If AI news has you thinking about your next move, the most useful thing you can do this month isn't another course. It's three conversations.

The reason is simple: most people learn what a field actually looks like from someone inside it, not from articles. And a conversation costs nothing but 25 minutes.

The plan: three conversations in 30 days.

Conversation 1: someone one step ahead of you in your own field. Ask: "What part of your week has changed most in the last year?" and "What do you now spend more time on?" You're gathering ground truth about your own line of work.

Conversation 2: someone in an adjacent role. Pick a job that uses your skills in a different setting: an analyst who could move toward product, a writer who could move toward research, an accountant who could move toward advisory. Ask: "What made someone good at this job that surprised you?"

Conversation 3: someone who hires. A manager or recruiter in a field you're curious about. Ask: "When you look at applicants, what tells you someone can work well with the new tools, and what tells you they can't?"

How to ask (keep it short):

"Hi [name], I'm thinking about how my field is changing and you're someone whose work I respect. Could I ask you three questions over a 20-minute call? Happy to work around your schedule."

Rules that make it easy:

• Send the request to people you already have a thread with: former colleagues, alumni, people you've met once.

• Prepare three questions. Send none of them a CV.

• Afterwards, send a two-line thank-you with one thing you'll act on.

• Write up notes the same day: what surprised you, what you'll test.

After three conversations you'll have a better picture of your options than a month of scrolling, and a few people who know you're looking around. That reduces the sense of having no control, which is a big part of why the uncertainty feels so heavy.

Related reading: https://straata.store/articles/should-i-change-careers-because-of-ai

Free 6-page PDF: https://straata.store/The-3am-Protocol.pdf

Published by Nightshift Guides.

Here is a 30-minute exercise that tells you more about your job's future than any headline about AI.

Write the handover document for your own role. Pretend you're going on leave for three months and a capable stranger is covering. What would they need?

Open a blank page and make three headings:

1. Steps anyone could follow. Recurring tasks with a clear input and output: weekly reports, formatting, first drafts, scheduling, data cleanup. Write each as a short checklist. If you can write it as a checklist, it is the part of your job most likely to be shared with software first.

2. Judgement calls. The moments where the stranger would phone you: which client needs a softer tone, which number in the report looks wrong, when to escalate, when a request from a senior person shouldn't be taken literally. Write down how you decide, not just what you decide. This is the hard section. It will take longer than you expect.

3. Relationships and context. Who trusts you, who owes you a favour, what was tried in 2023 and failed, what the boss actually means by "urgent."

Now look at the page. Most people find heading 1 is longer than they'd like and heading 2 is thinner than it should be, because judgement lives in your head and has never been written down.

That gap is your to-do list:

• Take the section 1 items and decide which you'd happily hand to a tool, so your hours go to section 2.

• Take the section 2 items and start writing two or three real examples per month, each with what you saw, what you chose, and why. That doubles as material for your next review or interview.

Keep the document. Update it in six months. It's a concrete record of where your time goes and what only you can do, which is a far better place to plan a career decision from than a vague sense of dread.

Related reading: https://straata.store/articles/you-are-not-your-job-title

Free 6-page PDF: https://straata.store/The-3am-Protocol.pdf

Published by Nightshift Guides.

If your team went through a layoff tied to AI tools and you're one of the people still there, the weeks afterward have their own shape — different from being the one who left, and rarely talked about directly. Here's a practical rundown of the decisions actually in front of you.

Your job description is now wrong, temporarily

The tasks that belonged to the people who left don't disappear; they land somewhere, often without anyone formally deciding where. In the first two weeks, write down what's actually landed on your plate versus what your manager thinks you're doing. That gap is worth a direct, unemotional conversation: "Here's what I've absorbed since the restructuring — can we talk about what stays with me and what gets redistributed?" Waiting for someone to notice on their own rarely works.

Don't volunteer for everything to look secure

A common reaction is taking on visibly more to seem indispensable. This usually backfires twice: it's unsustainable, and it teaches your manager that the new workload is simply normal for your role now, which becomes the baseline they expect going forward. Take on what's genuinely yours to take on, and say so out loud when you're near capacity.

The unease is common, and it isn't useful evidence

Feeling uneasy about being the one who stayed is a normal reaction, and it doesn't actually tell you anything about whether the decision was fair or about your own standing. It's worth noticing the feeling without treating it as a signal to act on — it isn't the same as evidence that your role is next.

Ask the direct question once

One conversation with your manager, framed plainly: "Is there another round coming, and if so, when might you know?" You likely won't get a full answer, but the response — a real answer, evasiveness, or reassurance — is itself information you can act on, which beats guessing from silence.

Update your own numbers regardless of the answer

Whatever you're told, this is a reasonable moment to check your runway and update your resume, independent of whether anything else happens. That's true after every reorg, not just this one.

More on checking that number honestly: https://straata.store/articles/financial-runway-job-loss-anxiety

Free 6-page extract, no card, no email: https://straata.store/The-3am-Protocol.pdf

If this has hit you harder than expected: US — call or text 988. UK — Samaritans 116 123. Elsewhere — findahelpline.com.

Published by Nightshift Guides.

If you freelance or contract, you've probably had a version of this conversation: a client hesitates on your rate, or a renewal, and mentions — sometimes as a real threat, sometimes just testing you — that they could "probably just use AI for this now." Here's a way to handle it that doesn't involve panicking or pretending it isn't a real question.

Separate the two things they might actually mean

"We could use AI instead" usually means one of two different things, and they need different answers. Either: (1) a chunk of what you deliver is now genuinely faster to produce with a tool, so the client has a real cost comparison in mind. Or (2) the client is using "AI" as a generic negotiating word for "we think this should be cheaper," the same way "I got a better quote" gets used. Ask one direct question before responding to either: "Which part of the work were you thinking AI would handle?" Most of the time this question alone reveals which situation you're in, because a vague threat can't name a part.

If it's real: reprice the task, not yourself

If they can name a specific deliverable a tool now handles faster — a first draft, a rough cut, a template — don't defend the whole relationship. Separate your rate into the part that's now faster (charge less for it, or bundle it as a smaller line item) and the part that's still entirely you: the judgment calls, the client-specific context, the catching of what's wrong before it ships. Quote those separately. This usually raises your effective rate on the work that's left, even as the total invoice shrinks.

If it's a negotiating tactic: say what you actually provide

If they can't name the part, say plainly what a tool can't currently do for them: knowing their brand's specific tone, catching the thing that would embarrass them with their own boss, the accountability of a named person they can call when it's wrong. That's not a defensive answer, it's an accurate one, and most clients respect a specific answer more than an apologetic one.

Either way, know your number before the call

Decide, in advance and calmly, the lowest rate you'd actually accept to keep the relationship, and the rate below which you'd rather lose the client. A number decided beforehand beats one negotiated live while you're worried about losing the income.

More on making the parts of your work that don't get replaced visible: https://straata.store/articles/job-crafting-instead-of-leaving

Free 6-page extract, no card, no email: https://straata.store/The-3am-Protocol.pdf

Published by Nightshift Guides.

This question is showing up in interviews and performance reviews now, and it catches a lot of people off guard for the same reason: they haven't decided what they actually think, so they either overclaim ("I use it for everything") or underclaim ("I don't really, I like doing it myself") on the spot. Neither answer serves you. Here's a way to prepare one that does.

Do the audit before you need the answer

You can't give a good answer to this cold. Spend ten minutes first: list five to eight things you did in the last month, and for each, note whether a tool helped, and what you still had to do yourself. This is the same production/judgement split from the task audit — where a tool did the first pass, and where you did the checking, deciding, or fixing.

Structure the answer in three parts

1. Where it actually helps you. Name one or two concrete, specific uses — not "I use ChatGPT," but "I use it to get a rough structure for a report started, then I rebuild the parts that need our specific context." Specific beats impressive.

2. Where you don't rely on it, and why. This is the part people skip, and it's the part that shows judgement. "I don't use it for the client-facing sections because that's where getting the tone wrong costs us the relationship" tells the interviewer or your manager exactly where you think the line is — and that you've thought about it, rather than either avoiding the tools or handing everything to them.

3. What you check, every time. One sentence on how you verify or edit what a tool gives you. This is the sentence that tells them you're the person who catches the error, not the person who ships it.

Why this version is actually true, for most people

If you've done any real task audit, you already have this answer — it's not spin, it's a description of a genuinely mixed practice. The three-part structure just organizes what's already true so it comes out clearly instead of as a nervous over- or under-statement.

In an interview specifically

Interviewers asking this are usually checking for one of two things: that you're not behind on the tools, or that you're not about to hand them unchecked output. The three-part answer covers both without you needing to guess which one they're worried about.

In a review

This is also a natural moment to ask the question back: "what would you want me spending the time I save on?" That's the same question worth asking after any tool adoption, and a review is a good place to ask it directly.

More on the underlying task audit: https://straata.store/articles/job-crafting-instead-of-leaving

Free 6-page extract, no card, no email: https://straata.store/The-3am-Protocol.pdf

Published by Nightshift Guides.

A common trap once you decide to "get better at AI": a new model or tool launches every few weeks, and it starts to feel like you're always one course behind. That treadmill is exhausting and it's optional. Here's a five-minute way to pick one thing worth your time this month, instead of everything.

Start from your own task audit, not the news

If you've already sorted your weekly tasks into production (output that could be specified and checked) and judgement (deciding what to produce, whether it's right, who needs it), you have the input you need. If you haven't, it's ten minutes: list what you did last week, mark each line P or J.

Rule one: learn tools for your production tasks, not in general

Skip "learn AI" as a goal. Pick the specific production task on your list that eats the most time, and learn the one tool or feature that compresses it. If your biggest time sink is first drafts of routine documents, that's the thing to get fast at this month — not whatever tool is trending on your feed.

Rule two: spend the time you free up on one judgement skill, not a new tool

This is the part people skip. Compressing a production task only helps you if you reinvest the freed time in something judgement-side and visible: getting better at the client conversation, learning the part of the process where you currently need someone else's sign-off, understanding the business context behind your team's decisions. That's the skill that doesn't get built by a tool release, and it's the one your role actually depends on.

Rule three: one skill, one month, then reassess

Pick one item from rule one and one from rule two. Give each a month. At the end, ask: did this change what I can do, or just what I've read about? If it's the second, drop it and pick something else. Most "AI upskilling" that doesn't stick was chosen because it was popular, not because it matched an actual task on your list.

What this isn't

It isn't an argument against curiosity. If you want to spend an afternoon poking at something new because it's interesting, that's a fine use of a Saturday. It's a different thing from believing you're falling behind because you haven't read about the tool that launched Tuesday. You're not behind. You're choosing on purpose instead of by headline.

More on separating what's actually changing in your role from what just feels urgent: https://straata.store/articles/job-crafting-instead-of-leaving

Free 6-page extract, no card, no email: https://straata.store/The-3am-Protocol.pdf

Published by Nightshift Guides.