Secure AI operations for investment organizations

Where the work becomes intelligence.

Genba is an agentic AI capability factory for the modern firm. Tell Genba the capabilities you want and it coordinates specialist agents that discover, develop and command systems within your hard constraints. What results is compounding institutional intelligence you own.

Managed single tenant · Credential-isolated execution · Firm-owned capability

GENBA / DELEGATED OPERATIONS GOVERNED

A stakeholder directs an outcome. Genba takes operational ownership, discovers approved services inside the firm’s secured perimeter, composes specialist AI agents, and plans or builds what is missing. Governed execution handles live system actions without exposing credentials to an AI agent. Every request, decision, approval, insight, and new capability remains as institutional intelligence.

01

Designed from operating experienceGenba is a clean-room implementation of lessons learned operating AI agents inside a leading hedge fund

02

Zero AI-related incidentsTwo years of operation with no credential exposure, unauthorized action, or security or compliance incident

03

10× principal leverageAI agents substantially expanded how much of the fund a principal could operate directly

What can I delegate?

Delegate the work around the decision.

Work that crosses sources, systems, and teams—too consequential for an ungoverned assistant, too persistent to leave manual. That is where Genba begins.

Working drawing: many separate holdings, managers and entities descending through one governed operating layer into a single consolidated view.

MFO · SFO · Investment holding companies

Institutional operating depth without institutional headcount.

Genba advances work across assets, managers, entities, and stakeholders, surfacing only the decisions that require human judgment while retaining the systems and knowledge behind them.

Own consolidated reporting

Unify portfolio, exposure, liquidity, and cash into one approved view.

Monitor managers and assets

Read letters, meeting notes, and KPIs; surface changes, exceptions, and follow-ups.

Run committee follow-through

Advance packs, decisions, actions, and stakeholder communication through approval.

The return: wider coverage, stronger continuity, and compounding operating capacity without building a larger office.

TŌRYŌ / GENBA'S CHIEF OF STAFF SYSTEM

Hit objectives while compounding capabilities.

Tōryō is the operating mind behind each objective, directing specialist AI agents and carrying every obligation to completion. It keeps work moving across systems and time, surfaces only decisions that need human judgment, and leaves the firm more capable each time.

Explore the Tōryō architecture
GENBA / INSTITUTIONAL EVOLUTIONCOMPOUNDING

DELEGATED OUTCOMES

INSTITUTIONAL CAPABILITY GRAPH

Service registryApproved capabilities
Reporting subsystemComposable component
AI agent skillReconciliation logic
Ownership graphDavid · Ops → Fund Admin
Operating insightEscalation pattern
Evaluation historyQuality + improvement

Every authorised team and AI agent can compose the next solution from what already exists.

THE WORK GETS DONE · THE CAPABILITY STAYS · THE NEXT OBJECTIVE STARTS AHEAD
This timeFinished work
Next timeComposable building blocks
Over timeA shared capability graph with lineage

DAIMON / GOVERNED EXECUTION

Let AI agents act. Keep the firm in control.

Daimon keeps AI agents within the firm’s controls. Every instruction, approval, action, result, and token is attributable to the objective.

Read the technical introduction to Daimon
DAIMON / CREDENTIAL-ISOLATED EXECUTIONAI AGENT ACCESS TO CREDENTIALS: ZERO

AI AGENT / INTENT

Retrieve Q2 exposure from the portfolio system
service: portfolioaction: readscope: Q2
NO CREDENTIALS

DAIMON GATEWAY

LIVE CREDENTIALNEVER LEAVES HERE

APPROVED SYSTEM

Portfolio managementEXECUTED BY DAIMON
PERMITTED RESULT → AI AGENT
THE REQUEST FIRES · THE RESULT RETURNS · THE AI AGENT NEVER SEES THE CREDENTIAL
01

Credentials stop at Daimon

No secret—standing or temporary—enters an AI agent’s context, runtime, memory, output, or logs.

02

Consequence stops at a person

Anything outside policy reaches a named owner with the case attached. Without approval, it does not happen.

03

Audit every action. Account for every token.

Reconstruct any assignment from instruction to result—including its sources, model usage, approvals, system actions, outcomes, and cost.

KYŌZŌ / INSTITUTIONAL CANON

The firm should never have to relearn itself.

Kyōzō is the corpus the firm keeps: what it knows, what it has built, and the conditions under which each may be used again.

Read the technical introduction to Kyōzō
KYŌZŌ / INSTITUTIONAL CANON 1,284 ITEMS96 CAPABILITIES100% ATTRIBUTABLE
THE FIRM’S CANON COMPLETED WORK FACTS FIRM DECISIONS DIVISION CAPABILITIES FIRM SERVICES APPROVED EVALUATIONS RESTRICTED Q1Q2Q3 Investor letter NAV pack DDQ response Reconciliation Board pack Manager review NAMED REVIEW — NOTHING ENTERS THE CANON WITHOUT IT source: q2-investor-letter/0117owner: Priya · IRversion: 2 · approved source: q2-nav-pack/0184owner: Marcus · Fund Opsversion: 5 · approved source: ddq-consultant-a/0203owner: Elena · Complianceversion: 1 · approved source: q2-recon-exception/0248owner: David · Ops → Fund Adminversion: 3 · approved source: q3-board-pack/0261owner: Priya · IRversion: 2 · approved source: manager-review-h2/0290owner: Sofia · Investmentsversion: 4 · approved NEW OBJECTIVE Next quarter’s investor update COMPOSED FROM 3 HELD ENTRIES · 2 JUST ADMITTED

SCROLL THE TIMELINE →

Six completed assignments arrive at the review line.A named owner approves each one in turn, with its lineage attached.Each approved contribution enters the canon, under the work that produced it.The next objective composes long-standing entries with the new ones.

01

What the firm knows

Facts, relationships, and decisions, each with its source and named owner.

02

What the firm can do

Workflows, skills, tools, and the approved services they run against.

03

When to use what

Scope, version, conditions, and the record of where each has worked.

Who is behind it?

Built by people who know the operating stakes.

Work directly with people who have led fund investing, engineering, and production AI.

Makoto Tominaga

Production AI · Funds · Delivery

Two years running AI agents in production inside a leading hedge fund; 25 years across kernel, security, and AI. Founder-CEO with delivery-organisation experience; owns Genba’s platform architecture and delivery.

Alex Norris

Investments · Quant

Head of Investments at a leading hedge fund, reached through quantitative research and Head of Quant across five years. Imperial MSc, with distinction.

Steve Lim

Engineering · Fund systems

Head of Engineering at the same fund, responsible for the trading, data and production infrastructure it runs on. UCL Computer Science, First Class.

QUESTIONS / DIRECT ANSWERS

Before AI enters the firm.

What investment leaders need to know before AI moves from assistance to accountable work.

01

What is Genba—and when should an investment organization use it?

Genba is a governed AI operating system for family offices, private equity, venture capital, corporate venture, and hedge funds. It is designed for work that must move beyond drafting an answer: Genba plans and completes a multi-step objective while preserving human authority, a complete audit trail, and firm ownership of every capability created.

02

What investment work can Genba take from objective to completion?

Genba can coordinate work across the investment lifecycle: research and thesis development, sourcing and diligence, investment or portfolio operations, fund reporting, investor relations, compliance, and back-office workflows. It is most valuable when the work crosses sources, systems, and teams but still requires accountable human judgment.

03

How is Genba different from ChatGPT, an AI copilot, or workflow automation?

Chatbots and copilots help a person complete individual steps; fixed automations repeat a predefined process. Genba accepts the outcome, assembles specialist AI agents, builds what is missing, governs work against live systems, and leaves the organization with reusable workflows, tools, skills, and institutional knowledge.

04

How can an investment leader delegate an outcome to AI without managing every agent?

Through Tōryō, Genba gives the leader one accountable AI Chief of Staff rather than a collection of disconnected agents. It translates the outcome into a plan, composes and supervises the specialist team, carries unresolved obligations forward, and brings only consequential decisions back for human judgment.

Explore Tōryō
05

How can AI agents use live systems without seeing credentials?

Genba’s control gateway, Daimon, keeps credentials outside the AI agent entirely. It interprets the requested action, checks the firm’s human-governed permissions and approval rules, executes through the approved system connection, and returns only the permitted result.

Explore Daimon
06

Can a regulated investment organization audit AI work down to the token?

Yes. Genba attributes every instruction, source, model call, token, policy decision, approval, system action, result, and cost to the objective that caused it. The firm can reconstruct what happened, why it happened, who authorised it, what it cost, and what changed.

07

How can AI work become reusable institutional knowledge and capability?

Genba carries each assignment forward through Kyōzō, its governed institutional canon. More than documents or model outputs, Kyōzō preserves the workflows, components, tools, skills, decisions, relationships, evaluations, and provenance that authorised people and AI agents can safely compose into future work.

Explore Kyōzō
08

How can an investment organization test Genba without a long IT project?

Start with a 30-day Genba One Experience Seat for one Principal in a dedicated, managed environment. If the evidence is there, continue with Genba One or add resource-scoped team tenancies through Genba Team. A sophisticated production Capability can then be co-developed with Genba FDEs; firm-wide adoption graduates to Genba Institution.

See the Experience Seat

Plans / Operating surface

Start with one seat of judgment. Grow into the institution.

Every plan includes Tōryō, Daimon, Kyōzō, managed operations, evaluations, and upgrades. Model and cloud usage are paid by the client and itemised separately.

One Principal

Genba One

£3,500 / month

A managed, single-tenant Genba deployment for one accountable decision-maker.

  • Dedicated AI Chief of Staff
  • Guided Activation
  • Standard connectors + resource mapping
  • Monthly Capability Clinic
Start with Genba One

Firm-wide operating surface

Genba Institution

Designed with you

A firm-wide Genba deployment with defined FDE capacity and the controls, integrations, and enablement required for institutional adoption.

  • Firm infrastructure + systems integration
  • Training + operating enablement
  • Guided Capability development
  • Firm-level governance + support
Design the deployment

Model and cloud usage pass through at provider cost and are itemised on the ledger.

Capability Build / FDE co-development

One sophisticated Capability. Live and accepted.

Add a fixed-outcome Capability Build to Genba One or Genba Team. Genba FDEs own architecture, integration, evaluation, deployment, training, and in-scope remediation through production acceptance.

Scoped with you

Fixed acceptance criteria · milestone billing · final 20% due on production acceptance

Define the Capability

進化する組織には、進化する能力が要る。

Start with evidence, not procurement.

Give one stakeholder 30 days. Let real work decide whether Genba deserves a larger role.

Request an Experience Seat