Click for Takeaways
- 40 hours upfront saves 100 hours later. If you’re doing the same thing every week, every month, every year — it can be automated. The math almost always works out. Most people just haven’t done it yet.
- The accountant of the future is a data wrangler. Bad data costs organizations an average of $12.9 million a year — not because of carelessness, but because operations are messy and the cleanup lands on accounting. That invisible upstream work is where the profession is heading.
- Do it the broken way first. Before you redesign any process, run it as-is for two full cycles. Assume nothing. Ask all the questions that sound stupid. The reason it’s clunky usually matters.
- AI is a great starting point. It’s a terrible solver. Manual data entry is still the top pain point for AP teams — cited by more than a third of practitioners. AI and automation can remove it, but only if you’re the one who knows how to design the replacement.
- Bad data structure is the real blocker. Manual invoice entry has already dropped significantly in just the last two years, and it’s accelerating. The technology isn’t holding back most finance teams that can’t implement AI — they’re being held back by inconsistent, unstructured data upstream.
Based on an interview with Emily Feinstone, Accounting Manager at Aventus Advisory Group
Finance teams live downstream. They receive journal entries, run variance analysis, build forecasts, and present to leadership — but they rarely see what it took to turn raw operational data into something clean enough to enter the general ledger. That upstream work is where Emily Feinstone operates.
Emily is an accounting manager at Aventus Advisory Group with nearly 20 years of experience in finance and accounting operations. She is, by her own description, more data wrangler than traditional accountant. She builds automation workflows, cleans high-volume messy data, and redesigns clunky inherited processes, including the month end close process, into lean, reliable systems. And rather than taking the traditional CPA route, she’s pursuing a degree in data science — a deliberate bet on where the profession is heading.
In a conversation with FP&A Today host Glenn Hopper, Emily talked about how she got into process automation before it had a name, what her standard toolkit looks like, how she thinks about controls and auditability, and why data structure — not AI itself — is the real bottleneck holding most finance teams back.
The Origin Story: Union Reports and One Excel Course
Emily’s entry into process automation was not planned. She was working in her family business, had grown up office-adjacent, and got thrown into producing union benefits reports for five unions running three different shifts.
“After the first two times I did it by paper, I said, ‘There has to be a better way.’ I think I had one Excel course in my whole entire life, and I somehow built this Excel template that I could reuse every week.”
That was 2006. The instinct — spend time once to save time indefinitely — has defined her approach ever since.
The arithmetic is straightforward, even when the investment feels large.
“40 hours upfront could be the difference of saving 100 hours the rest of the year. It’s a no-brainer for me.”
The constraint she keeps running into is that most people are either unaware of what’s possible or reluctant to invest the upfront effort. They tolerate the repetition because they don’t realize it can be removed.
What “Data Wrangling” Actually Means in Practice
When Emily describes herself as a data wrangler rather than a traditional accountant, she’s describing something finance teams rarely think about: the work that happens before the journal entry exists.
“You have a client and they have all of these systems that they operate in and all of the data that comes out of those systems can be kind of messy. So if you have to transform the same data the same way every month, you are wrangling data. You’re not just making a journal entry — you’re saying, does all the data make sense? Is everybody using the boxes correctly?”
Operations are messy. Salespeople get lumped into a single column with slashes and dashes. Departments use fields inconsistently. Systems that weren’t designed to talk to each other produce exports that require significant translation before they’re usable. That translation lands on the accounting team.
Poor data quality costs organizations an average of $12.9 million per year, according to Gartner — a problem that the best financial analysis software is specifically designed to prevent by governing data before it reaches the finance team. For accounting teams operating on tight close cycles, that cost is often invisible but constant.
“If you guys actually had to see the actual data before it turned into the journal entry, you would be astonished. Sometimes I feel like a magician.”
The Two-Month Rule: Don’t Fix What You Don’t Understand
When Emily inherits a broken process, her first instinct is not to fix it. It’s to run it.
“My approach is: for two months, I do it the way it is. I know I’m inheriting a clunky process, but I don’t know why the process is clunky. On paper, it could just look clunky and I could assume why, but I’ve learned that I assume nothing — because I could assume X and there’s so many variables about why it was done that way that we might actually need to keep that part of the process and build around it.”
This principle extends to the people running the process. A new person arriving and immediately declaring something broken creates friction. The people who built the workflow usually had reasons — constraints, edge cases, legacy system quirks — that aren’t visible from the outside.
“It’s more than just going and saying, ‘This is broken, this is how we’re gonna fix it.’ It’s almost getting to know the people, the why, how it got designed this way, how it evolved into this. You learn more from that.”
Only after understanding the system from the inside does she begin redesigning it, a discipline that sits at the core of FP&A best practices for process improvement.
A Concrete Example: Invoicing in Two Hours Instead of Two Days
The clearest illustration of her approach is a recent invoicing process she inherited. The client was tracking commissions with multiple salespeople crammed into a single column — names, percentages, and dollar amounts all in one cell, separated by slashes and dashes. Mistakes were common, and the process consumed two full days every month.
Emily ran the broken process for two months, asked all the questions that seemed obvious (and some that didn’t), then rebuilt the structure from scratch: breaking out each variable into its own column, reformatting all historical data, and creating a single source of truth that feeds directly into her QuickBooks Online import tool, her subcontractor payment tracking, and her commissions reports.
“Invoicing used to take two days. It takes me maybe two hours on a bad day now. We spend our time now making sure we got it right instead of going back and forth. We have time for insight rather than just preparation.”
The system didn’t require an ERP upgrade or a new platform. It required someone willing to understand the process before touching it.
The Standard Toolkit: Power Query, Amalgam, and Power Automate
For most of her client work — all currently on QuickBooks Online — Emily’s stack is compact and consistent.
Amalgam is an Excel plugin that creates a bidirectional connection with QuickBooks: download GL detail, push journal entries, run bank recs and account reconciliations, all from inside a familiar spreadsheet. What used to take 15 minutes of toggling between systems now happens in one place.
Power Query handles everything Amalgam doesn’t: normalizing data from multiple sources, formatting payroll reports into journal entries, automating repetitive transformation steps. It’s the data analysis tool in Excel she recommends most emphatically for anyone doing repetitive finance work.
“Even if it’s the first time you’re hearing about Power Query — if you have repetitive processes, repetitive data you’re cleaning every week, definitely look into the tool. It will be a game changer for your life and for the people you report to.”
Power Automate (part of the Microsoft Power Platform) fills the finance AI workflow layer: managing emails, saving files, triggering actions that don’t require human attention.
The bar for when to move up to a full ERP is simple: capacity.
“I always look at that from a capacity standpoint. If I can still do all of that work and take on other client work, I don’t think we’re at a breaking point yet. I start missing deadlines — that’s the breaking point.”
Controls: Build Them In, Don’t Add Them Later
Automation without controls is just faster failure. Emily’s approach is to embed checks at every stage of a workflow rather than relying on a final review to catch problems.
The principle: every tab should have a control on that tab. If something breaks — a missing GL account, an unexpected value — it should turn red before it propagates downstream. The cover sheet becomes a dashboard: all green means go.
“You can take what your cash output is supposed to be and tie it to your GL account inputs every single time. If something turns red, you know you have a problem. Go check that tab.”
This matters more as automation gets more sophisticated. Manual data entry still tops AP pain points, with 37% of AP professionals citing it as their biggest challenge — but the replacement for manual entry only creates confidence if the controls structure tells you immediately when something’s wrong.
On process documentation: every automation she introduces comes with written step-by-step instructions. Not for her benefit, but for everyone else’s.
“Step one is: put the data where it belongs. Then you refresh the data source and it cleans it, it does all these steps. It’s very important that as you introduce these things in your job, you clearly spell out what you’re doing so people aren’t scared of it. That’s half the battle of my life right now.”
How Emily Actually Uses AI
Emily uses AI applications in finance as a starting point, not a finisher. The distinction matters.
“If I’m ever stuck at where to begin a project, I go to AI. I say: this is what I’m trying to achieve, these are my inputs, these are the variables — where should I start? It always has a great starting point, and then you can build your own ideas off of that.”
For code generation — particularly Power Query M code — AI tools for Excel are excellent. She gives it a description of her columns and the transformation she wants, and it produces a working starting point in minutes. Tasks that would have taken her two days come in at two and a half hours.
What she doesn’t use it for: doing the accounting.
“I definitely learned the hard way — going to school and trying to have it help me with calculus. It’s wrong sometimes. It just calculates things wrong. So I’m definitely not going to trust it to do my accounting journal entries for me.”
On data privacy: she never puts client data into ChatGPT. If she needs help building a template, she provides the column headers and describes the structure — not the actual records.
“I give it the headers of both systems and say, ‘I would like you to make me a spreadsheet where I can find the differences between the two systems.’ It gave me a spreadsheet where I could paste the raw data and it told me the difference between the two systems, impeccably.”
The address normalization that followed — avenue vs. ave., state names vs. abbreviations — still required Power Query and a manual final pass. But the project that would have taken two full days took under three hours.
The Future of the Profession
Emily chose data science over the CPA for a simple reason: it aligns with what she’s passionate about and where the work is going. She’s not dismissing the CPA — she leans on CPAs every day and works closely with controllers — but for the operational, systems-oriented work she does, the data science credential is the more useful investment.
The broader shift she’s watching: manual invoice entry into ERP and accounting systems has already dropped from 85% to 60% since 2023, and the trajectory only continues. The question for accountants isn’t whether automation will change the work — it’s whether they’ll be the ones designing it or displaced by it.
“Think about accounts payable. Where do you think it’s gonna be in 15 years? With Ramp and Bill.com and just the stuff QuickBooks Online is implementing — once the workflow is complete and the computer’s pushing the workflow, do you need the accountant to push the workflow anymore? So what is the accountant’s role in accounts payable?”
Her answer is the same one that’s kept her relevant for 20 years: be the one who designs the finance AI workflows, not the one who runs them manually.
“I want to be the one that teaches the AI, not the one that is replaced by it.”
The 90-Day Roadmap for Finance Professionals
For anyone who wants to move in this direction without going back to school full time, Emily’s starting point is deliberate and free.
First stop: Microsoft Learn. It’s the platform where she developed her own understanding of the Power Platform and automation tools, and it reorients the way you think — from accounting-first to process-first.
Second: Power Query. She has mentioned it throughout the conversation because she means it.
“If you don’t know Power Query, learn Power Query. It will be a game changer for your life and for the people you report to. If you’re giving the same clean data every time, they can create actionable insights off that data.”
The deeper point is structural:
“Most companies can’t implement AI or month end close automation because the data structure isn’t there. Some of the boxes have inconsistent information all the way down the columns. With tools like Power Query and proper financial consolidation tools, we can clean that up and put it into the financial systems in a responsible way that makes sense – where you can generate powerful insights off of it. And that’s where everybody wins.”
Where Datarails Fits In
The work Emily describes — wrangling messy inputs, building reliable transformation workflows, creating a single source of truth — is what makes the downstream finance function possible. But even with clean data flowing in, finance teams still need a platform that can consolidate it, analyze it, and surface the story inside it without requiring everyone to start from scratch.
Datarails is the AI-powered FP&A platform built for Excel users. It connects financial data from ERPs, accounting systems, and spreadsheets into a governed source of truth — giving finance teams real-time visibility, dynamic scenario planning, and AI-generated narratives without abandoning the Excel workflows they already trust.
When the data upstream is clean and the platform downstream is intelligent, the finance team stops spending all month translating systems and starts spending that time on the insight that actually moves the business.
To learn more, visit datarails.com.
About Emily Feinstone
Accounting Manager at Aventus Advisory Group, where she specializes in finance process automation, data normalization, and operational accounting for small and mid-sized businesses. Emily brings nearly 20 years of experience in accounting and finance operations, with a hands-on approach to turning clunky, manual workflows into clean, repeatable systems. She is currently pursuing a degree in data science.Connect with Emily on LinkedIn.
FAQs
If you’re doing the same transformation with the same data in the same format every week or every month, it can almost certainly be automated. The test is simple: if you had to explain every step of what you do, and none of those steps involve actual judgment, the steps without judgment should be handled by a tool.
Never put actual client records into a public AI tool. Instead, share the structure: column headers, the shape of the data, and what you’re trying to achieve. AI can write Power Query code, design comparison templates, or generate starting-point logic based entirely on structure — no real data required.
Power Query, according to Emily, is one of the most underused data analysis tools in Excel, and the process-oriented thinking that comes with learning it. The Microsoft Learn platform is free and will shift how you approach repetitive tasks. The goal isn’t to become a data engineer. It’s to understand enough about data structure that you can clean it, standardize it, and make it usable for the people who depend on it.
It depends on the work. For audit, compliance, and public company finance, the CPA remains important. For operational and systems-oriented accounting, the value of a data science background or automation skillset may be higher, and the day-to-day work will reflect that more than the credential will.
