An AI readiness assessment at a hedge fund answers one question: where can AI improve a decision or save real time, and what is in the way? At most funds the answer comes down to four things. The data cannot be traced back to its source. Key workflows exist only in people's heads. Controls were written before anyone used models. And a short list of use cases is ready now, next to a long list that is not.
What the assessment covers
A useful assessment follows the full path from raw input to investment decision. It looks at five areas:
- Data: sources, ownership, lineage, point-in-time integrity, and entitlements.
- Workflows: how research, portfolio construction, risk, and operations actually run each day.
- Technology: where data lives, how systems connect, and what the team can build and maintain.
- Governance: model risk, compliance review, recordkeeping, and vendor oversight.
- People: who will own each system after it ships.
Finding 1: The data is good, but its history is thin
Hedge funds pay for high quality data. The gap is usually in how that data is stored. Vendor files are overwritten when figures are revised. Internal estimates sit in spreadsheets with no timestamps. Research notes are scattered across email and chat. A model trained on this data can see the future by accident, so any backtest built on it is suspect. The fix is point-in-time storage with clear lineage, and it has to come first because every later use case depends on it.
Finding 2: The best workflows are undocumented
The most valuable processes in a fund are often the least written down. Examples include an analyst's earnings prep, a PM's pre-trade checklist, and the risk team's morning review. Each one runs daily, and each person does it a little differently. AI can shorten these tasks once someone defines the inputs, the steps, and what a good output looks like. The assessment maps these workflows in detail, and the map is often useful on its own.
Finding 3: Governance lags the tools already in use
Staff are usually using AI tools before the firm has a policy for them. Analysts paste text into public chat interfaces. Developers use coding assistants. Compliance often finds out late. The risks are concrete: confidential information leaving the firm, material nonpublic information handled without controls, and communications that fall outside recordkeeping. The assessment records what is actually in use and sets out the controls needed. That means approved tools, data classification rules, logging, and a review path for new models.
Finding 4: Use cases split into ready and not ready
Once data, workflows, and controls are mapped, the candidate use cases fall into two groups. The ready group has a few things in common. The inputs are text or structured data the fund already controls. A person reviews the output before it affects a position. The time saved is easy to measure. Examples include summarizing filings and transcripts, drafting research notes from existing models, reconciling operational breaks, and searching internal research archives.
The not-ready group usually involves generating signals or automating decisions. These projects can work. They also need clean point-in-time data, a validation framework, and a clear owner on the investment side. If a fund starts here before those pieces exist, it can lose months.
Finding 5: Ownership is the most common failure point
A system with no owner decays. Vendors change file formats. Models drift. Prompts that worked last quarter stop working. Before anything is built, the assessment names an owner for each proposed system on both the investment side and the technology side. If no one will take ownership, the system does not get built.
What the output looks like
A good assessment ends with a short document the investment committee can act on:
- A ranked list of use cases, each with expected impact, dependencies, and an owner.
- The data and infrastructure work needed to unlock each one.
- A governance framework that compliance has reviewed.
- A sequenced plan that puts a working system into production early, so the firm learns from real use.
I have run a portfolio and I have built these systems. From both sides, I judge an assessment by one test: does it tell the PM what will change in their week, and when? If it cannot, it is just a slide deck.
Where to start
To find out where your fund stands, book a discovery call.