Deploying AI at Your Company
Welcome to issue #017 of New Age Accounting. Most teams want to start. Few know how. Here's the framework that actually works.
Everyone knows AI is important.
Most companies have made that declaration at an all-hands or a board meeting. Leadership is bought in. The budget conversation has started. The pressure to do something is real.
The harder question — the one nobody has a clean answer to — is where do you actually start?
The teams getting it right aren’t the ones with the biggest budgets or the most technical talent. They’re the ones who started small, built deliberately, and didn’t try to automate everything at once.
Here’s what that looks like in practice.
The three mistakes most companies make
Before the framework, it’s worth talking about what gets in the way. Because most companies don’t fail at deploying AI because they didn’t care enough. They fail because they started wrong.
Mistake 1 — No clear direction
They know AI is important so they start throwing things at the wall. A tool here, a pilot there, a vendor demo every other week. Nothing connects. Nothing compounds. Six months in they have a collection of subscriptions and no system.
The energy is there. The intent is there. The direction isn’t.
Before you touch a single tool, define what you’re trying to produce. What does success look like in 90 days? What workflow do you want to stop doing manually? What does your team spend hours on every month that a system should be handling?
Answer those questions first. Everything else follows.
Mistake 2 — Biting off more than you can chew
They want every bell and whistle. The most advanced tools. The most ambitious use cases. They overspend, underdeliver, and lose momentum before anything is actually working.
The mistake isn’t ambition. It’s sequencing. The basics done well beat the advanced done poorly every single time. One workflow running cleanly is worth more than ten half-built automations that nobody trusts.
Get one thing right. Then build the next one.
Mistake 3 — Automating before the foundation is solid
This is the most common and the most costly.
They try to build automations on top of broken processes. They vibe code solutions when tools already exist for exactly what they need. They skip the step where they leverage what they already have — and end up duplicating effort, creating new problems, and spending money they didn’t need to spend.
Before you build anything new, look at what you already have. Your ERP, your spend management platform, your communication tools — most of them already have AI capabilities you haven’t touched. That’s your starting point. Not a new tool. Not a custom build. What’s already in your stack.
Use what exists before you build what doesn’t.
How to actually deploy AI
Once you’ve avoided the common traps, the path forward is more straightforward than most people expect. Here’s the framework.
Step 1 — Find your LLM and commit to it
Pick one. Claude, ChatGPT, Gemini — it matters less than you think which one you choose. What matters is committing to one and learning it deeply. The teams that switch constantly never build real proficiency. The ones that go deep on one tool get dramatically better outputs than the ones who dabble in five.
Pick one. Go deep.
Step 2 — Identify your repetitive tasks
Make a list. Everything your team does that is rule-based, high-volume, and repeatable. The month-end close. The reconciliations. The variance commentary. The reporting packages. The data pulls.
These are your automation candidates. Write them all down before you prioritize any of them.
Step 3 — Rank by impact and ease
Not everything on that list is worth automating first. Rank each item by two factors: how much time does it take and how straightforward is it to automate.
The sweet spot is high time cost, low complexity. Those are your first builds. They produce the fastest wins, build team confidence, and prove the concept to anyone who needs convincing.
Step 4 — Build one thing at a time
Take the top item on your list and build it properly. Define exactly what the output should look like. Write the instructions. Test it. Refine it. Get it working before you move to the next one.
The temptation is to build everything at once. Resist it. One thing built well is the foundation for everything that follows.
Step 5 — Track what you’re spending
AI tools have real costs. As you deploy more of them those costs add up faster than most teams expect. Track what you’re spending, what you’re getting, and where the ROI is actually coming from.
This isn’t optional. It’s how you justify continued investment, identify what to cut, and have an honest conversation with leadership about what’s working and what isn’t. Treat your AI spend the same way you’d treat any other line item — with scrutiny and accountability.
Step 6 — Expand deliberately
Once the first workflow is working, repeat the process. Another workflow. Another build. Another refinement cycle.
The teams that get this right compound every month. Each build makes the next one faster and easier. The knowledge carries forward. The confidence grows. What felt overwhelming at the beginning starts to feel like just how work gets done.
That’s the goal. Not a one-time implementation. A continuous build cycle that makes the team more capable over time.
What this looks like in practice
Here’s a real example from a lean finance team.
The starting point was one painful monthly workflow — cash analysis. Hours of manual work every month. Downloading bank statements. Building pivot tables. Reconciling accounts. Writing commentary.
Instead of trying to automate the entire finance function at once, the team started there. Defined the output. Connected the data sources. Built the workflow in Claude. Tested it. Refined it through multiple iterations until the output was trustworthy.
That one workflow is now automated. The hours it used to take are now minutes. And that success became the template for the next build, and the one after that.
The stack didn’t get built in a day. It got built one workflow at a time, each one compounding on the last.
That’s the model.
A note on AI spend
As you deploy more tools and build more workflows, your AI spend will grow. That’s expected and worth it — but only if you’re tracking it.
Build a simple log. Every tool you’re paying for, what it does, what workflow it supports, and whether that workflow is actually producing the outcome you expected. Review it quarterly. Cut what isn’t working. Double down on what is.
The companies that get in trouble with AI spend aren’t the ones who invested too much. They’re the ones who invested without tracking. Don’t be that team.
Building for the future
Deploying AI at your company doesn’t require a massive budget, a technical team, or a six-month implementation project.
It requires clarity on where to start. Discipline to start small. Patience to build one thing properly before moving to the next. And the willingness to track what’s working honestly.
The companies that figure this out aren’t the ones with the most resources. They’re the ones that started — and kept building.
Start this week. One workflow. One build.
See what’s possible.
Here’s my question to you:
What’s the first workflow you’d automate if you knew exactly how to do it? Drop it in the comments — I read every one.
The purpose of New Age Accounting is simple: to empower accountants — at every level — to become builders, not bookkeepers. Whether you’re a staff accountant, a controller, or a CFO, there’s something here for you. Some topics will be high level, others will come with step-by-step guides, and some will include the exact prompts and tools you need to start building today.
If you’ve made it this far, you’re already thinking differently about this profession. Subscribe and come build with us.