Polarity VenturesGet in touch →
Field Note · Vol. 01 · Apr 2026
01 / ThesisLiving document · Vol. 01

The Migration Thesis

AI strips the repeatable work.
The edges still pay.

The repeatable middle gets compressed. What stays human: judgment when situations are unclear, the trust that holds client relationships, and the stitch across multiple systems. Organizations that move up to those edges win. The rest get compressed.

ServiceNow’s CX Shift study (34,000 voices, 2026) confirms the gap: 50% of customers cite lack of empathy as their top frustration. Only 16% of organizations report meaningful AI progress on emotional connection. The data says what the thesis says.

Every field I touch is splitting into two piles. One pile gets automated to near-zero in eighteen months. The other pile gets more valuable because the work that surrounded it just disappeared.

This is not a prediction document. It is a lens. Updated as the work surfaces new signal across SN, AI tooling, and the operator stack.

Last update · 2026-04-30

02 / FrameworkFive tiers of disruption
T0

Tier 0

Untouched (for now)

Work that requires physical presence, trust bonds, regulated human-in-loop, or irreducible taste. Plumbing, elder care, surgery, the moment a therapist reads a face. Not permanent. When the law changes or the receiver adapts, these move to Tier 1.

T1

Tier 1

Reshaped (role survives, shape changes)

AI eats the part of the job that was well-defined enough to specify. The lower half automates. The surviving half moves up: policy design, audit boundaries, drift remediation. The accountant keeps advisory. The SN consultant keeps governance architecture. The marketer keeps narrative.

T2

Tier 2

Displaced (the work goes away)

The task stops being done by humans at all. Not reshaped. Gone. L1 service desk, junior SN configuration, basic SDR sequences, boilerplate legal drafting, first-draft copy. The question isn't whether the task disappears, it's whether the staircase to the next role still exists.

T3

Tier 3

Emergent (did not exist at scale in 2024)

Categories appearing now, most without formal titles yet. Agent trainers. Memory architects. AI auditors. Agent orchestration engineers. Personal AI OS builders. Governance-as-code authors. AI-native SI implementers. Trust-broker ops.

T4

Tier 4

Compounding (more valuable as AI spreads)

Work that gets MORE valuable because AI cuts everything around it. Taste and editorial judgment. Deep domain expertise in regulated fields. Cross-platform orchestration. Narrative and positioning. Relationship capital. Systems thinking across business and tech. Founder-level leverage.

03 / Stay relevantFor any role

The exercise

Map your work
to the tiers.

Decompose your role into tasks. Map each task to a tier. Count how much of your current income comes from Tier 2 work. That is your migration debt.

Deprecate

Any skill a subscription tool already does better than you. Work you can specify clearly enough to hand off. Titles that exist to manage displaced work.

Double down

Judgment across ambiguous tradeoffs. Specific domain depth in regulated or high-consequence fields. Narrative and positioning. Relationship capital. Systems thinking across separate platforms.

Build fresh

Proof of taste in public. A personal AI OS or equivalent leverage stack. A public thesis you update.

The thesis is a living document

Want to talk through where you sit?

Tell me what you are working on and we will map it to the tiers together.