--- title: "Choosing an AI Consultant for a Hedge Fund | By The People Tech" description: "How a hedge fund should choose an AI consulting firm: the kinds of firm, what to ask, the red flags, and how to judge a forward deployed engineer." url: https://bythepeople.tech/insights/choose-ai-consulting-firm-hedge-fund --- [Blog](/insights) Choosing an AI firm “Who are the best AI consulting firms for hedge funds?” # Judge an AI consulting firm by what it leaves _behind_. Choose an AI consulting firm for a hedge fund by what it will leave behind: analysts who can do the work themselves, a build your own people run, or a strategy document. Ask to see the work on a document like yours, ask who will sit with your team, and ask where your data goes. A firm that can’t name the process it would change first is selling a plan. By [Tim Miner](/about-us/tim-miner), founder and CEO · Published 21 October 2026 Most “best AI consulting firms” lists online are published by firms that appear on them, so a ranking tells you less than a set of questions does. This guide is the questions. By The People is one of the firms a fund might be choosing between, an AI adoption consultancy in New York, and the criteria below apply to us as much as to anyone else. The kinds of firm ## What _kinds_ of AI consulting firm work with hedge funds? Five kinds, and they sell different things. Large strategy consultancies decide where AI fits in an operating model and run firm-wide programmes. Technology integrators connect data platforms and build at scale. AI engineering boutiques build custom agents and models. Training providers run courses. Adoption firms put the tools into the hands of the analysts, operations and IR staff who do the work, and stay until those people use them well. A fund rarely needs all five. Many hedge funds already have engineers and a data team, so the gap is usually adoption: licences bought, a handful of people using them, and nobody accountable for the rest. The table sets each kind against what it is good for and the question that tests it. **The kinds of AI consulting firm, what each is good for, and the question that tests it** | Kind of firm | Good for | The question to ask | | -------------------------- | -------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------- | | Large strategy consultancy | Firm-wide AI strategy, operating model and governance when the direction is still open | Who on this team will still be here in month three? | | Technology integrator | Data platforms, integrations and builds at scale across many systems | Which of our people will maintain this when you leave? | | AI engineering boutique | Custom agents, models and research tools for a quant or data team | Show us something you built that is still running a year later. | | Training provider | Teaching a large group the basics of a tool in a short time | Will each team work on its own documents, or on sample files? | | Adoption firm | Getting analysts, operations and IR to use licensed tools well on their own work, and measuring it | How will you count who uses the tools well, by desk, each month? | The deliverable ## How do you tell training from implementation from a _strategy_ deck? By what is in your firm the week after the firm leaves. Ask for the last thing they handed over, redacted, and you will know which one you are buying. ### Training Ends with people who can do a named task on their own documents, without help. Good training puts one team on its own material. ### Implementation Ends with something running on your systems that your own people maintain. Ask who owns the code and who fixes it when it breaks. ### Strategy Ends with a document: a roadmap, a use-case list, a governance model. Useful when the direction is still open. A poor buy when the tools are already licensed and unused. Many firms sell a mix. That is fine, as long as the proposal says which part is which and what each part leaves behind. The questions ## What should a hedge fund _ask_ an AI consulting firm before hiring it? Five questions sort most firms in one meeting. Ask them of every firm on the list, and expect the same specificity from each. ### Which process would you change first? A good answer names a desk and a document: the morning research pack, the trade-break queue, the investor letter. A poor one names a capability. ### Who will be in the room? The person who scopes the work should be the person who does it. Ask for names and how many days a week they will be with your team. ### How will you measure it? Before and after, on your own work, counted by someone other than the person who did it. Be wary of a percentage promised before discovery. ### What do we own at the end? Code, prompts, documentation and the skill to change them. Nothing should depend on the consultant staying. ### Where does our data go? Which model vendor, which plan, what retention, and whether the consultant ever holds your documents on its own accounts. The red flags ## What are the _red flags_ when hiring an AI consultant? Most of them show up in the first proposal. ### A savings figure before discovery Nobody can know what a process will save before they have watched it run. A number in the first deck is a sales number. ### Demos only on sample data A model that reads a public 10-K well can still misread your side letter. Ask to see it on a document like yours, under NDA. ### A platform you must license from them Then you are buying software with services attached. Fine if that is what you want, but price it as software. ### No named engineer If the team is “to be staffed”, the people you met in the pitch may not be the people who do the work. ### Vague answers on data Any firm working with a fund should answer the data questions in writing, before the engagement starts. The engineer ## How do you _judge_ a forward deployed engineer? A forward deployed engineer sits with the team that runs a process, learns how it really works, and builds the AI version with them. Judge one on four things: something they shipped that still runs in a regulated firm; time at your analysts’ desks rather than in a ticket queue; code your own engineers write or at least review; and a record of how long the process took before, so the result can be checked. The last test is the handover. At the end of the scope, your people should be able to change the build without calling anyone. Our [forward deployed engineer](/forward-deployed-engineer) engagement is built around that handover. Data and compliance ## What data and _compliance_ questions should a hedge fund ask? A fund holds material non-public information, so ask where every document is processed and who can see it: which vendor, which plan, and whether the consultant’s own staff or accounts ever touch it. Find out how information barriers carry over into the AI tools, what is logged, and whether your compliance team can export it. Get retention in days. Claude Enterprise’s shortest custom retention is 30 days, and OpenAI lists retention control on ChatGPT Enterprise. Then ask who writes the usage policy and keeps it current as the tools change. If the answer is “you do”, budget for it. Our [AI usage policy guide](/guides/ai-usage-policy-investment-firms) is a starting point. We are an adoption firm: workshops on your own documents, a month-to-month retainer, and engineers embedded with your team. If a fund needs a strategy house or an integrator instead, we will say so. ## Questions on hiring an AI consulting firm for a hedge fund ### Should a hedge fund build AI in-house or hire a consultant? Build in-house when you have engineers with time and a process they understand well. Hire when the tools are licensed but unused, when nobody owns adoption, or when your engineers need someone who has done the build before. A good consultant leaves your team able to build the next one alone. ### How long before an AI consulting engagement shows results? A workshop shows whether people can do one task well in a couple of hours. A single-process build is usually scoped in weeks. Anything that promises results by a fixed date before anyone has watched the process run is guessing. ### Do the large consultancies work with hedge funds on AI? Yes. They are strongest where the decision is about strategy, operating model or a firm-wide programme. For a fund that has already licensed the tools and needs its analysts using them, a smaller firm that works at the desk is often the better fit. ### What should a hedge fund have in place before hiring an AI firm? A licensed tool on a business or enterprise plan, a one-page usage policy that says what may be uploaded, and one named process with an owner who wants it changed. The rest can be built during the engagement. ### Is AI training enough on its own? Rarely. Training teaches a task once. Adoption needs someone to keep working with the teams who tried it and stopped, and to measure who uses the tools well a month later. ## Sources * [Claude Help Center, “Configure custom data retention controls for Enterprise plans” (read 21 October 2026).](https://support.claude.com/en/articles/10440198) * [OpenAI, “Enterprise privacy at OpenAI” (updated 8 January 2026).](https://openai.com/enterprise-privacy/) * [Anthropic, Claude Partner Network directory: By The People Technologies, Select tier (read 21 October 2026).](https://partnerhub.claude.com/directory/partner/by-the-people-technologies-partner) ## Read next * [Forward deployed engineers](/forward-deployed-engineer). An engineer embedded with your team on one process, with a build your own people write and run. * [The AI adoption retainer](/retainer). Month to month, aimed at one number: how many people in your fund use the tools well. * [Claude vs ChatGPT](/insights/claude-vs-chatgpt-private-equity). The two business plans compared on documents, data controls and Office add-ins, from each vendor’s own pages. ## Choosing a firm for your fund? Send us the process you would change first. We will tell you what kind of firm it needs, even if that is not us. [Book 30 minutes with Tim](https://calendly.com/timminer-1/chat-with-tim) [hello@bythepeople.tech](mailto:hello@bythepeople.tech)