29 Predictions: How AI Will Reshape the Finance Skills that Actually Matter
Predictions #22–24: The Skills Landscape Is Going to Shift Far Faster Than I Think Most People Realize
Last issue, I focused on what’s being displaced — shrinking analyst pools, commoditized modeling skills, AI running month-close in the background. This issue is about what replaces it, and more importantly, what you can do about it.
Because the questions I hear most often — from analysts, directors, CFOs, and everyone in between — is about how they should pivot:
>What skills are actually going to matter? Do I have them? Where should I focus my efforts on developing myself? Am I in danger of being replaced? Where does AI win and where are humans indispensable?
Here’s my honest attempt to answer that.
Disclaimer: The following views and opinions are my own and do not reflect the views or opinions of any professional organizations or enterprises I work/partner with.
Prediction #22: Finance and Strategy Will Merge — but Only at the Senior Level. Finance Professionals will Become Less Technical and More Interpretive.
I remember as a young financial advisor, my Senior Managing Director would make a joke in client meetings. He’d say: “Carl does all the work. I get all the credit.” He meant it in good humor, but it subtly rubbed me the wrong way. Because for as long as I can remember, there’s been an informal hierarchy in most organizations: strategy people make the decisions, finance people validate them.
I predict that separation is eroding. And AI is going to accelerate it.
The reason finance and strategy were historically distinct functions is that financial analysis was time-consuming and specialized. You needed people dedicated to building models and running scenarios just to keep up with basic reporting requirements. Strategy could afford to operate at a higher altitude.
That’s where my Senior Managing Director sat, while I sat in finance.
But when AI handles the data gathering, modeling, scenario analysis, and routine financial reporting, the finance professional has time and capacity to do something different. They move up in altitude. They’re not building the models anymore — they’re overseeing the process, asking what questions the model should be answering, and connecting the findings to business decisions.
But the 1-trillion-dollar question I keep asking myself is: if young professionals are going to defer much of the skill-building work to AI, how are they going to level up to positions of influence that require domain expertise? I don’t have a clear answer. And that’s one of the biggest unresolved questions in this transition.
In the current cohort of Claude, Copilot, and Dynamic Excel for Finance that I’m facilitating, I’m revealing to the attendees the ability to codify our finance skills. Put more simply, I can take my own FP&A and modeling frameworks and have Claude replicate it reliably and at scale.
What’s somewhat concerning is that it takes me out of the driver’s seat as a builder and repositions me as a reviewer. But the reason I can still sit confidently in the seat of the reviewer is that I know the underlying finance very well.
I believe that on the higher rungs of the corporate ladder, finance and strategy will increasingly converge rather than stay separate – mainly because the value of the financial doer will not carry the weight it once did.
I believe the CFO of 2030 will be less focused on their heavy controllership background and more focused as a finance-strategy hybrid. As controller work shifts toward oversight rather than execution, the CFO function becomes dramatically more powerful and more demanding.
Prediction #23: Prompt Engineering Becomes a Core Finance Competency for the Next One to Two Years…and Then Hardly at All
Years ago, getting consistently good results from AI required following what I called a four-step framework:
- Make the prompt specific.
- Provide context.
- Provide guidance.
- Review the results.
Pretty easy, right? Not so fast.
I often found myself providing increasingly detailed inputs to get the AI to produce the result I wanted. And once I’d invested that much time crafting the prompt, I’d be reluctant to abandon it, even when the output was clearly taking me in the wrong direction.
The quality of prompts mattered. And at the time, knowing how to structure those prompts was a genuine skill. Last week I was with around 25 people from a SaaS company, helping them distinguish one set of prompts from another. When they ran prompts that followed the 4 rules, they ended up with better results. When they kept the prompts vague, Claude went its own way.
Knowing how to effectively direct AI tools is going to be a real, meaningful differentiator for finance professionals over the next few years. It begins with prompts. Analysts who understand how to frame a question, structure context, and evaluate the quality of an AI-generated output are going to operate at a different level than the ones who don’t.
But what many people aren’t realizing – because I’m seeing it and hearing it live in the groups I work with – is that effective instructions to AI go well beyond basic chat. The full capability of AI extends into frameworks that include skills, workflows, and subagents. Some of these will be deterministic, guided by humans. Others will be augmented or automated.
Perhaps the greatest irony I see is that, as with so many of my predictions, the skills that separate the trailblazers from everyone else over the next few years will eventually become commonplace and cease to be an advantage at all.
Looking at my personal experience in AI and finance, I sometimes feel like a trailblazer. But it’s only a matter of time before the whole population catches up. Which, of course, is critical for our collective prosperity.
Prompt engineering as a specialized skill will commoditize as the tools get better at inferring intent, interfaces improve, and the baseline expectation for every finance professional rises. We’re already seeing, as more and more people upload their content into these engines, Claude, ChatGPT, and Copilot keep getting better with less guidance from us.
By 2030, saying “I know how to prompt AI” will sound like saying “I know how to use Google.” It won’t distinguish you from anyone.
The reason to build these skills now is not that they will remain a permanent advantage. It’s that the professionals who understand AI deeply during this transition period will develop judgment about AI outputs that proves valuable even after the skill itself becomes standard. They’ll have seen enough to know when to trust the output and when to push back.
I see the window being short – just a handful of years. So I suggest we use it.
Prediction #24: Restructuring and Turnaround Advisory Becomes One of the Most AI-Resistant Disciplines in Finance
I’ve done restructuring and crisis management work for almost 20 years. It’s the part of my career that taught me the most, and it’s also the part that is hardest to describe to someone who hasn’t done it. I’ve also had the privilege to train professionals at many of the top turnaround and restructuring firms across the country.
So for this prediction, I’m probably going to be biased…I’m just saying.
Restructuring is finance under conditions of uncertainty and crisis. A company is in trouble. Relationships are strained. Trust with lenders or private equity has eroded. The data is often unreliable. Decisions are high-stakes and sometimes irreversible.
You’re negotiating with lenders who are trying to protect their position, management teams who are trying to protect their jobs, and creditors who disagree about who deserves to get paid first. The stakes are higher than in most other situations and the room we’re in can be uncomfortable.
In my experimentation with AI — notably Copilot and Claude — I’ll give it this: it can build really good cash flow models. It can run the scenarios. It can synthesize financial data faster than most analysts.
What it cannot do is read the room and it struggles with nuance. It cannot sit across from a lender’s representative at 8pm and understand whether resistance is principled or positional. It cannot sense that a management team is understating an inventory valuation because they’re afraid, not dishonest. It cannot hold a negotiation together when there’s animosity between the parties.
Those are human skills built on situational awareness and accumulated experience. They don’t transfer from the immeasurable amounts of data that AI is trained on. The restructuring advisor who combines deep financial expertise with that kind of judgment is going to be in higher demand, not lower, as AI handles more of the technical work.
In addition, AI’s struggle with nuance means we have to teach it edge case after edge case and exception after exception. This is almost impossible to do, when every situation is different from the last. Just because AI can handle complex situations well, doesn’t mean we should go to it when the stakes are extremely high. Human involvement and judgment will likely outweigh the speed and simplicity of an AI engine.
As our world becomes more volatile, uncertain, complex, and ambiguous — what’s commonly called VUCA — those who remain calm and involved, in the face of nuance, are best positioned.
