AI for the regulated front office.
Shipped on the Control Plane.
The Investment & Wealth Workbench is an active-management and wealth accelerator that ships on the AlphaBitCore Control Plane — 23 agents, 95 workflows, 78 skills, 12 models, and 10+ MCP connectors preconfigured for research, portfolio, advisory, and compliance work, with FINRA 2026 alignment built in. Your CAIO signs the procurement; your advisors, analysts, PMs, and reviewers use working AI on day one. The Control Plane underneath is vertical-agnostic and available standalone for firms whose AI use cases aren’t front-office active research or wealth.
The Control Plane is the product. The Investment & Wealth Workbench is an active-management and wealth accelerator on top.
The Control Plane is sold standalone and runs against any AI workload — banking, insurance, custody, back-office, passive index, or any front-office work a firm builds itself. The Investment & Wealth Workbench is the optional accelerator for active-management and wealth firms: a working front office, ready on day one, configured to your firm by ABC field engineers.
The Control Plane.
Planner, Orchestrator, Agent Runtime, Skills & Tool Registry, Gateway, Event Stream, Contract Registry, Revocation Log, Key Registry, EAC issuance. Vertical-agnostic. Every effectful AI action passes through it; every governed execution seals to it. Sold and deployed without the Workbench whenever a firm’s AI use cases sit outside front-office active research and wealth.
The Investment & Wealth Workbench.
23 agents, 95 workflows, 78 skills, 12 models, 10+ MCP connectors. Preconfigured for active-equity research, portfolio, advisory, and compliance work, with FINRA 2026 alignment and primary financial-data integrations built in. Advisors, analysts, PMs, and reviewers use it the day the firm turns it on.
Adapt it. Then build your own.
ABC forward-deployed engineers adapt the Workbench (or the bare Control Plane) to your firm’s data sources, permission model, supervisory program, disclosure rules, and approval paths. From there, firms modify, customize, and author their own agents, workflows, and skills on the same Control Plane.
Where the vertical shows up in the work.
One platform, four audiences inside a regulated financial firm. Each gets its slice of the workbench — the agents, workflows, and skills for that function — and its own regulator-legible evidence trail.
Research
Move from generic AI answers to governed financial research workflows. Company and thematic research, equity screening, valuation modeling, manager selection.
Portfolio
Exposure analysis, construction, attribution, rebalancing rationale — with every assumption traced and every output replayable.
Advisory & client communications
Portfolio updates, meeting prep, client emails. Compliance scanning and disclosure injection happen inside execution, not after.
Compliance operations
Supervisory review at the rate content is generated. Examiner-ready evidence packages from the sealed event stream. FINRA 2026 mapping structural, not procedural.
Auditable by architecture. Not by afterthought.
FINRA's 2026 Regulatory Oversight Report identifies four categories of risk for agentic AI. AlphaBitCore addresses all four — architecturally, not procedurally. Parallel regulatory mappings — HIPAA, EU AI Act, NIST AI RMF, NAIC model AI rules, ISO 42001, SR 11-7 — are in preparation, targeting Q3 2026 publication.
AI workflows selecting intermediate actions that substitute for human supervisory review (Rules 3110 & 3120).
Every intermediate action requires explicit authorization and Gateway traversal. The event stream records the complete decision chain — supervisory review becomes structurally enforceable, not procedurally aspirational.
Gap between complexity of AI decision processes and the trace logs firms retain. “Outputs alone are insufficient.” (Rule 4511 / SEC 17a-4).
Every event carries identity, scope, authority, decision logic, result, and state delta. The trace is append-only and sealed. Not just the output — the full causal chain Compliance and Audit can replay.
Firms must reconstruct the full chain of AI agent activity, not just final outputs.
Any AlphaBitCore-governed execution can be deterministically replayed from its event stream. Audit moves from log interpretation to execution replay — probabilistic to deterministic.
AI agents acting beyond user's intended scope; regulators require guardrails to limit agent behaviors and decisions.
Authority-separated architecture enforces scope at the runtime level. Denied executions produce zero state change — the formal guarantee that makes guardrails more than a policy document.
Eight of 95, sampled across the four functions.
Every workflow composed of typed building blocks: an ABC workflow (wf:), a skill (skill:), or an MCP connector (mcp:). Use as shipped, configure, or fork.
Company Profile
Pull entity, business segments, ownership, and key metrics for a single ticker. Foundation for any research agent.
Equity Screen
Describe what you're looking for in plain language. Get a ranked, filterable universe back.
Exposure Decompose
Decompose a portfolio across factor, sector, geography, and characteristic exposures.
Performance Attribution
Decompose returns into allocation, selection, and interaction effects with narrative.
Household Consolidate
Group accounts into household views with ownership, registration, and tax-status detail.
Client Comms Draft
Draft client-facing communication with tone, length, and disclosure constraints honored.
Compliance Scan
Scan content against firm and regulatory policies with rule-level citations.
Approval Queue
Stage items into a tiered supervisory approval queue with role-aware routing.
Three questions firms ask before any deeper conversation.
Works with your data
Bring your own data and policies, or start with pre-built workflows and connect over time.
Connects to your systems
Plug in market data, documents, custodial feeds, and internal databases. The platform fits between systems your teams already use.
Hosted or local — your choice
Hosted frontier models or self-hosted local models, picked per task by sensitivity, cost, and firm policy.
Workflow catalog
All 95 workbench workflows by function. Click any for the detail view — job, output, building blocks, screenshot.
Workbench agents
The named AI personas that combine workflows, skills, and MCP connectors into front-office work.
Firm-specific builders
Custom workflows, firm controls, custom data connections, and custom model routing. How the vertical becomes your vertical.
Ready to see it against your front-office work?
The FS briefing walks through the workbench against your research, portfolio, advisory, or compliance team — plus the FINRA 2026 evidence trail an examiner would read and the FDE-led configuration path from out-of-box to firm-specific.