Your Next Finance Hire Might Not Be a Person

By Chris Fenster, Founder & Executive Chairman

Datarails published research this week analyzing 5,000 finance job listings across the major portals. The headline: nearly one in three corporate finance roles now requires AI or machine learning skills, up from one in four a year ago. For FP&A specifically, it’s 43%. Accountant roles saw the fastest growth, with AI mentions surging from 18% to 30% in twelve months.

Datarails also found that CFO compensation is rising, while controller compensation is compressing: a finding they summarized as the market “concentrating value at the top of the finance function while applying automation pressure to the roles beneath it.” The specific salary numbers in their report skew low for what I’d consider a qualified controller at a scaling company, but the directional trend matches what we see in the market every day. Strategic judgment is getting more expensive, and execution-layer skills are getting repriced.

According to Gartner, acquiring and developing AI talent is now the top challenge for CFOs. PwC found an 80% increase in job postings seeking what they call “fusion skills”: accountants who can also think about data, technology, and cross-functional collaboration. Meanwhile, the AICPA has been reporting a nearly 20% decline in CPA exam candidates since 2019. The pipeline that produced traditional finance talent for decades is thinning at the exact moment the job requires a completely different skill set.

W2 or Outsource? Wrong Question

Most scaling companies approach their finance function as a binary decision: hire a W2 or outsource. Build the team internally, or rent it from someone else.

Every board conversation, every investor check-in, and every founder decision tree starts here. But this binary decision framework was designed for a world that isn’t coming back. Both options have failure modes that everyone pretends don’t exist.

W2-only means you’re paying fulltime salaries for roles where the job description is changing every six months. Let’s say you post for a controller in January, interview in March, and onboard in May: by August, the role you hired for has shifted underneath the person you just hired. (I wrote about this in The Talent Paradox: by the time they reach $100M in revenue, most companies will have cycled through two or three finance leaders.)

Outsourcing has its own problems. You get bodies without context, deliverables without continuity, and a fresh learning curve with every engagement. The consultant zooms in, produces a deck, and leaves, and whatever institutional knowledge they’ve acquired walks out the door with them.

I saw a CFO Brew piece last week profiling what PE search committees actually want in a portco CFO. The recruiter, Mark Jansen at Frederick Fox, said this is the most competitive CFO search market he’s seen in 25 years. He listed four CFO non-negotiables:

An operational copilot who can partner with the CEO

Advanced AI experience (not “I’ve been looking into it”, but actually having built a tech-enabled finance function)

Transaction readiness at 72-hour notice

The backbone to push back on the board when the numbers say what the board doesn’t want to hear

I read his list and thought, you’ve got to be kidding me. Someone who’s simultaneously a strategic executive, a technology architect, a deal-ready operator, and a diplomat with backbone? Good luck finding that at any salary.

The hire-a-W2-or-outsource binary just doesn’t work when we expect finance professionals to be mystical creatures. Companies need to stop looking for one person and start building a system that delivers all four.

The operational copiloting would come from an embedded team with deep company context.

The AI capability would come from infrastructure, purpose-built and tested for AI-forward finance functions.

The transaction readiness would come from standardized processes that don’t need to be assembled from scratch when the clock is running.

The backbone would come from a long-term partner, incentivized by the company’s success because they’re not worried about job security.

We’ve Seen This Pattern Before

Every decade or so, finance undergoes a major structural change across four distinct transitions.

In the 1990s, the shift was integration. ERP systems brought disparate departmental databases into a single source of truth. The trade-off was brutal: monolithic software, 18-month implementations, and rigid processes. But the companies that invested had a fundamentally different operating capacity than the ones still running on spreadsheets.

In the 2000s, the shift was digitization. Finance was still largely analog — desktop QuickBooks, paper-based processes, Excel as the operating system for everything. The real work was getting transactions into systems at all. When we started Propeller in 2008, most of the companies we saw were running on spreadsheets and shoeboxes. The companies that moved early built the foundation for everything that followed.

In the 2010s, the shift was cloud and composability. SaaS changed how finance software was delivered: cloud-based, subscription-priced, and accessible from anywhere with NetSuite, Intacct, and Workday. I remember debating in 2012 whether anyone would trust their general ledger to the cloud. By mid-decade, APIs, data warehouses, and modern ETL meant the finance stack became something you could assemble from best-of-breed components — but someone had to assemble it, and someone else had to know what to do with the output once it was assembled.

Now, we’re in the fourth transition, where the shift is agentic. AI agents don’t just process data or generate reports: they reason about data, detect anomalies, surface patterns, and draft analyses.

The technology layer is no longer something people operate during business hours. It operates continuously — reconciling at 2am, flagging exceptions before the team arrives, executing workflows that used to wait for someone to be at their desk. That shift demands a corresponding one from the people around it: from operating tools to governing systems that run without them. Neither side works alone. Agents running on ungoverned data are dangerous. Skilled people without agentic infrastructure are bottlenecked. The companies that get this transition right will rebuild both simultaneously.

Each of these transitions followed a similar pattern: the technology arrived before the organizational model caught up. You didn’t need the best tools to survive, but you did need to rebuild your team around the new reality before your competitors did.

When we started Propeller in 2008, we saw companies struggling with the shift from analog to digital. They didn’t need another software license, they needed embedded infrastructure: people, processes, and systems woven into their operations. I think this moment is structurally identical: different technology, but the same organizational problem.

What the Team of Tomorrow Actually Looks Like

I’ll give you a real example. Hero is a growth-stage client that had brought their entire finance function in-house. Good team, solid W2 hires; it was the right call at the time.

Then they came back to Propeller: not because the W2 team had failed (the team was doing exactly what they were hired to do) but because Hero’s leadership realized that the combination of their internal people and our embedded team produced measurably better outcomes than either group could alone. They lived and breathed the business every day, and we brought pattern recognition from hundreds of similar companies, standardized processes, and a platform that was getting smarter with every engagement.

This was the unlock. The answer to the W2-vs-outsource question is “both”: integrated, with a shared intelligence layer underneath.

The finance function Hero is building (and that I think most scaling companies will need over the next few years) has four elements: all necessary, none sufficient on its own.

A company’s own people anchor the function with institutional knowledge, cultural context, and day-to-day accountability. They’re the people who know that revenue dipped in Q3 because three enterprise deals slipped simultaneously, or that the VP of Sales is probably going to get managed out. As AI takes over the mechanical work, these roles don’t disappear; they get elevated. The controller becomes the orchestrator of a system, not the operator of a process.

The embedded partner brings what no individual W2 hire can carry: cross-company pattern recognition. When you’ve worked inside 1,400+ companies over 18 years, you’ve seen the current crisis before. You know how to standardize a company’s bespoke processes so the data is clean, the workflows are consistent, and the whole system is ready for agents to operate reliably. That’s not something a company can do for itself, because it requires seeing the pattern across hundreds of companies, not just your own.

AI agents handle the work that used to consume 50% of a finance role’s bandwidth: reconciliation, variance analysis, anomaly detection, and rolling forecasts. But an agent without standardized processes and clean data is just expensive autocomplete: only as reliable as the infrastructure underneath it. If you bolt an AI tool onto a chaotic finance stack, you get faster chaos. Put it inside a well-structured system with a data layer built from years of normalized financial engagements, and it becomes a genuine force multiplier: one that compounds as the data gets richer.

The intelligence layer ties it together. Cross-client benchmarking, longitudinal pattern data, processes that make agents reliable across different companies at different stages. This is the compounding asset. Every month of clean data from every engagement feeds back into the system, making it smarter for the next client. It’s also the piece that can’t be replicated by any individual hire or standalone tool — you can’t compress 18 years of financial data from 1,400 companies into a software subscription.

Now, imagine any one of these four elements in isolation: it doesn’t work.

A great W2 team without the intelligence layer is flying blind. A powerful AI agent without embedded human judgment is a liability. An embedded partner without data infrastructure is just a more expensive consultant. The advantage is how they’re wired together.

Don’t Just Take My Word For It

The largest professional services firms in the world are all arriving at the same conclusion…and it isn’t replacing humans with AI. It’s building hybrid systems.

CB Insights published a report on the future of professional services documenting this shift. EY is deploying 150 AI agents to 80,000 tax professionals through what they call a “client zero” approach: test it on themselves first, then roll it to clients. KPMG’s Swami Chandrasekaran described the near-term future as “AI-native problem solvers who blend domain expertise, creative thinking, and AI.” Accenture has trained 500,000 employees on generative AI. These aren’t AI-replacement strategies; each one of them is a hybrid human-AI model like the one I’ve described above.

These firms have the resources to retool at a massive scale, but the scaling startup doing $10M or $50M in revenue does not. It’s impressive that HPE built an internal agentic AI tool called “Alfred” that cut their financial reporting cycle by 40%, but HPE is a $30 billion company with a dedicated engineering partnership with Deloitte. The founder running a sub-100-employee company with a three-person finance team doesn’t have that option.

This is the gap Propeller exists to fill. Last month I wrote about why the infrastructure matters as much as the technique (the foam pit that makes the Fosbury Flop possible). We’re building infrastructure so that companies at $5M or $50M 0r $150M can operate with the same structural advantages that HPE and EY are building for themselves.

What Comes Next

I’ve been at this for 18 years now; that’s long enough to have been wrong about a lot of things and right about a few. The thing I feel most confident about is that the finance function of 2028 will not look like a bigger version of the finance function of 2026. The roles will be different, the team composition will be different, and the tools will be different. Companies that start building the integrated model now will have compounding advantages that late movers can’t replicate.

The binary choice between building in-house and outsourcing was always a false constraint. What’s emerging is something stronger: a system where humans supply judgment, relationships, and accountability, where agents supply speed, memory, and precision, and where the intelligence that connects them gets smarter with every month of data.

If you’re running finance at a scaling company and the question on your whiteboard is, should I hire a controller or outsource? you’re asking the wrong question. The better question is, What system am I building, and will it compound as we grow?

This is the question we’re trying to answer: not just for our clients, but for ourselves. We’re rebuilding Propeller’s own operating model around this thesis right now, because we think you have to drink your own champagne before you can credibly serve it to someone else.

So far, the results are tasty. More on that soon.