How Martin Goodwin used ChatGPT, Power BI and DataSights to transform reporting at Open Box Software.
In the last few weeks, weâve focused on where to start with AI and how to scale its use across your finance function without turning it into chaos. But theory can only take us so far.
Itâs time to go practical.
This week, weâre kicking off a new series in CFO Techstack: AI in Action. Each edition will feature a finance professional who used AI in one very specific way, whether to remove manual work, fix a pain point or completely transform a workflow.
Weâre starting with Martin Goodwin, Group Finance Manager at Open Box Software.
From spreadsheets to on-demand reporting
Martin joined Open Box Software with a clear mandate: âclear backlogs, find automation wins and improve finance processes.â Like most finance teams, the usual suspects came first: fixing bank feeds and tightening up how Xero was used. But the real challenge? Reporting.
Open Box operates across multiple entities and currencies. And while Xero and Syft got them part of the way there, finance couldnât solve a very specific need: standardising reporting at a fixed exchange rate using the source currency of each transaction. As Martin explains:
âWe like to report at a standard exchange rate, so every year we use the same exchange rate, so that you can see the actual growth without the impact of currency movementsâ
The workaround was tedious. Export from Syft, manually manipulate the data. Strip out converted revenue. Recalculate everything manually using custom exchange rates. âIt took days,â Martin says.
He needed a new approach, one that could automate this process, provide flexibility, and work securely with sensitive data. Thatâs when he came across a post from Tyler Caskey, whose name he recognised from the CFO Techstack community, recommending DataSights as a powerful way to extract and manipulate financial data from Xero for reporting.
The learning curve: ChatGPT as an AI-powered tutor
Martin had never used Power BI or SQL before. But he knew this was the stack he needed.
Rather than hand it off to the dev team or bring in someone to do it for them, which wasnât an option due to data sensitivity, he decided to upskill himself, with the help of ChatGPT.
âI did a short Udemy Power BI course, but itâs not very practical... It taught me the basics and where to click, but not how to write the code. So I started using ChatGPT.â
Initially, the answers were rough. âTerrible,â in Martinâs words. But with repetition and structured prompts, providing full context, table structure, column names, he started getting better results.
âEventually I learned how ChatGPT gives you a proper answer. You have to treat it like a developer and give it every bit of info you have so that it has context.â
ChatGPT became a sounding board and tutor, helping Martin think through problems, identify mistakes in logic, and iterate. Even when the outputs were wrong, they pointed him in new directions.
âNow I donât use it as much, I can write the SQL myself. But back then, I wouldnât have been able to do any of it without AI.â
Why domain knowledge still matters
What stood out in Martinâs process wasnât just the AI prompts, it was the judgment behind them. As he put it:
âIt didnât give me the right answer a lot of the time, but it gave me the start⌠and when itâs wrong, you still have to know, okay, why didnât it work?â
This is a crucial insight. AI can generate answers, but only someone with domain knowledge can judge if those answers are any good. It takes experience to know whether the output makes sense, where itâs falling short, and how to refine it. Even when the output wasnât usable, Martin says it helped him approach the problem differently, think through the logic more clearly, and better define what he was actually trying to achieve. In many ways, the biggest value came from how ChatGPT made him think, pushing him to sharpen his understanding and improve the way he communicated his problem.
From scratch to scalable
Martin built a custom reporting system using DataSights (for extraction and transformation), SQL (for logic and structure), and Power BI (for visualisation). The result?
A fully custom Reporting and FP&A platform, built in-house, with hundreds of live reports, built exactly how each user requires it:
- Fully consolidated P&Ls and Balance Sheets
- Divisional and departmental P&Ls
- Actual vs. budget vs. prior year
- Currency impact analysis
- Revenue Analysis by division, customer, location, service type, currency
- Customer concentration tracking and growth
- Debt ageing and working capital metrics
- Live dashboards for each division & department with full access control
- All reports have full transaction level drilldown
Itâs fast. Itâs automated. Itâs secure. And, most importantly, itâs now the platform used by leadership to review the business.
âOur Exco meetings start with Power BI. Itâs not just reporting anymore. Itâs where we do FP&A.â
Where itâs going next
With the success of the reporting rebuild, Martinâs now moving further upstream, connecting project, billing, CRM, and payroll data into a central SQL database. The goal is to create a single source of truth that is all connected.
Eventually, that foundation will allow for more advanced AI tools that can analyse, detect patterns, or even flag anomalies before humans spot them.
But none of that happens without clean data.
âThe data has to be good before you can actually use AI. Garbage in, garbage out.â
Final thought
Martinâs story isnât about buying a fancy AI tool. Itâs about using the tools already available, ChatGPT, SQL, Power BI and applying them to a real, everyday finance problem.
He didnât wait for the perfect software or a dedicated AI budget. He started with a pain point, figured out what needed to change, and learned the skills required to make it happen.
Thatâs what this series is about.
One process. One tool. One win at a time.
Next week, weâll feature another finance professional solving a different problem with AI. Until then, if youâve got a use case worth sharing or just want to learn how others are doing it, join the conversation in The Stack Exchange.