Methodology & model scope
AICompass is a multi-module research stack: (1) procurement fit with heterogeneous treatment adjustment on vendor ROI; (2) CVC-style equity diagnostics; (3) structural portfolio decomposition (Σ_obs, Σ_do, confounded mass); (4) M&A with explicit DAG / interventional readouts where identified. CPA and M&A embed causal structure by construction; fit and investment layers combine MCDA with causal regularization — not single-equation OLS storytelling.
Four workstreams: mandate & deliverables
1. Strategic fit (procurement / MCDA)
Mandate:
Operating procurement of an AI capability. Weighted MCDA across strategy, technology, and supervision; ROI claims pass through a confounding / HTE adjustment layer before capital approval.
Deliverables:
Pillar scores, causal ROI lattice, risk register, formal recommendation string, board-ready narrative, optional live-market overlay.
Open fit module →2. CVC / equity diligence
Mandate:
Minority or strategic equity sleeve — distinct from procurement. Composite scoring (fit × economics × risk) plus an explicit venture-signal vector; not a substitute for securities law diligence.
Deliverables:
Equity read, scenario-conditioned returns, payback surface, sponsor-demand heuristic, confounder disclosure.
Open equity module →3. Portfolio analytics (CPA / SCM)
Mandate:
Attribute covariance to shared latent factors vs. idiosyncratic shocks; contrast MV portfolios on Σ_obs versus interventional Σ_do for regime-shift diagnostics.
Deliverables:
Volatility spread, φ̄ (average confounded fraction), weight vectors, visualization suite.
Open CPA module →4. M&A transaction intelligence
Mandate:
Full-diligence stack for depositories / CUSOs with DAG-backed interventional estimates on post-close efficiency and ROE where the identifying assumptions hold.
Deliverables:
Recommendation, signal index, modular financial / risk / synergy / member-impact workbooks, value-driver attribution, do-style scenario table.
Open M&A workbench →Fit Analysis Modules (Product Fitment)
The Product Fitment Analysis combines your institution profile with vendor data to produce: Strategic Alignment, Technical Fit, Regulatory Readiness, Risk Assessment, and a Fit Recommendation. Each module produces scores and explanations tailored to your institution's size, systems, and maturity.
1. Strategic Alignment
How well the vendor matches your strategic priorities
How it's calculated
The overall score combines four sub-scores with these weights:
- Priority Match (30%) — Does the vendor's category align with your primary strategic priority? (e.g., Member Experience + Conversational AI = high match)
- Member Relevance (30%) — Based on your digital adoption rate, average member age, and product mix
- Timing Appropriateness (20%) — Your AI maturity and implementation experience vs. vendor complexity
- Board Readiness (20%) — Board AI literacy and governance policy vs. vendor transparency
Score interpretation
70+ = Strong alignment; 50–69 = Partial; <50 = Misalignment or timing concerns
2. Technical Fit
Can you technically deploy this vendor?
How it's calculated
- Core Integration (35%) — Does the vendor support your core banking system? Production = 90, Beta = 75, Unknown = 40–60
- Infrastructure Readiness (20%) — Cloud posture vs. vendor deployment model
- Data Readiness (25%) — Your data quality self-assessment and API gateway
- Team Capacity (20%) — IT staff count vs. typical implementation weeks
Risk levels
Core integration can be Low, Medium, High, or Blocker (no documented compatibility).
3. Regulatory Readiness
Are you and the vendor prepared for regulatory scrutiny?
How it's calculated
- Model Risk Alignment (25%) — MRM framework, SR 11-7 alignment for lending AI
- Fair Lending Compliance (20%) — Adverse action support, bias testing (for lending vendors)
- Vendor Compliance Posture (25%) — Vendor's regulatory score, SOC2, NCUA exam readiness
- Examination Readiness (30%) — Your exam rating and governance vs. vendor readiness
4. Adjusted ROI (Causal Adjustment for Confounding)
Multi-criteria fit + causal adjustment of vendor ROI for your institution
Vendors show their best customers. We compare your CU to those cases and adjust the expected benefit up or down based on your size, digital adoption, and data quality.
What we adjust for
- Factors that may change outcomes — Institution size, digital adoption, data quality. Case study FIs may differ from you.
- Our estimate for your institution — The same product may perform differently at your CU than at their typical success story.
- Financial scenarios — Conservative, expected, and optimistic projections with payback and 3-year ROI.
How much we trust the vendor's claims (1–5)
5 = Gold standard (RCT or quasi-experimental); 3 = Pre/post without controls; 1 = Marketing claim only
5. Risk Assessment
What could go wrong? (Lower score = higher risk)
- Vendor risk (25%) — Vendor risk flags (concentration, regulatory history, etc.)
- Implementation risk (25%) — Core integration risk level
- Regulatory risk (20%) — Governance gaps
- Strategic risk (15%) — Vendor lock-in, switching costs
- Member impact risk (15%) — Fair lending, bias (for lending AI)
Mitigations are tagged as Immediate, Pre-Launch, or Ongoing.
6. Decision Recommendation
The final verdict
Recommendation logic
- Do Not Proceed — Core integration is a Blocker
- Strong Proceed — Avg alignment ≥75 and risk inverted ≥70
- Proceed with Conditions — Avg ≥65, risk inverted ≥60 (conditions listed)
- Proceed with Caution — Avg ≥50 (pilot recommended)
- Defer — Avg <50 or risk >70
The recommendation includes conditions, prerequisites, and a suggested implementation approach (Pilot, Phased, etc.).
Portfolio Analysis (CPA): How It Works
Causal Portfolio Analysis — regime robustness, not static optimization
CPA splits observed covariance (Σ_obs) into regime-invariant interventional component (Σ_do) and confounded component. Portfolios built using Σ_do tend to stay steadier when market regimes shift. Most valuable for regime robustness — geometric sufficiency governs static mean-variance efficiency; causal decomposition adds value specifically when conditions change.
What you do
- Search for companies (stocks, ETFs) by name
- Add them to your portfolio (2 or more required)
- Run analysis — we fetch ~5 years of monthly returns and decompose the covariance
What you get
- Standard vs. causal volatility — Compare typical minimum-variance (Σ_obs) vs. regime-stable (Σ_do) portfolio risk
- Shared market influence — How much co-movement comes from macro factors; higher values mean the causal approach helps more when regimes shift
- Recommended allocation — Weights for each holding to reduce risk under regime changes
Investment Analysis: CVC & Venture Signal
How we quantify how institutional and VC investors are likely to view a vendor
The Venture Signal answers: "If we invest in this vendor as a CVC, how would institutional and VC investors see it?" We combine fit scores, ROI projections, and risk into a single signal tuned for credit-union and CUSO realities.
Signal level
Strong VC interest likely → Moderate → Neutral → Weak VC interest. Based on growth trajectory (ROI, payback), strategic/technical fit, and regulatory/risk posture.
Vector: VC demand, exit potential, round risk
- VC demand — How attractive the vendor is to institutional/VC investors (High / Moderate / Low)
- Exit potential — Likelihood of favorable exit (IPO, acquisition) based on ROI and payback
- Round risk — Risk around the next funding round (driven by overall risk score)
FI translation
We translate the signal into credit-union / CUSO terms: e.g., "Attractive as CUSO-led consortium bet," "Viable as CUSO co-investment," "Better as pure vendor, low equity upside," or "Vendor-only; equity exposure not recommended."
Key drivers & caution flags
Key drivers (e.g., strong ROI trajectory, fast payback, strategic adjacency) and caution flags (regulatory risk, extended payback, unclear pricing power) help you assess the investment thesis.
M&A workbench: structure & interventional readouts
Multi-dimensional fit with explicit DAG; do-operator effects on efficiency and ROE where identified
The module scores strategic, financial, operational, and supervisory alignment. The causal subgraph (merge → scale / headcount / branch rationalization → cost synergy → efficiency, ROE) supports interventional summaries such as E[Efficiency | do(merge)], E[ROE | do(merge)], and E[Efficiency | do(headcount_reduction)].
Inputs
- Acquirer profile (saved institution object or manual entry)
- Target balance-sheet, membership, profitability, efficiency, core vendor, examination posture, and related fields
- Execute run — full tensor and narrative briefs materialize in one pass
Deliverables
- Strategic Fit — Geographic overlap, member demographics, product mix synergy
- Financial Synergies — Cost savings potential, scale economies, revenue synergy, estimated savings %
- Operational Fit — Core system compatibility, branch overlap, integration complexity
- Regulatory Readiness — Approval likelihood, exam rating alignment (NCUA/FDIC; CUSO: NCUA CUSO rule)
- Value Proposition — Premium range, accretion/dilution, ROE impact, payback, NPV
- Revenue, Loans, Deposits & Asset Quality — Extended metrics (CUs/Banks); CUSO Service Fit for CUSO targets
- Causal Analysis — Primary value driver (scale vs cost vs revenue), regime sensitivity
- M&A Signal — Deal readiness, integration complexity, timeline risk, key drivers, caution flags
Reference universe: issuer-level scores
Displayed on universe cards and issuer detail routes
Cross-sectional scores are estimated off the public corpus and are institution-agnostic. Strategic fit combines these tensors with your saved institution object.
- Overall Vendor Score — Technology (25%) + Regulatory (25%) + FI Fit (20%) + Value (15%) + Risk inverted (15%)
- AI Genuineness (1–10) — Real ML vs. rules-based. 10 = proprietary ML, published research; 1 = no real AI
- Technology, Regulatory, FI Fit — Component scores for the vendor's capabilities
Launch a workstream
Strategic fit • CVC / equity • Portfolio (CPA) • M&A workbench