above prompt engineering. But Sarah gets there through doing something adjacent to her current work, at a different scale. Not through training.
The second part: what Sarah already understands
Sarah understands how work flows between systems and people. She’s built relationships with clients or her team. She’s developed judgment about what matters. She’s got three to six years of track record, which says: this person can stick with something difficult.
None of this means her old job continues. But it’s raw material for something else.
The third part: the positioning work
Here’s where Sarah’s choice actually matters.
She’s got roughly three to six months before the tool is fully rolled out in her company and the coordination work isn’t really a role anymore. In that window, she can do something or nothing.
Nothing: she waits, she hopes the automation doesn’t happen as fast as expected, she sends out a CV to ten companies doing the same coordination work at scale, she hopes the next company hasn’t automated this yet. Statistically, this is what most people do. Statistically, it doesn’t work out well.
Something: she stops waiting and starts thinking about what problem her judgment and her relationship networks actually solve that doesn’t depend on her doing coordination work.
For someone with Sarah’s background, this usually looks like one of three things. Not retraining. Just repositioning.
First, the client relationship angle. Sarah knows clients and what they actually want. Companies often need someone to hold that client relationship at a different level than transaction management. Account director, new business development, or internal operations role that understands client needs. It’s different from her current role but uses the fact that she knows clients and they know her.
Second, the operations angle. As AI handles routine workflows, someone has to ensure the infrastructure’s right. Is the tool being used correctly? Is the output being checked? Is there a human in the loop where needed? These aren’t flashy roles, but they’re durable. Sarah’s five years understanding workflow is a real advantage.
Third, the something-adjacent angle. What does her client actually need that she delivered as part of coordination work, but that isn’t coordination? Strategy, audit, analysis, or just continuity because the client trusts her. There’s usually something in the existing work that was always interesting but got obscured by coordination overhead.
None of these are guaranteed. Some Sarah’s won’t find a slot because there isn’t one in their company. But the point is: Sarah’s choice determines whether she’s in the 30 per cent who get landed somewhere else, or the 70 per cent who get displaced and never find their footing again.
The fourth part: what the company has to do
The company’s choice is stark. Automate and keep headcount unchanged (Sarah eventually leaves), or automate and restructure (find her a new role, or have an honest conversation about it).
The honest conversation is the one that doesn’t usually happen. “The coordination work is being automated. That’s not something we can stop. But we need someone in the client relationship role, or the operations role, or the analysis role. Do you want that job, or do you want to look elsewhere?”
This doesn’t pretend automation isn’t happening. It doesn’t offer a fake job. It acknowledges change and offers a choice.
Some Sarahs will say yes. Some will say no. Both are better than pretending the old role exists. Companies that handle this well survive with culture intact. The ones pretending the old structure exists while Sarah does something new without support are where culture breaks.
The fifth part: the realism about timing
Here’s the timeline as I see it.
Sarah’s got about six months. Maybe a year if she’s lucky and her company’s slow to move. In that window, the tool is being deployed, the efficiency is becoming obvious, the leadership is doing the maths about headcount.
If Sarah’s done her positioning work, there’s a decent chance she lands somewhere. Not the job she expected. Not the trajectory she planned for three years ago. But a job that uses what she knows and pays something like what she’s earning now.
If she hasn’t done that work, the window closes and she’s looking for work in a contracting market. She’s looking for jobs in coordination at companies that haven’t automated yet, which is rapidly becoming a list that only includes companies in serious trouble. She’s competing with people who have six months more experience than her in roles that are being squeezed right now.
This isn’t a lecture about personal responsibility. This is describing the structural reality Sarah’s facing. Some of it is in her control. Some of it isn’t. The companies that handle this reasonably well are multiplying slowly. Some of it is chance.
But doing nothing, waiting, hoping the change doesn’t happen as fast as everyone’s predicting, that’s the one choice that’s almost certainly worse.
The sixth part: what I actually tell Sarah
When Sarah asks me the question directly (which happens, sometimes), here’s what I say:
Your job as it existed is being automated. That’s not my prediction or my preference. That’s what the tools do. Your company will probably automate this. If they’re smart, they’ll restructure so you have somewhere to go. If they’re not smart, they’ll pretend the automation doesn’t require any structure change and you’ll eventually leave frustrated.
Your best move right now is to figure out what about your current role you actually want to keep doing, and go talk to your company about whether there’s a place for that. Or start looking for a place that needs what you know.
Don’t retrain. Don’t become a prompt engineer. Don’t assume the role will come back. Don’t wait for someone to tell you what to do.
Figure out what problem your judgment and your relationships solve. Find a company that has that problem. Go solve it.
That’s the answer I give. Not comfortable. Not “don’t worry, everything will be fine.” Not “learn new skills and you’ll be fine.” Just: here’s what’s actually happening, here’s what you can control, here’s what you should do about it.