“AI in finance” doesn’t have to mean magic dashboards or a six-month project. Most wins come from fixing the everyday friction: getting a straight answer quickly, closing the loop on a task, and keeping everyone on the same page without flying 30 spreadsheets around.
Below is a simple playbook you can run this quarter. It’s written for real teams with real deadlines, not labs. We mention Simon once because it’s the tool we know best, but you can apply the ideas with whatever stack you have.
1) Start with the questions people actually ask
Before you talk models or vendors, list 10 questions your team asks every week. Keep them specific:
- “Why did Opex exceed budget in July—by department and vendor?”
- “Which customers are >60 days overdue and highest impact?”
- “Which SKUs will stock out in <14 days based on recent sell-through?”
If your AI tool can’t answer these cleanly in one place (Slack, Teams) and show the data behind the answer, stop there. Fix that first.
What ‘good’ looks like: plain-English Q&A that returns the result, the drivers (not just a number), and the next step (“open tasks for owners,” “draft payment run,” “schedule a weekly digest”).
Avoid: vague “insights,” charts without context, or anything that forces you back into Excel to make sense of it.
2) Automate the follow-through, not just the analysis
Answers are only half the job; the other half is getting the action moving.
- Collections: When you ask, “Who’s >60 days?” your tool should create call tasks, queue dunning emails, and set follow-ups for promises-to-pay—using your approval rules.
- AP: From “What’s due next week?” go straight to a proposed payment batch that respects cash limits and sign-offs.
- Inventory: From “What’s under 14 days of cover?” generate draft POs based on reorder logic and vendor lead times.
If your tool can’t draft the next step inside the system of record, people will stop using it. Analysis without action is a detour.
3) Close faster by fixing three sticky points
Most month-end delays trace back to the same culprits. Use AI to remove the manual grind:
- Reconciliations: “Show accounts not reconciled and the delta.” Assign owners and due dates.
- Accruals/GRNI: “Receipted not invoiced by vendor, last 30 days.” Draft accrual journals for review.
- Variance narrative: “Budget vs actuals by department; highlight >10% and list top drivers.” Store the narrative with the report so the story travels with the numbers.
This is mundane work. That’s the point. Automate it and free people for judgment calls.
4) Forecasting: less guesswork, more version discipline
AI should help you version the forecast, compare scenarios, and keep the conversation honest.
- Snapshot a weekly forecast automatically.
- Ask, “What changed since last week? Revenue, COGS, hiring—rank by impact.”
- When someone tweaks assumptions, log who/what/why so you can learn instead of arguing.
You’ll still make calls. AI just makes the calls traceable and reversible.
5) Bring non-finance into the loop without opening the ERP
Budget owners don’t need another tool; they need clarity.
- Let a marketing lead ask: “My spend vs budget MTD; top 5 vendors,” or “Open POs waiting on my approval.”
- Send a monthly budget digest to their channel.
- Allow them to request a reclass with one click—controller approves in the system.
If this feels too simple, that’s a good sign. Simple is what gets adopted.
6) Guardrails: roles, approvals, and audit trail
The boring stuff is what earns trust.
- Answers must inherit the same permissions people have in your ERP.
- Drafts (journals, payments, POs) must follow your approval chains.
- Every action needs an audit trail you can show to auditors without drama.
If a tool hand-waves this, it’s not ready for finance.
7) A rollout you can actually do in two weeks
Week 1
- Pick 5–10 questions per role (CFO, Controller, FP&A, AR/AP, Ops).
- Turn on Q&A, two meaningful alerts (e.g., unusual spend, late POs), and a weekly digest.
- Map each alert to an owner and a default next step (task, draft, approval).
Week 2
- Add three “do” actions: draft payment run, draft accruals, and schedule collections tasks.
- Cut one manual report you no longer need because the digest covers it.
- Put a 30-minute training on the calendar—show real examples, not slides.
Measure adoption with two simple metrics: (1) questions asked per week; (2) tasks created from answers. If both rise and close times drop, you’re on track.
8) What to skip (for now)
- Boiling the ocean. Don’t integrate every system on day one. Start where the data already lives (usually the ERP).
- New dashboards. If people aren’t opening the old ones, a shiny version won’t help.
- “AI strategy” documents. Run a pilot, publish the before/after, then write your strategy in one page.
9) Where Simon fits
If you want an off-the-shelf way to do the above: Simon sits on top of your ERP and works where people already are (Slack/Teams or the web app). NetSuite is the first ERP integrated in Simon, so all answers and actions respect NetSuite roles, permissions, and approvals. You ask in plain language; Simon returns the result, explains the why, and can draft the next step (payment runs, accruals, tasks, digests). That’s it. No new dashboards to babysit.
10) A few prompts you can steal
- “AR aging with owners—who’s >60 days and highest impact?”
- “Opex vs budget in July—biggest drivers by department and vendor.”
- “Items with <14 days of cover; show vendor and last lead time.”
- “Budget vs actuals by department; highlight >10% and summarize the why.”
- “Create a payment batch for bills due next 7 days; respect approvals and cash limit of €X.”
- “Draft accrual journals for receipted not invoiced last 30 days; send to controller for approval.”
Final thought
The best use of AI in finance isn’t a moonshot—it’s removing the daily friction that keeps smart people from doing their real job. Start with the questions your team actually asks, make the next step automatic, and keep the guardrails tight. If you do that, “AI in finance” stops being a project and becomes the way you work.
