◭
Expansion-signal detector
Customer successScans a customer's product usage, support tickets, and call notes for buying signals that justify an expansion play (new framework, new region/entity, subprocessor sprawl, seat growth, trust demand) and emits a typed signal with the exact upsell and the line to open the conversation.
The routing recipe
Job typeComplex reasoningPolicyBalanced
routes to
NVIDIA Nemotron 3 Ultra 550B-A55B nvidia
nvidia:nvidia/nemotron-3-ultra-550b-a55b
90.8
match score
96.5
task fit
100
cost
59.5
speed
Why this model from benchmarks & capabilities
- "Complex reasoning" leans hardest on reasoning; NVIDIA Nemotron 3 Ultra 550B-A55B scores 100/100 there — top-1 of 40 chat models.
- Strength on this class of work shows up on GPQA Diamond, MMLU-Pro, BIG-Bench Hard, where NVIDIA models are competitive.
- Cost: runs on NVIDIA free credits / self-hostable NIM — effectively $0 marginal. Speed: ~70 tok/s (estimated).
- Chosen over NVIDIA Nemotron 3 Super 120B-A12B for higher task-fit (96.5 vs 90.7) & more context.
| Capability the job needs | Weight | NVIDIA Nemotron 3 Ultra 550B-A55B | Catalog rank |
|---|---|---|---|
| reasoning | 1.00 | #1 of 40 | |
| math | 0.60 | #3 of 40 | |
| knowledge | 0.50 | #4 of 40 |
Relevant benchmarksGPQA Diamond MMLU-Pro BIG-Bench Hard LiveBench (reasoning)
EconomicsCost: runs on NVIDIA free credits / self-hostable NIM — effectively $0 marginal.
Speed: ~70 tok/s (estimated). vs runner-upChosen over NVIDIA Nemotron 3 Super 120B-A12B for higher task-fit (96.5 vs 90.7) & more context. (NVIDIA Nemotron 3 Super 120B-A12B, score 89.5)
Speed: ~70 tok/s (estimated). vs runner-upChosen over NVIDIA Nemotron 3 Super 120B-A12B for higher task-fit (96.5 vs 90.7) & more context. (NVIDIA Nemotron 3 Super 120B-A12B, score 89.5)
Example result
Sample input ▾
Account: NorthSaaS (Growth tier, SOC 2 Type II live since Jan). Last 30 days — Product events: created 2 new 'evidence scopes' tagged "EU-prod" and "frankfurt-rds"; invited 4 users with title "Privacy/DPO". Tickets: "Does iCompaas map controls to ISO 27001 Annex A?" (x2), "Can we generate a GDPR Art.30 record of processing?". Calls: champion said on QBR "we just signed our first two German logos and legal wants an AI-governance story for the EU AI Act by Q4." Billing: 18/20 seats used.
Signal: NEW_REGION_ENTITY · Confidence: High
Evidence: New "EU-prod"/"frankfurt-rds" evidence scopes + 4 DPO invites, plus QBR quote — "first two German logos… EU AI Act story by Q4."
Secondary (flagged): NEW_FRAMEWORK — two ISO 27001 Annex A mapping tickets show an ISO add-on warming (~60%+ controls already shared with their live SOC 2).
Recommended upsell: EU module bundle — GDPR (Art. 30 RoPA + DPA workflows) + EU AI Act / AI-Governance pack, scoped to the Frankfurt boundary so they reuse existing SOC 2 evidence instead of standing up a second audit. Pre-stage the ISO 27001 cross-map as the follow-on.
Seat note: 18/20 seats used; the 4 DPO hires tip them over — attach a seat true-up to the same order.
Opening line: "Saw you're standing up Frankfurt and legal wants an EU AI Act story by Q4 — want me to show how we'd reuse your live SOC 2 evidence to stand up GDPR + AI-governance in one boundary, no second audit?"
Test it on your own data
· confidence: \nevidence: \nsecondary (flagged): \nrecommended upsell: \nseat note: \nopening line: \n\nrules: never invent signals not present in the input. if primary is none, output only the signal line plus a one-line reason and stop. be terse. tie every recommendation to evidence in the input — no generic plays.","task":"reasoning","optimize":"balanced"}'>