--- title: "AI Agents for Funds" canonical: "https://www.fume.finance/ai-agent-funds" description: "AI agents, delivered and managed. Instead of selling software, Fume delivers the finished task: we find the process that eats your team's hours, build an agent…" --- # AI Agents for Funds AI agents, delivered and managed. Instead of selling software, Fume delivers the finished task: we find the process that eats your team's hours, build an agent for it, deploy it into your stack, and manage it while it runs on its own. ## How it works 1. **Find the process**: We sit with the people who actually do the work and map every manual step. The process that eats the most hours is where the first agent goes. 2. **Build and deploy the agent**: We build the agent around how your team already works and wire it into your existing systems. No migration, no new platform to learn. 3. **Let it run**: The agent performs the task on its own, every cycle. The output lands in the drive, inbox or platform your team already opens. 4. **We maintain it**: We monitor every run, fix issues before they reach you, and ship upgrades as your tools evolve. You get the result of the work, not another tool to babysit. ## Not a chatbot ChatGPT and Copilot are useful, and they still leave the work with your team. An agent has access to your systems, a defined task, and permission to complete it end to end. Your team reviews the output instead of producing it. ## Where we start: the NAV lifecycle Four examples across the NAV lifecycle. Wherever data moves by hand in your operation, an agent can take it over. 1. **Ingest agent**: Pulls data from brokers, banks, email inboxes, and onchain transactions every morning. Manual logins and exports disappear. 2. **Structure agent**: Takes any statement format and turns it into structured transactions, across crypto and traditional assets, in the exact layout your platform expects. 3. **Push NAV agent**: Pushes structured data into Paxus, NTAS, Geneva, Investran, or any NAV engine. The platform runs the calc, the agent returns NAV, fees, and the cap table. 4. **Tokenize agent**: Feeds NAV results, payouts, and the latest unit register into Fume's tokenization engine, fully onchain and end to end. ## More agents we build - **Compliance agent**: Monitors regulatory changes across jurisdictions and flags issues before they become problems. - **Document agent**: Reads legal documents, term sheets, and investor reports. Summarizes, extracts, and flags the terms that matter. - **Risk agent**: Live risk assessment across your portfolio. Spots concentration, correlated exposures, and unusual patterns before they hit. - **Communication agent**: Drafts investor updates, quarterly reports, and responses to common questions in a consistent voice, in a fraction of the time. Need something else? Fume builds bespoke agents wired to your stack. ## How we build agents - **Wired into your stack**: Works with your custodian, admin platform, trading venues, and data providers. No forced migration, no rip-and-replace. - **Bespoke, not off-the-shelf**: We write, deploy, and hand over custom agents built around your process. No 200-page reports, no generic product. - **Guardrails that hold**: Agents act inside limits you set. Every action is logged and auditable, so your compliance team can trace each run. - **Managed by us**: We monitor every run, catch the edge cases, and keep improving the agent over time. You get the result, not a maintenance burden. ## Built to pass your compliance review - **Paperwork before data**: We sign a data processing agreement and share a security overview of the exact setup before any client data flows. - **Certified infrastructure**: Client data runs on a short list of ISO 27001 and SOC 2 certified providers, hosted EU-first and encrypted in transit and at rest. - **Your data never trains models**: Model calls run with zero data retention. The processing is transient, and nobody stores your data or trains on it. - **Humans stay in the loop**: Every run is logged and auditable, the output goes to your team for review, and your manual process keeps running as the fallback. Full details: [Security documentation](https://www.fume.finance/docs/company/security). ## Fully private: agents that run locally Some funds can't send data to a public AI cloud, and they shouldn't have to. Fume deploys the same agents on open-source models that run entirely inside your own infrastructure, with the same guardrails and audit logs. The only difference is that no data ever leaves. - **On your infrastructure**: Open-source models run on your own servers or in your private cloud. Investor data, positions, and documents stay inside your firewall. - **No third-party AI cloud**: Your data never touches a public model API, and nobody trains on it. Since there's no AI vendor in the chain, compliance has nothing new to review. - **Proven in production**: Fume already runs fully on-prem deployments for managers: Docker stacks on their own infrastructure, with their own keys, monitoring, and database. ## The Fume stack for AI Every part of Fume is built for programmatic access, so the agents we deliver can lean on each layer. Agents on the other side of the table use the same stack to browse, subscribe, and invest. - **Fund Builder**: Agents pick legal, admin, custody, and audit providers through the Fund Builder. - **Tokenization Engine**: Free tokenization. Agents get programmable ERC-6909 shares they can issue and distribute. - **OmniClear**: Instant stablecoin settlement. Subscriptions and redemptions clear in the same block. - **Fund Registry**: A public directory of tokenized funds. Agents browse, evaluate, and invest.