Upskill or Hire? The AI Talent Decision

1. The Real Question Behind the Hire

At some point the conversation turns to talent. AI is clearly going to matter, the team is busy, and someone asks the natural question: should we just hire somebody who knows this stuff. It feels like the responsible move.

It is also where firms most often skip a step. A hire feels like progress because a job posting is something you can point to. But a posting is not adoption, and a new face is not a strategy. The real question is not "should we hire an AI person." It is "what capability do we actually need, and what is the cheapest, fastest way to get it that sticks."

Frame it that way and three paths appear, not one. You can train the people you already have. You can hire a dedicated specialist. Or you can bring in an outside partner who already has the skill. Each fits a different firm at a different stage, and the expensive mistake is defaulting to the hire because it is the most familiar shape.

A quick boundary, because it matters. This guide is about the talent decision, train versus hire versus partner. The separate question of who should own AI day to day, the operating model and where the role sits, is its own decision, covered in who should own AI at your firm. Here we are deciding where the skill comes from, not where the org chart puts it.

2. Upskill: Train the Team You Have

For most firms, most of the time, the answer starts with the people already in the building. They know the work, they know the firm, and they are the ones whose hours you are trying to free.

Upskilling works when the capability you need is using AI well, not building it. Drafting the memo the firm's way, screening a CIM against the box, answering a DDQ from the library, reading a monthly pack faster. These are judgment-plus-tool jobs, and the person who already owns the judgment is the right person to add the tool. An analyst who learns to screen with AI is worth more than a hire who knows AI but has never screened a deal.

It also solves the adoption problem at the root. When the people who do the work are the ones who got good at the tool, there is no handoff, no translation, no resented mandate from someone who does not share the desk. The habit forms where the work happens. That is why a structured training program aimed at the real workflows beats a generic skills course or a single clever hire for the bulk of what a firm needs.

Upskilling has a ceiling, though. It makes your people excellent users. It does not, on its own, produce someone who can architect a custom build, wire live systems together, or run a firm-wide program. When you genuinely cross into that territory, training alone stops being enough, and the question becomes who supplies the deeper skill.

3. Hire: When a Dedicated Skill Pays Off

A dedicated AI hire is the right call past a real threshold, and the threshold is higher than most firms assume.

A full-time specialist starts to pay off when there is genuinely a full-time job: a continuous pipeline of things to build and connect, a firm large enough that workflows multiply faster than a trained desk can absorb them, and a maintenance load that needs a permanent owner. If AI is becoming the way the firm runs, not a tool a few teams use, a real role is justified, and it should be a real role with real cover, not a side title on someone already stretched.

The mistake is hiring against that picture before you are in it. A small or mid firm that hires a senior AI specialist to do work that does not yet exist gets an expensive person waiting for a mandate, building things nobody asked for, and slowly disengaging. The role outran the need. You hire the specialist when the build-and-scale work is already there and overflowing, not in the hope that hiring one will create it.

And even at the right threshold, a hire is necessary but not sufficient. A specialist can build the system. A specialist cannot, by themselves, make a skeptical desk change how it works, which is the part that actually determines whether any of it pays off.

4. What an AI Hire Actually Does

It helps to be precise about what a good AI hire delivers, because the title hides a lot and firms buy the wrong thing.

A strong hire builds and maintains: custom workflows, connections to your live systems, the setups that turn a general tool into the firm's tool, and the upkeep that keeps all of it from rotting. That is real, valuable, full-time work once there is enough of it. It is also, notice, mostly engineering. It is the supply side of capability.

What a single hire cannot do alone is the demand side: getting twenty busy, skeptical investment professionals to actually adopt the thing. Adoption is a behavioral problem, not a technical one. It runs on trust earned through checkable wins, a champion the desk respects, and patient change management, and it does not arrive because a talented engineer joined. Firms that hire purely for the build and forget the adoption end up with excellent tools nobody uses, the most expensive outcome of all.

So when you scope the role, scope the whole job or plan to cover the rest another way. A builder plus a credible internal champion is a working pair. A builder alone, dropped into a firm that prizes individual judgment, is a half-bought solution.

5. The Options Compared

Three paths, plus the one most firms actually end up wanting, which is some of each.

The figure makes the real point: these are not four rival camps, they are ingredients. The right answer for most firms is a blend that shifts over time, and treating it as a single forced choice is what produces the premature hire and the unused tool.

6. The Talent You Will Struggle to Keep

Suppose you decide to hire. There is a second problem waiting, and it is the one firms underweight: keeping the person once you have them.

Genuinely skilled AI people are in heavy demand and have a lot of options, many of them at technology companies and well-funded startups that can pay more and offer work the person finds more interesting. A talented specialist at a mid-size investment firm can feel professionally isolated, the only one doing their kind of work, with no peers to learn from and a ceiling they can see. That is a recruiter's dream, and they will get the call.

So the hire carries a hidden tail cost. You pay to find them, pay to keep them, and absorb real risk that they leave in eighteen months and take the context with them, leaving you to start the search again. None of this means do not hire. It means count the full cost honestly, including retention and key-person risk, and ask whether a single point of failure is the structure you want for something the firm is coming to depend on.

For many firms, the answer to the retention problem is not a better comp package. It is a structure that does not rest the whole capability on one person you have to keep happy, which is the case for the next two sections.

7. The Fractional Alternative

There is a path between training your team and committing to a full-time hire, and it solves several of the problems above at once: a fractional partner.

The idea is straightforward. You bring in an outside partner who already has the skill, for a fraction of a full-time role, to do the part the firm cannot yet do itself: build the setups, run the training, and shepherd adoption on the desk. You get senior capability now, without a six-month search in a hard market, without a six-figure salary plus the retention tail, and without betting the whole capability on one hire who might leave.

It fits the reality of most firms better than a hire because the build-and-scale work is real but not yet full-time. A fractional partner flexes to the actual load: heavier while you are standing things up, lighter once a workflow is won and running. And critically, a good partner covers both sides of the job, the build and the adoption, rather than the engineering alone. This is exactly what the AI Operating Partner retainer is, fractional AI leadership that brings the skill and runs the change.

The fractional path also de-risks the eventual hire. If the load grows to a genuine full-time job, you make that hire later, from a position of knowing exactly what the role needs, with workflows already proven and a partner who can help you scope and onboard the person. You climb to the hire instead of leaping to it. For the deeper logic of when to keep this outside versus bring it in-house, see AI: build, buy, or partner.

8. What Each Path Costs

Put rough numbers on it, because the talent decision is partly a math problem and the obvious option is not the cheapest.

Upskilling is the cheapest by a wide margin. A structured training program for the deal teams is a one-time investment measured in thousands, not a recurring salary, and the people you train already have the institutional knowledge. The cost is mostly their time and attention, and the return is the desk doing the work the new way.

A full-time hire is the largest commitment: a senior AI salary, plus recruiting, plus the retention tail, plus the cost of the role sitting underused if you hired ahead of the need. It is the right spend when there is a full-time job to do, and an expensive way to wait when there is not.

A fractional partner sits in between by design. A retainer that flexes to the load gives you senior capability without a full salary or a search, and it pairs naturally with upskilling: the partner builds and trains, your people carry it. For most firms the cheapest path to real capability is upskilling plus a fractional partner, with a hire considered only once the load clearly justifies one.

The number that should drive all of this is the loaded cost of the hours you are trying to free. Measured against an investment professional's time spent on assembly work, every one of these paths is cheap if it lands. The expensive outcome is not any single price tag, it is paying for capability that never reaches the work.

9. A Decision Rule by Firm Size

A rough rule, because firm size is a decent proxy for how much build-and-scale work actually exists.

Emerging and small funds. Upskill the team, full stop, and add a fractional partner for the work the desk cannot yet do. There is no full-time AI job here, and a hire would be an expensive person waiting for one. Get your people excellent and bring in outside skill for the standing-up.

Mid-size firms. Upskill plus a fractional partner is still the center of gravity, often paired with one internal champion who carries the habit between sessions. This is where a hire gets discussed and is usually still premature: the build work is real but rarely a full forty hours a week yet. A fractional partner plus a champion covers it without the salary or the retention risk.

Large firms. A dedicated hire starts to make sense, because the build-and-scale load is genuinely continuous and a permanent owner is justified. Even here, the hire is for the build, and the adoption and program still need real attention, which is why large firms often run a hire and a partner together rather than treating the hire as the whole answer.

The rule underneath the rule: hire when the full-time job already exists and is overflowing, train always, and use a fractional partner to bridge the gap, which for most firms below the largest tier is the gap, not a temporary phase.

10. Where to Start

Before you write a job description, write down the capability you actually need this quarter, and be specific. Is it your desk using AI well inside known workflows, or is it continuous custom building and scaling. Most firms, honestly answered, need the first, and the first is an upskilling problem, not a hiring one.

If you need the deeper build skill too, do not default to a search in a hard market. Bring in a fractional partner who has the skill now, get the workflows proven and the desk adopting, and make any full-time hire later, from knowledge instead of hope. You will hire better, if you hire at all, once you know exactly what the role must do.

The fastest path for most firms is both at once: a training program that makes your people excellent, run alongside an AI Operating Partner who supplies the senior skill and the adoption without the salary or the search. If you are not yet sure which capability you need, an AI Readiness Sprint answers that first, so the talent decision follows the work instead of guessing at it.