When I started hearing that AI was going to replace accountants, I’ll be honest, it scared me.
I was afraid of the unknown, and of it replacing accountants like everyone was saying. I run a finance firm. I’ve built a team. I’ve spent years developing processes, training people, and building something that really works for the companies we serve. The idea that a tool could make all of that irrelevant wasn’t something I took lightly.
But I’ve spent the last few years inside this. Building AI workflows at Cypher, testing agentic AI, having conversations with other accounting firms about what they’re doing, and watching what actually happens when growth-stage companies try to automate their finance function without the right foundation underneath it.
What I’ve seen has changed how I think about all of it.
TLDR
- Finance teams that understand AI will outperform the ones that don’t.
- Process and documentation have to come before automation.
- Agentic AI running without human oversight produces dangerous errors that compound until month-end close.
- The value of a finance team lives in the decisions it drives, not the data it produces.
- AI is a tool that makes good finance teams more effective, and exposes weak foundations faster.
- Spreadsheets just got a lot more powerful.
The Profession Has Adapted to Every Shift. This One Is No Different.
Spreadsheets came in the 80s and moved accountants away from manual ledger work and toward analysis. Cloud platforms came in the 2000s and moved finance teams away from desktop software and toward real-time reporting.
Each shift elevated the profession. AI is doing the same thing, just faster and bigger.
And here’s something I keep having to say out loud: I hear “spreadsheets are dead” constantly, and every time I do, I think about how Copilot in Excel just let a non-technical user build a pivot table in thirty seconds by asking for it in plain English.
The spreadsheet became the interface for AI, not the thing AI replaced. For anything where exact numbers carry real accountability, financial models, forecasts, pricing, people still need a model they can interrogate and audit.
AI hallucinates, and spreadsheets are deterministic.
The accounting firms that are building AI workflows now, training their teams to work alongside the tools, and using automation to handle volume, are the ones that will serve more clients at a higher quality in five years.
The ones skipping the foundational work and handing AI the keys without oversight are the ones producing errors at scale.
What AI Actually Does Well in Finance
I want to be specific here, because the conversation around AI in accounting tends to be either fear or hype, and neither is useful to a CEO trying to make a real decision.
There are real tasks where AI adds genuine value today:
- Vendor coding in accounts payable, particularly when you have clean historical data to train it on
- Transaction categorization and reconciliation at volume
- Payroll automation, once the underlying setup is correct
- Expense management templates
- KPI dashboards, though there are many steps that have to happen before you get one that is automated and well thought through
Every item on that list has something in common. It requires a well-documented process, clean underlying data, and a human who understands the business well enough to know when the output is wrong.
AI handles pattern recognition well, but a human still has to build the pattern.
Process Has to Come Before Automation
This is the part I find myself repeating most often to leadership, and it’s the part that gets skipped most often.
I had a CEO come to us after trying to automate their entire finance function before their workflows were documented. No SOPs. No clear process for how information flowed between systems. No technical expertise documentation. No one had thought through what the data needed to look like before it got handed to a tool.
It cost them more time and money than doing it manually would have.
I asked them to take a step back. We reworked the walkthroughs, shaped up the processes, cleaned up the underlying data and the flow of information between each system, and then introduced AI on top of that foundation and the right team to manage it and maintain it. And it worked.
That experience is not unique; I see a version of it constantly in growth-stage companies.
Founders and CEOs often don’t find out about the errors until much later, sometimes not until an investor asks a question they can’t answer, or until month-end close surfaces a reconciliation problem that traces back months.
The question I use to help them figure out whether they have an automation opportunity or a process-and-data-quality problem is: “Can you walk me through exactly how this works today, step by step?”
A vague answer means the process problem comes first. Fix that before you introduce any automation.
Agentic AI Still Needs a Finance Team Behind It
A founder recently asked me: “What are you going to do that Claude isn’t already doing?”
It’s a fair question.
Agentic AI, meaning AI that is set up to run a process end-to-end without human oversight, produces errors when it runs unsupervised. Errors accumulate over time: until month-end close, until you’re in a board meeting trying to explain a figure that doesn’t reconcile, or until you’re in due diligence and someone asks for clean financials you can’t produce.
To set up automation that actually works, you need:
- Templates built and validated by finance professionals
- FP&A logic developed by someone who understands the business model
- The right coding and AI layer on top of that, once the foundation is solid
- A finance team managing the output, catching exceptions, and making the judgment calls the tool can’t make
The final outcome of a finance team is decisions, not data. A good finance team analyzes what’s in the reports and tells leadership what to do about it. That’s what AI, running on its own, can’t deliver.
The thinking still has to come from somewhere.
Journal entries, accruals, final closings, vendor changes, and anything requiring professional judgment needs human review before it’s posted. AI makes the need for strong SOPs and internal controls more important, not less.
Junior Finance Talent Has a Real Role in an AI-Enabled Team
I also want to address something I hear from people inside finance functions, not just from leadership.
Junior accountants coming into the profession right now are going to be among the most valuable people in the room.
Even when you automate workflows, someone needs to maintain those workflows and manage them. The people entering finance right now are coming in with the technical fluency to do that naturally. There’s also real creativity involved in customizing AI workflows, rewriting the logic, tweaking the rules, and making sure the output reflects the business it’s serving. That’s skilled work.
What I’m focused on at Cypher is making sure everyone on the team, including every person touching our client workflows, understands AI well enough to use it well. The firms that win in the next ten years will be the ones where every team member works alongside these tools, not around them.
The Differentiator Is the Thinking Behind the Tool, Not the Tool Itself
Every accounting firm has access to the same AI platforms, the same automation capabilities, and the same software.
The expertise to train those tools correctly, the process discipline to build the right foundation underneath them, and the judgment to catch what they miss, that’s what separates a finance partner from a software subscription.
This is something Cypher has always believed. We’ve built our entire model around customizing workflows to each client’s business, not forcing them into someone else’s process. Cloud and SaaS tools largely asked you to conform to their structure.
AI flips that, enabling customization at a level that wasn’t possible before, which means the tailored processes we’ve always built for our clients can now be maintained, automated, and scaled without losing what makes them specific to each business.
Documenting those processes, understanding what the business needs, and introducing AI in a way that supports customized workflows, that’s the work.
The tool is the last step, not the first.
Companies choose the finance partners they work with because of how they think, how they implement, and how they handle the complexity that tools can’t handle alone.
At Cypher, AI is helping us serve more clients at a higher quality. The expertise behind the work is still ours.
Answering the Internet’s Top FAQs
Will AI replace accountants?
Finance teams that understand AI and build with it will outperform the ones that don’t. The shift is away from low-value, repetitive tasks and toward strategy, analysis, and decision support. Spreadsheets and cloud platforms made the same shift, and the profession was improved for the better both times.
What finance workflows are ready for automation today?
Vendor coding in AP, transaction categorization at volume, payroll processing inside well-configured tools, and expense management templates are all real opportunities today. KPI dashboards are possible but require significant foundational work before the automation is meaningful. Start with the workflows where you have clean historical data, documented processes, and someone who can validate the output.
How do I know if I have an automation opportunity or a process problem?
Walk through the workflow step by step. If the answer is vague or inconsistent, the process problem comes first. Automation built on an undocumented or inconsistent process produces errors faster, and those errors are harder to trace.
What needs human review in an AI-enabled finance function?
Journal entries, accruals, final closings, vendor changes, and anything requiring professional judgment. These should always be reviewed by a human before being posted. The audit trail still matters. Controls still matter. AI increases the need for strong SOPs, but doesn’t replace them.
Are spreadsheets becoming obsolete?
No, and I’d push back hard on anyone saying otherwise. AI is making spreadsheets more powerful, not replacing them. Tools like Copilot in Excel and Gemini in Sheets mean a non-technical user can now write complex formulas, generate pivot tables, and clean data just by describing what they need. The spreadsheet becomes the interface for AI. And for financial models, forecasts, and pricing, where exact numbers carry real accountability, spreadsheets remain the standard. A CFO signing off on a budget needs a model they can interrogate and audit. AI alone can’t give you that
What do growth-stage companies need from a finance partner when it comes to AI?
A finance partner that is genuinely building with AI, not just talking about it, and that can help you think through your processes and workflows before introducing automation. The approach has to be customized to your business model. And the team managing it still needs the finance expertise to know what good output looks like.
The Finance Function That Scales
The companies building the strongest finance functions right now are the ones treating AI as a layer on top of a solid foundation.
This foundation is still the processes, the data quality, the reporting structure, and the judgment. AI makes a strong finance function more efficient and exposes a weak one faster.
If you’re building a finance function that scales, uses AI well, and actually supports the decisions you need to make as a CEO, book a call and let’s talk about what that looks like for your business.
