The control plane for agentic investing

Your AI desk.Your decision.

An AI investment desk that researches continuously, explains every decision, and never crosses the execution boundary without you. Specialist agents, portfolio context, verification, and risk controls meet in one evidence-driven workflow you control through AI Desk.

Explore how it works Request access
BullyBearAI bear watching over a network of market evidence ResearchOwner gateReconciled

Connected intelligence

How BullyBearAI turns evidence into a bounded decision.

Sources enter once. Specialist desks qualify and challenge them on a shared evidence graph. Portfolio context creates a recommendation; verification, risk, human approval, bounded action, and reconciliation determine what happens next.

FilingsSEC EDGAR10-K, 10-Q, 8-K evidence
MarketPrices & technicalsfreshness and benchmark context
NewsPublic news deskcatalysts, dissent, attribution
SentimentSocial researchbounded read-only signal context
EventsMacro & disclosuresrates, earnings, public activity
Shared evidence graph
RATERISKLEARNVERIFY
ContextPortfolio-aware ratingBuy, Hold, Sell, Avoid, Review
ProofLineage verifiersources match the exact candidate
LimitsRisk monitorcaps, freshness, duplicates, window
PermissionOwner approvalfresh, explicit, hash-bound consent
TruthReconciliationbroker, ledger, portfolio, benchmark
Agents communicate through evidence.They exchange normalized facts, confidence, dissent, freshness, and safe reason codes—not hidden authority.
External sources stay bounded.Unavailable, stale, malformed, or restricted inputs remain visible blockers instead of invented signal.
The owner remains in the graph.Research can inform a candidate, but only protected approval and execution workflows can authorize action.
Meet the desksScroll horizontally → each desk hands evidence to the next
01Research

Research Desk

Looks atUniverse, catalysts, earnings, disclosures
ProducesRanked research queue
Cannot doCreate an order or approval
02Market

Market & Technical Desk

Looks atPrice, volume, regime, benchmark
ProducesFresh market context
Cannot doTurn a quote into authority
03Fundamentals

Filings & Fundamentals

Looks atSEC filings, estimates, events
ProducesCited thesis evidence
Cannot doInvent missing proof
04Signals

Sentiment & News

Looks atNews, public sentiment, dissent
ProducesAttributed catalyst context
Cannot doBypass evidence quality
05Portfolio

Portfolio Context

Looks atThesis, holdings, concentration
ProducesBuy · Add · Hold · Trim · Sell
Cannot doTreat a label as an order
06Proof

Verifier

Looks atLineage, freshness, exact candidate
ProducesVerified or Review Required
Cannot doApprove a candidate
07Limits

Risk Monitor

Looks atCaps, duplicates, window, kill switches
ProducesPass or fail-closed reason
Cannot doOverride owner policy
08Truth

Execution & Reconciliation

Looks atApproved candidate and broker truth
ProducesBounded action and durable receipt
Cannot doAct without protected approval

All desks publish into the same evidence graph; AI Desk is how the user interrogates and directs them.

A portfolio rating, in motion

Why did the rating change?

Ratings are outputs of evidence plus portfolio context—not opaque model opinions. Watch the responsible desks change the label as price, fundamentals, catalysts, concentration, and evidence quality change.

Illustrative market replay · ACMEAgents observing
Illustrative stock-price chart with Buy, Add, Hold, Trim, Sell, and Review Required research labels A synthetic price path rises with volatility while BullyBearAI agents evaluate evidence and produce advisory portfolio labels. No real security, performance, or trade is represented.
BUY · 82%fresh cited evidence
ADD · 86%thesis strengthening
HOLD · 74%concentration watched
TRIM · 77%risk budget elevated
SELL · 79%thesis break detected
REVIEW REQUIREDevidence stale or conflicting
ACME +18.4% synthetic pathBenchmark +7.1%
01 · BuyEvidence earns conviction.Evidence Analyst qualifies fresh filings, news, fundamentals, and price context.
02 · AddThe thesis strengthens.Portfolio Rating sees new evidence while Risk confirms room inside concentration limits.
03 · HoldDissent tempers action.Sentiment and Risk surface mixed signals; the desk waits instead of forcing a trade.
04 · TrimExposure outruns conviction.Risk Monitor detects elevated concentration while the thesis remains partly intact.
05 · SellThe thesis breaks.Evidence and Verifier connect a material change to the exact rating and its proof.
06 · Review RequiredProof is no longer sufficient.Verifier stops progression when evidence becomes stale, incomplete, or conflicting.

Illustrative synthetic replay—not a current portfolio, real security, performance record, recommendation, or autonomous trade. Research labels are advisory. Any live-capable action remains separately bound to genuine evidence, an exact candidate, hard caps, broker review, fresh owner approval, and reconciliation health.

The human boundary

Automation stops here. You decide whether it continues.

Research can prepare and test a candidate. It cannot approve itself. The protected human decision is a hard break in the path, and every downstream gate must still pass afterward.

AdvisoryResearch recommendation
ProofVerifier
LimitsRisk monitor
CandidateProposed action
Hard boundaryHUMAN APPROVAL
BrokerPreflight
BoundedPlacement
TruthReconciliation
Research is not an order.
Verifier pass is not approval.
Notification delivery is not approval.
Preflight cannot place an order.

AI Desk

One conversation with the whole desk.

Ask why, refresh evidence, inspect missing proof, edit your mandate with confirmation, audit portfolio fit, and run safe verifier or risk workflows without hunting through operator screens.

Explain any rating and its evidenceRefresh bounded research inlineRun verifier and risk workflowsShow exactly what needs human attention
Synthetic AI Desk sessionNo private visitor data retained
You

What needs me?

AI Desk · Inbox

Three items: LUNR research is stale, its rating moved from Add to Hold after concentration increased, and one verified candidate is waiting for your protected review.

You

Refresh LUNR and explain the change.

AI Desk · Research + Portfolio

Bounded refresh complete. New filing evidence remains constructive, but catalyst timing weakened and position concentration rose. Rating: Hold. Sources, dissent, and freshness are attached.

You

Run the verifier and risk monitor.

AI Desk · Verifier + Risk

Lineage passed. Risk blocked promotion because the quote is stale. No candidate advanced and no approval or broker action was created.

You

Prepare the approval request when the quote is fresh.

AI Desk · Next step

I can prepare the exact request for review after freshness passes. Preparing or delivering it will not approve or place the trade.

Trust is a system property

It explains itself. Then it asks.

Source-linked and freshness-aware.Missing or stale evidence becomes Review Required, not invented confidence.
Private context, protected authority.Your thesis and portfolio inform research; exact approval remains a separate gate.
Bounded action, broker truth.Preflight, hard limits, kill switches, and reconciliation leave an auditable receipt.

Your research desk should explain itself before it asks you to act.

Join the BullyBearAI private beta to test a supervised AI investment workflow built around evidence, conversation, and explicit human control.

Request accessView public demoReturning customer? Log in above.