AI & Accounting
November 18, 2026

AI Is a Magnifier, Not a Magic Fix: Paul Barnhurst on Risk, Context, and Finance Judgment

Use a red-yellow-green framework to decide which finance work AI can automate, where review is required, and what people still need to own.
Bohdan Sitarskyi
Product Lead and Head Account Manager
AI Is a Magnifier, Not a Magic Fix

Key Takeaways

  • AI magnifies whatever surrounds it: strong data and process improve the output, while weak inputs scale the error.
  • The right level of automation depends on the cost of being wrong, not on what the technology can technically do.
  • Finance teams can use a red-yellow-green framework to separate low-risk automation from work that needs review or human ownership.
  • AI can draft a budget or forecast, but it cannot know management’s unrecorded plans unless people supply that context.
  • As routine preparation gets faster, judgment, communication, and relationships become more valuable.

AI can draft a budget, summarize a variance, or build a first-pass model in seconds. That is useful. It is not the same as knowing whether the source data is reliable, which assumptions are missing, or what a decision means for the people involved.

The best way to think about AI in finance is as a magnifier. It makes good data, clear processes, and real expertise more useful. It can also make weak assumptions, poor controls, and bad information more damaging.

That was the central idea in Episode 3 of Prompt & Loss, a Veltrix AI podcast hosted by Bohdan Sitarskyi. Paul Barnhurst, known as The FP&A Guy, joined Bohdan to discuss budgeting, forecasting, automation, and why human judgment still matters when the tools become faster.

Paul is an FP&A professional, trainer, consultant, and speaker. He holds an MS in Information Management and an MBA in Finance, along with FPAC, FPAP, and AFM credentials. He is also a Microsoft MVP for Excel and shares finance education with an audience of more than 100,000 LinkedIn followers.

His message was not that finance teams should avoid AI. It was that they should use it deliberately, match automation to the cost of an error, and keep people responsible for decisions that require context or trust.

Watch or Listen to the Full Conversation

‍

Watch Paul and Bohdan discuss AI, budgeting, data quality, finance judgment, and the future of FP&A.

Watch the episode on YouTube, listen on Spotify, or listen on Apple Podcasts.

What does it mean to call AI a magnifier rather than a magic fix?

‍

AI does not repair a broken process simply because it produces an answer quickly.

Paul put it plainly:

AI is a magnifier. It’s not magic. It magnifies the good and the bad.

If a company has reliable records, clearly defined work, and people who understand the business, AI can help them move faster. It can organize data, draft analysis, find patterns, and make a strong process more efficient.

If the inputs are wrong or the process is unclear, the same speed works in the other direction. A flawed assumption can spread across a forecast. A bad mapping can affect every report. A confident summary can make an unreliable result look finished.

This is why a polished output is not enough. The first questions should be about the process around it: Where did the data come from? What assumptions were made? What was the tool asked to do? Who checked the result?

AI scales what is already there. The quality of the result still depends on the system, information, and expertise around it.

How should finance teams decide what to automate?

‍

The fact that a task can be automated does not mean it should run without review. The better question is: what happens if the answer is wrong?

Paul described a simple red-yellow-green framework for making that decision.

Green: low risk and easy to check

Green tasks have a limited downside if the first result is imperfect. The output is easy to review or correct. Drafting a meeting agenda, summarizing notes, or suggesting spreadsheet formulas can fit here.

Yellow: useful, but review is required

Yellow tasks can save meaningful time, but an error could affect a report, recommendation, or internal decision. Variance explanations, a first-pass forecast, management commentary, or anomaly detection often belong in this category.

AI can prepare the work. A finance professional should check the source data, assumptions, calculations, and business meaning before anyone relies on it.

Red: material consequences or human relationships

Red tasks involve a high cost of being wrong or depend heavily on trust, accountability, and context. Final investment decisions, board-level recommendations, sensitive employee conversations, and advice with legal or regulatory consequences need human ownership.

AI may support the analysis, but it should not become the decision-maker.

The framework is useful because it begins with consequence rather than enthusiasm. A finance team does not need one rule for every use of AI. It needs a clear level of review for each kind of work.

Why can’t AI own a budget or forecast?

‍

AI can help build a budget. It can bring historical data together, apply assumptions, identify a pattern, draft scenarios, and flag numbers that need attention. What it cannot do is know every plan that has not yet appeared in the records.

A budget is not just last year’s numbers adjusted by a percentage. It includes choices that management has not made public or documented in a system: a planned hire, a delayed launch, a pricing change, a new contract, an expected customer loss, or a decision to enter another market.

Paul used planned hiring to make the gap clear. Historical payroll data can show the current cost base. It cannot reveal three roles the company intends to add next quarter unless someone supplies that information.

The same problem appears in forecasts. The model may find a pattern in the past. A finance professional still has to ask whether the future will follow it.

AI can make the first version faster. People must provide the business plans, challenge the assumptions, and decide which scenario is realistic.

What context is usually missing from the data?

‍

Financial systems record what happened. They rarely capture every reason it happened or what management plans to do next.

A revenue decline might reflect weaker demand, a deliberate exit from an unprofitable product, a delayed shipment, or a sales team focused on a larger deal that has not closed. A higher cost might be waste, or it might be a planned investment.

The numbers alone do not settle the question.

Paul’s point is especially important for finance teams because their work sits between data and decisions. They need to understand operations, strategy, and the people behind the plan. They need to know when a trend is meaningful and when it is an artifact of timing, classification, or a one-time event.

This is also why data quality is more than a technical concern. Good data includes definitions, ownership, and enough business context for someone to interpret it correctly.

When AI works from an incomplete picture, it can still produce a complete-looking answer. Human review is what separates a useful draft from a conclusion the business can trust.

How does AI change the value of finance work?

‍

As routine preparation gets faster, the value of finance shifts toward judgment, communication, and influence.

A professional who spends less time gathering data can spend more time asking why a change occurred, explaining what it means, and helping management choose a response. The spreadsheet or report still matters, but it becomes the beginning of the conversation rather than the final product.

Paul described FP&A as part science and part art. The science includes models, formulas, data, and analysis. The art includes understanding the business, communicating clearly, and working with people who have different incentives and priorities.

AI is strong at parts of the science. It can help draft, summarize, compare, and calculate. It does not remove the need for the art. In many cases, it makes that part more visible because less time is spent on the mechanical preparation.

This does not mean every finance role becomes strategic overnight. Teams still need accurate reporting and disciplined processes. But professionals who learn to combine those foundations with business understanding will be better placed to use AI well.

What does this mean for accounting and advisory relationships?

‍

The same principle applies beyond an internal finance team.

An accountant can use AI to prepare a first review of a client’s records, compare periods, or find a change worth investigating. The output still needs someone who understands the client, checks the evidence, and decides whether the issue matters.

Clients may have access to many of the same tools. What they often do not have is the accountant’s experience, knowledge of their history, or ability to connect a financial result with a business decision.

That makes the relationship more important, not less. A useful adviser does not simply deliver more numbers. They help the client understand which number deserves attention, what information is still missing, and what to do next.

Veltrix supports that division of work. The Client Business Health Check can provide a fast first review of connected client data, while the accountant opens the supporting information, adds context, and leads the conversation. Firms can manage separate client workspaces through Veltrix for accountants.

The software helps with preparation. The accountant remains responsible for the interpretation and advice.

Where should a finance team start using AI?

‍

Start with one repeated task where the team already knows what a good result looks like.

Classify it using the red-yellow-green framework. Define the source data, the expected output, the review step, and the person responsible for the final result. Then test the complete workflow rather than judging only the speed of the first draft.

For a green task, the goal might be to save time on meeting preparation. For a yellow task, it might be to create a first variance explanation that a finance professional checks against the ledger and operational context. A red task should keep the final recommendation and decision with a person from the start.

Track what the AI missed, not only what it got right. Note where instructions were unclear, data was incomplete, or the reviewer had to supply context. Those gaps show what needs to improve before the same workflow is used more widely.

The point is not to automate the most impressive task first. It is to build evidence that the process is reliable.

Does better AI make finance work less human?

‍

It can make parts of finance less manual. That is not the same as making the work less human.

Budgets, forecasts, and analyses exist to support choices. Those choices involve tradeoffs, uncertainty, priorities, and people. Someone still has to explain the result, challenge the assumptions, and take responsibility for the recommendation.

Paul’s magnifier is a useful test for any new AI workflow. If the process is strong, the data is reliable, and the review is clear, AI can make good work faster. If those foundations are weak, automation can spread the problem before anyone notices.

The practical path is to automate according to risk, keep people close to material decisions, and use the time saved for better questions and clearer conversations.

That is not a limit on AI. It is how finance teams turn speed into work the business can trust.

Where can you learn more from Paul Barnhurst?

‍

Visit The FP&A Guy for Paul’s training, resources, and finance content, or follow The FP&A Guy on YouTube.

You can also watch the full episode on YouTube, listen on Spotify, or listen on Apple Podcasts.

Accounting professionals can also explore Veltrix for accountants or run a free Client Business Health Check.

Find what’s worth a client conversation

Use Veltrix to spot changes, investigate what caused them, and turn client data into advisory opportunities.