Primitive 01
Unit of Potential (UoP)
A single, governed AI deployment decision — outcome-anchored, constraint-aware, scored. Each UoP carries an Alignment Score and a maturity stage: Generated → Configured → In Production.
Read the spec →01 · What r.Potential is
The alignment layer between AI capability and enterprise work.
We help executives decide what AI should do, where, for whom — and measure whether it actually worked.
Not a marketing site. An operator’s tool. Take five minutes — we’ll walk you through how to read it.
02 · Why this exists
We slice the world by work, not by industry.
Most AI vendors slice the world by industry. We slice it by work archetype — because automation maps to work, not to logos. The same UoP for legal drafting ships across financial services, insurance, and government — and the alignment math is the same.
The AI industry assumes work = tasks. Wrong. Work = coordination. Trying to match tasks and automate them is most of the time useless. What matters is alignment between stakeholders — only then are tasks derived. We see companies as complex adaptive systems, not task lists.
Our opinionated stance
Industry is downstream of work. Vendors who can’t articulate the work archetype they’re solving for are selling you a logo, not a deployment.
03 · The four primitives
Four primitives. Then the rest reads itself.
Everything else in the platform is a surface for one of these four. If you know these, you can operate it.
Primitive 02
Workbench
The operator console. Where to deploy? · Who's ready? · Vendor Allocation · UoP·Live · Cockpit. One surface for the whole AI-deployment decision loop.
Open the Workbench →Primitive 03
Vendor Allocation
The honest multi-vendor shelf, per archetype. OpenAI is one of N. Anthropic leads SWE code-gen at 54%. We tell the truth — the allocation is the work, not the badge.
See the allocation →Primitive 04
UoP · Live
The proof case. TAG × Harvey legal drafting — $30M realized, 1,240 roles, 0 displaced. One UoP in production today, measured against its business case.
See it live →04 · The 4-tier confidence ladder
Every score wears its confidence on its sleeve.
We don’t hide what we don’t know. Every UoP, every cell, every recommendation carries one of four tiers — visible to anyone reading it.
- 1 · EstimatedScore capped ≤ 60%
External data only. Lightcast postings + O*NET DWAs + Eloundou α/β/γ. No org context yet.
e.g. Apple SWE: Estimated 58/100. The ceiling holds because we haven't talked to anyone there. - 2 · ModeledOrg context added
Enriched with firmographics, geography, revenue band, sector-specific absorption signals.
e.g. JPMorgan Compliance Drafting: Modeled 71/100 — financial-services regulatory weight folded in. - 3 · ConfiguredStakeholder-validated
Sponsor identified, Procensus consensus run (L6), governance + redeployment plan in place.
e.g. Adecco × Workday HR Triage: Configured 79/100 — sponsor + HR + IT + legal aligned. - 4 · MeasuredLive telemetry
In production. Outcomes measured against the business case. Realized Potential vs. Estimated Potential reported.
e.g. TAG × Harvey: Measured 78/100 · $30M realized · 1,240 roles · 0 displaced · 4 quarters of telemetry.
Today: 3,315 of our 3,316 scored cells are Estimated. One — TAG × Harvey — is Measured. The progression up the ladder is the product.
05 · Emergence co-creation
The secret of the Alignment Score: it’s not a prediction.
The Alignment Score is a proxy for whether the conditions are right for humans + AI to co-create value. When the score is high, three conditions are met:
Value Clarity
Everyone — sponsor, operator, worker — agrees what good looks like. The outcome metric is named, owned, and measurable.
Agency
Humans + AI both have meaningful contribution. Workers shape the deployment, not just absorb it. AI does what AI does best; humans do what humans do best.
Substance
Real outcomes, not deck theater. The UoP ships value the business actually feels — measured, governed, surfaced in the ledger.
Realized Potential exceeds Estimated Potential when emergence happens. We don’t predict — we set the table.
That’s why the primitive is called Unit of Potential, not Unit of Plan. Potential = capacity for value to emerge when conditions are set right. The plan is the floor; emergence is the ceiling.
06 · Redeployment plan
Every UoP carries a named redeployment plan.
Dignity-with-productivity, not zero-displacement absolutism. The workforce side of every UoP is a first-class artifact — named pathways for every role affected, surfaced before the score lands.
The Worker POV is built into every score. If we can’t name what happens to the people, we don’t ship the UoP.
Worked example · Measured
TAG × Harvey · Legal Drafting · 1,240 roles · 0 displaced
| Role | Affected | Path | Outcome |
|---|---|---|---|
| Junior associate | 412 | Reskilled toward client-facing legal ops | Augment |
| Paralegal | 318 | Reskilled toward AI-supervised drafting review | Augment |
| Document specialist | 187 | Reskilled toward case-management orchestration | Reskill |
| Contract reviewer | 162 | Augmented — throughput +3.4× with human-in-the-loop | Augment |
| Legal admin | 98 | Retained — coordination layer untouched | Retain |
| Knowledge manager | 63 | Expanded — AI corpus curation becomes core scope | Expand |
0 displaced · 1,240 redeployed · measured across 4 quarters · governed under TAG-LGL-001
07 · The journey ahead
Five steps. Five minutes.
The Workbench answers five questions in order. Each step is a surface — click any to jump in.
- 1Where to deploy?Cold roadmap (14,200 enterprises) → Hot roadmap (2,180 targets) → today's curated 30. Revenue-band proxy, sector view, dignity-baked enterprise drawer.Demand Map · Workbench surface 01
- 2Who's ready?Of the visible 30, who's ready for which UoP. 30 × 7 = 210 scored cells. Each carries a confidence tier and a reason chip if it's gated.Absorption Heatmap · Workbench surface 02
- 3Which vendor?Honest multi-vendor allocation per archetype. Per-archetype zoomed-out view. Per-enterprise deep-dive. 42-slice grid below.Vendor Allocation · Workbench surface 03
- 4One in productionTAG × Harvey · $30M realized · 1,240 roles · 0 displaced. The proof case — outcomes measured against business case across 4 quarters.UoP · Live · Workbench surface 04
- 5The cockpitOperator + partner real-time view. Funnel header (Generated · Configured · In Production · Gated · Blocked) + market-signal pipeline.UoP Cockpit · Workbench surface 05
08 · Provenance
We tell you where every number comes from.
No marketing badges. Every source feeding the Global Labor Graph, with row counts and refresh cadence.
| Source | What we use | Coverage | Refresh |
|---|---|---|---|
| Lightcast | Postings · Profiles · Skills · Firmographics | 1.28B postings (16-yr) · 594M profiles · 28,795 skills · 7.3M companies | Daily |
| O*NET | DWAs · Tasks · KSA ontology | 19,265 DWAs · 1,016 occupations | Per release |
| BLS / OEWS | US workforce + wage data (CPS · OEWS p10–p90) | 396 metros · all SOC codes | Per release |
| ILOSTAT | International workforce employment | 172 countries with employment data | Per release |
| OECD | Sector employment + economic indicators | 38 OECD members | Annual |
| Eloundou et al. 2023 | GPT exposure α / β / γ per occupation | 100% of O*NET occupations | Static (research) |
| WEF Future of Jobs 2025 | Employer-reported skills trajectory | 1,000+ CHROs surveyed | Annual |
| TAG (proprietary depth) | HR telemetry · candidate interactions · placements | 100K enterprise clients · 104M placements/yr · 300M candidate interactions/yr | Continuous |
Today: 530 enterprises live · Target cohort: 14,200 (Q3 2026)Full data catalog →How sources feed the methodology →
09 · Step in
Ready? The Workbench is the operator’s tool.
We work with hyperscalers, AI labs, workforce platforms, and the enterprises deploying both. If this is useful — for your team, your customers, or your platform — talk to us.