Firetower · Custom portfolio intelligence
A private portfolio operating system, built around a wealth manager’s holdings, income objectives, and risk rules. It observes continuously, explains what moved, models decisions before they happen — and preserves discipline when markets get disorderly.
$K
$K/mo
DRIP ON
buys gated in YELLOW
41%
semis & AI · flagged
The engagement · Client 002
Client 002 is a private wealth manager running a small family office: a leveraged income portfolio with real margin, real distributions, and no research desk behind it—one person responsible for a family’s capital, with obligations that don’t pause for market hours.
The raw material was all there — live positions and margin metrics inside the brokerage, distribution announcements scattered across issuer sites, market catalysts across news and social feeds, and the house rules living in documents and memory. Six sources of truth, zero shared clock. Every volatile morning meant manual math under pressure.
Firetower turns fragmented brokerage data into a governed portfolio operating system.
Every figure on this page is illustrative — the shape is real, the dollars are not.
The operating model
Four capabilities, one governed workflow. Ordinary broker tools stop at the first.
One live portfolio truth.
Explain what actually moved.
Model the decision before making it.
Keep the strategy intact under pressure.
A day in the system
Overnight futures and global context, portfolio risk condition, distribution outlook, verified catalyst calendar.
Live movement, verified holding and sector drivers, updated LTV and liquidity, material changes since morning.
Final performance, income engine and upcoming ex-dates, day-in-review narrative, next session’s watch list.
Session-aware: knows regular hours from after-hours, overnight futures, weekends, and holidays — and answers accordingly.
“No catalyst” is a claim too — the audit demands evidence in both directions, so silence is never just a shrug.
Inside the cockpit
Recreations of the client’s actual dashboard tiles, with the numbers swapped. Margin health lives in the command center above; here’s exposure, income, and the decision-support layer.
$K/mo
6 events
BUY · $K · covered-call fund
Modeled against live broker margin data. Preview only — nothing is submitted.
YELLOW mode caps new buys at the solver ceiling — $K today. This add survives the base case but breaches ORANGE under a −10% stress.
Advisory only · human approval required for all actions
The governed playbook
The client’s rules, moved out of their head and into one tested definition — thresholds, modes, and a hard dollar floor. The same definition renders the dashboard, constrains the what-if solver, shapes briefing language, and wakes the watchdog. Two languages, one source: nothing on the desk can disagree about risk.
Case study · a high-volatility session
Semiconductors fell nearly 5% while the broad market finished positive — and the portfolio’s concentrated AI exposure amplified the day into its largest drawdown of the quarter. The exact scenario the playbook was written for.
The system identified a sector rotation, not a systemic crash, quantified what leverage did to net liquidation value, and tracked excess liquidity from the morning low through the afternoon recovery. When the bounce came, governance held: a recovery is not new buying capacity. The client made one pre-planned move, on time, with the evidence attached — and next-morning acceptance checks reconciled every share count and balance against broker records.
Trust architecture
Every endpoint carries a timestamp and cache age. Stale data is labeled stale. A source failure cannot silently masquerade as a live value.
Distributions confirm from issuers or press releases. Market claims carry sources. Broker-reported risk stays separate from modeled interpretation.
Core reports never depend on a model completing. Research degrades independently without blocking the facts. Every scheduled report is audited after delivery.
Broker access is read-only. Previews never submit orders. AI researches; it doesn’t get to assert or act. Material actions require explicit approval.
A consulting installation, not a subscription
Every installation begins with your accounts, holdings, leverage structure, income objectives, decision rules, and how you like to be told things. The software is the artifact; the operating model is the product.
Accounts and custodians, portfolio construction, income and withdrawal objectives, leverage structure, and the operating rules you already half-follow — written down at last.
Brokerage integration, your custom risk model, distribution intelligence, dashboards on your own hardware, scheduled and on-demand reporting, alerts and audits.
Decision playbooks, human-authorization boundaries, source and freshness policies, change control, and ongoing system health — documented so it outlives any one tool.
Firetower is a reference installation. Big Sky AI takes on a small number of private builds like it — for portfolios, and for businesses that run on the same discipline.