How Shift works — and what we don't claim.

The three pressures we score

For every role, at three time stops (2026 / 2028 / 2030), we score three independent dimensions:

Automation pressure

How much of the role's day-to-day work AI is taking over. The skills you used to bill for that a model now does in seconds.

Augmentation pressure

How much AI is changing how the work gets done — without removing the role. New tools, new workflows, new expectations of speed.

Human resilience

How much of the role stays stubbornly human — taste, trust, presence, judgement, accountability, the bits AI is bad at and people pay for.

These three are not a pie. A role can be high-automation AND high-augmentation AND high-resilience at the same time — that's a job in upheaval whose human core nevertheless holds. We deliberately don't make them sum to anything.

The five bands

Each of the three dimensions, at each year stop, gets one of five bands:

1. LOW Almost no movement on this axis.
2. MEDIUM-LOW Visible, contained.
3. MEDIUM Real change, working its way through.
4. MEDIUM-HIGH Reshaping the role.
5. HIGH Dominant force at this stop.

Bands are intentionally chunky. They reflect that we cannot tell you whether automation pressure on copywriters in 2028 is 62% or 67% — and neither can anyone else.

The three confidence levels

Every individual claim — every "AI is automating X" line — carries one of three confidence pills:

• High The claim is backed by a named source we cite inline (industry report, company announcement, peer-reviewed paper).
• Medium The claim is consistent with multiple secondary sources but we're not citing one specific anchor. Treat it as informed observation.
• Low The claim is our analyst take on where things are heading, not a measurement. We label it so you can weigh it accordingly.

Sush's honest take

About one in five role pages includes a "Sush's honest take" pull-quote. The voice rule is strict: neutral by default, opinion only on conviction. Most roles ship without a take. That's deliberate — silence is more honest than forced commentary.

Where a take is present, it never restates a numeric claim. It tells you something the instrument panel can't, in plain English, with no fluff and no fake confidence.

What Shift does NOT claim

• We don't predict who keeps their job. Roles aren't people. A role being under heavy automation pressure doesn't mean every person doing it is replaceable — usually the opposite.
• We don't tell you whether to switch careers. The "adjacent dispatches" section shows where the bridges are if you want one. It is not advice.
• We don't have proprietary unemployment numbers, salary projections, or headcount forecasts. If a site is showing those, it's pretending — including some big names.
• We don't claim our 2030 numbers are right. We claim they are useful for thinking. That's the bar.
• We don't sell your data. Shift has no signup, no tracking pixels, no third-party scripts. Just static pages.

Refresh cadence

Each role carries a last_verified date. The intent is monthly review; the cadence we hit in practice is whatever's honest. If a role hasn't been reviewed in over six months, take its claims with extra salt — and tell us, because we'd rather know.

How to disagree with us

Shift is built by one analyst with strong views and limited time. Some of our calls will be wrong. If you read a role page and think we've got something materially off — a skill we listed as automating that isn't, a force we missed, a citation that's stale — please write to A Guide to Cloud. We update publicly, with the date and the reason.