◆ Supervised Specialist
Builds the valuation that anchors the equity story … trading comps, DCF and sum-of-the-parts … and runs the discount/premium scenarios the syndicate-desk agent needs for pricing. Frames the range; the syndicate-desk agent sets where in it to land.
Memory
Working The valuation build and the live pricing scenario grid.
Episodic Prior ECM valuations in the sector and how they priced vs. the range.
Semantic IPO-discount conventions, sector multiples, float dynamics.
Procedural Valuation-build playbooks per offering type.
Store File-based memory tool + market-data warehouse
Orchestration
MCPA2A
Harness · Managed Agents … modelling session with sandboxed computation; assumptions persisted in a NOTES file across the deal timeline.
Tools
⌘ Market data (CapIQ / FactSet) MCP ›_ Valuation sandbox Code exec ⇄ Comps set A2A ⇄ Syndicate-desk-agent review channel A2A
Evals & guardrails
- Range reconciliation across comps / DCF / SOTP; divergence must be explained.
- Pricing scenarios sanity-checked against recent sector IPO outcomes.
- The syndicate-desk agent re-derives and owns the price; this agent frames the range.
Offline reflection
Replays where deals actually priced within the proposed range to calibrate its IPO-discount assumptions per sector and market regime.
Frontier edge
- ▲World-model simulation: simulates investor response across the price range, modelling demand build and aftermarket trade before the desk sets guidance.
- ▲Causal reasoning: counterfactual analysis of how float size and lock-up structure move the achievable discount, not just historical correlation.
- ▲Continual learning: eval-gated calibration of IPO-discount assumptions per market regime from where deals actually printed (SEAL-style).
In numbers
60
Valuation builds / quarter
median ~4 h
Range-build latency
Handoffs
Fed by ← Comparable Companies Agent
Hands to → Bookbuild & Allocation Agent