I wrote a version of this article eighteen months ago.
The argument back then was simple. The question of AI or business partnering was the wrong question. The right one was AI and business partnering. Two things you needed to manage together, not pick between.
Eighteen months on, the question keeps coming up. Different audience, different room, same question.
Andrew, where should we be putting our investment? On the AI side or the people side?
The answer hasn’t changed. It’s still both. But the technology has changed enough that I think the question deserves a fresh answer, because what we now mean by “AI” in a finance context is not what we meant in late 2024.
In late 2024, AI in finance mostly meant ChatGPT. An assistant. You typed a question, it gave you a draft, and you tidied it up. The decision about what to do with the output was still yours.
In 2026, AI in finance increasingly means agents. Autonomous systems that don’t respond, but execute. They pull data, run analyses, draft commentary, post journals, flag exceptions, and, in some cases, initiate transactions. The work doesn’t sit in a chat window waiting for you. It runs in the background and arrives at your desk already done.
That shift sharpens the AI-versus-business-partnering question in a way it hasn’t been sharpened before. Because if AI is no longer suggesting – but doing – the question of what’s left for the human finance professional gets more pointed.
Let me make the argument for why the answer is still both. And why the polarity hasn’t dissolved, it’s just got sharper.
The Agentic AI Shift
The phrase you’ll hear most often in CFO circles right now is agentic AI. The World Economic Forum’s framing is the cleanest one I’ve come across. Agentic AI goes beyond generative AI by enabling autonomous decision-making, collaboration, and learning, without constant human guidance.
In plain English: a generative AI assistant waits for you to ask. An agent doesn’t.
The deployments are real now, not theoretical. WEF reporting from Davos 2026 noted Goldman Sachs is developing autonomous agents to handle trade accounting and client onboarding, and Lloyds Banking Group expects enterprise-wide deployment of agentic AI to add £100 million in value in 2026 by automating fraud investigation and complex complaints. Embedded copilots are now standard in SAP, Oracle, Workday, and Microsoft’s finance stack.
So the technology has moved on. Two years ago, AI was the smart graduate who could draft your commentary. Today, it’s a small team of graduates who can also do the variance pulls, populate the deck, run the sensitivity analysis, and have it on your desk before the morning meeting.
Which poses the question: If a team of autonomous agents can run the close, build the forecast, and write the commentary, where does the finance business partner add value?
This is where most of the noise in the market gets the answer wrong.
Why Agentic AI Doesn’t Resolve the Polarity
The argument I keep hearing is some version of, with agents now in the picture, the human side matters less. If AI can execute, not just advise, then the workflow has fewer human touch points. So the value of the human shrinks.
It’s the wrong read for two reasons.
The first is that agents are still only as good as the information and the context you give them. They can run the variance analysis brilliantly. What they don’t know is that the variance in Region B was caused by a competitor pulling out of the market. Why? Because that conversation happened in a coffee with the sales director on Tuesday.
It doesn’t know about the working capital covenant that the bank just renegotiated. Or that the operations manager is under-resourcing the line that produces your most profitable SKU. The context that makes commercial commentary commercial sits in human relationships and informal conversation, not in your ERP.
I covered this in detail in my post on AI and data quality. But the short version is this. Faster, cleaner, more autonomous AI sitting on data that doesn’t capture the commercial reality of the business will only give you faster, cleaner, and more confident wrong answers.
The second reason is more important.
When AI agents handle the production work, such as the variance reports, the commentary, and the forecast updates, the value of finance shifts even further toward what AI cannot do. Sit in a room with an operations leader and a CEO. Explain the financial impact in clear, practical terms. Push back constructively on a sales director who wants to drop a price. Get a board to action a recommendation. Build the trust that makes someone bring you into a decision before they make it, not afterwards.
None of that is what agents do. All of that is what business partnering is.
So the polarity hasn’t softened. It’s the opposite. The faster and more autonomous the AI side becomes, the more the human side has to be deliberately, visibly, demonstrably about the work AI cannot do. The teams that don’t make that pivot don’t become irrelevant gradually. They become commodity work overnight.
That risk is the subject of finance’s AI wake-up call. The danger is not that AI replaces the finance team. It’s that the business stops bringing the interesting decisions to a finance team that hasn’t visibly evolved its role.
What AI-Involved Finance Business Partnering Looks Like in 2026
Let me make this concrete, because too much of the writing on this is abstract.
A business partner working in an agentic-AI-enabled finance team in 2026 looks something like this.
They walk in. The agent has already produced the variance analysis for the month, drafted the commentary in three different lengths (one-pager for the CEO, half-pager for the operations leader, two lines for Slack), flagged the three accounts that look unusual relative to forecast, and pre-populated the commercial review deck. That work used to take two days. Now it’s done before the kettle boils.
The business partner spends ten minutes reviewing it. They check the drivers the AI has attributed to each variance (the discipline we cover in is your AI hallucinating), because plausible-sounding AI commentary can attribute variances to drivers that don’t exist in the data. They cross-reference against what they know from last Friday’s commercial review. They spot that the agent has missed something about a customer’s payment terms that sales mentioned in passing.
Then they go and have three conversations:
- One with the operations manager about why a key SKU is underperforming
- One with the regional sales lead about whether the customer issue is structural or one-off; and
- One with the CFO about whether the forecast needs adjusting in two months’ time.
They come back, refine the recommendation, and walk into the executive meeting at 2 pm with a sharp commercial view rather than a finance report.
That is what AI-enabled business partnering looks like in 2026.
The agent did the production work. The human did the work that changes what the business does next.
Neither alone is enough. Without the agent, the business partner is buried in the production work and doesn’t get to the three conversations. Without the business partner, the agent’s outputs are technically correct, beautifully formatted, and disconnected from action.
It is, and remains, an and.
A Quick Diagnostic for Finance Leaders
If you’re a CFO or head of FP&A, here’s the honest test for where your team sits today.
When AI saves time on a production task on your team, where does that time go?
If the answer is “we deliver more reports, faster”, you have a problem. You’re using AI to be a more efficient version of the team you already were. The risk is that the function gets smaller in the next restructure, because nobody can articulate what the team is doing with the saved capacity that the business actually wanted.
If the answer is “the team has more conversations with operations, sales, and the exec”, you’re using AI properly. You’re letting the technology do what it does well, and reallocating human time to where humans still beat machines.
If the answer is “I don’t know yet because we haven’t measured it”, get measuring. It’s the single most important productivity metric for a modern finance function, and almost nobody tracks it.
What to Do If You’re Heavy on One Side
I see two failure modes in the teams I work with.
Heavy on tech, light on partnering.
The team has rolled out Copilot for Finance, integrated AI into the ERP, and run pilots with agentic workflows. The reporting is faster, the commentary is cleaner, and the forecast cycle has shrunk. But the team is still mostly behind their desks, and the business goes elsewhere for the conversations that change decisions.
Symptom: the function looks impressive in board slides about AI adoption, but business unit leaders don’t book time with finance the way they used to. The fix is making partnering capability explicit, deliberate, and measurable. We cover the strategic argument for this in why AI is not the most important thing.
Heavy on partnering, light on tech.
The team has strong relationships with the business and is in the right rooms. But they’re slow, they’re stretched, and they’re producing the analysis the same way they did five years ago.
Symptom: senior partners with great commercial instincts are spending half their week on production work that agentic AI could do in 30 minutes. The fix is curiosity about the technology. What it can take off your plate, what it cannot, and how to plug it in without breaking trust in the numbers.
For a structured starting point on both sides, our guide to using AI in finance walks through where to start as an individual and as a function.
The Polarity, Sharper Than Ever
The original argument I made in 2024 was that AI and business partnering are a polarity, not a competition. Two things you need to manage in balance. Lean too far one way, and the function gets brittle.
The agentic era hasn’t dissolved that polarity. It pulled the two poles further apart.
The technology side is now genuinely powerful and in production at the biggest financial institutions in the world. The human side, in response, has to be genuinely good at the things that aren’t automatable. Context, judgement, conversation, influence, the ability to walk into a room and change what gets decided.
If your function is investing in agentic AI but not investing in business partnering capability, you are building a faster way to be irrelevant.
If your function is investing in business partnering but ignoring the agentic shift, you are building an artisan team that can’t keep up with the pace the business needs.
You need both. The CFOs and finance leaders I work with who are getting this right are running parallel investments in the technology stack and in their people’s commercial capability.
When I started building my finance business partner training programs, I kept coming back to one thought:
Once the robots arrive, the only thing left will be business partnering.
That feels even more true now.
The robots have arrived. So the question for your team is no longer whether to invest in business partnering. It’s how fast you can get the investment moving, because the technology side is already running.