How Phil Walker uses AI to deliver advanced analytics without the enterprise price tag
When you're running a fractional CFO practice, you encounter a fundamental challenge: your SME clients need enterprise-level insights, but they don't have enterprise-level budgets. Phil Walker, co-founder of Mettryx, has found a way to bridge that gap using AI as his technical co-pilot. Mettryx equips ambitious business leaders with strategic financial clarity and hands-on support to drive growth, profitability, and long-term value, no matter their size or stage.
His approach isn't about replacing human expertise, it's about removing the barriers that prevent finance professionals from delivering sophisticated analysis to smaller businesses.
The reality check on AI adoption
Phil brings a refreshing dose of realism to the AI conversation. While the finance community buzzes with cutting-edge possibilities, he's keenly aware of the gap between what's being discussed and what's actually happening in practice.
"I'm always cognisant of the fact that there's a bell curve of capability out there," he explains. "AI might be at the forefront of digital transformation, but there's still a tail of organisations that haven't embraced even basic digital changes. Believe it or not, we’ve recently encountered businesses using paper records for parts of their process."
This perspective shapes how he approaches AI implementation; cautiously, strategically, and always with the client's reality in mind.
Where AI actually delivers value
Beyond the obvious uses like proofreading and tone adjustment, Phil has found AI's sweet spot in technical assistance. Working across multiple client systems means constantly encountering new software, and AI has become his instant help desk.
"We can't always have tons of experience in all the systems we encounter," he says. "AI essentially has all those help documents and user manuals ready for us to ask questions about the tools we're using."
But the real breakthrough came with code writing support.
Deep-dive analytics without the enterprise overhead
Phil's most successful AI implementation centers on data analysis and reporting. Using ChatGPT to help write SQL queries and DAX code for Power BI, he's able to deliver sophisticated management reporting that would typically require expensive enterprise tools.
One client needed comprehensive profit analysis across multiple dimensions, margins by channel, geography, and customer groups. Traditional solutions would have required significant investment in data analytics suites and implementation costs that simply weren't viable for an SME.
Instead, Phil used AI to help write the SQL queries needed to extract data from the client's system and bring it into Power BI. The AI then assisted with creating the DAX queries that powered the analytical insights.
"We would have got there without AI, it just would have taken us much longer," he reflects. "We're able to provide the value and the insight back to the client more quickly and more cheaply because we didn't have to invest in a large implementation of a data analytics suite."
The assistance mindset
Phil emphasises that success with AI requires treating it as an assistant, not a replacement.
"You've got to have context, background, and be clear about the output. Just like with a junior team member performing a task, you can't give a one-sentence instruction and expect them to come back with a fully developed solution exactly how you envisioned it. They need guidance."
This approach extends to his technical work. When using AI for code writing, he provides table names, structures, and defines the output requirements clearly. The result isn't perfect code that runs immediately, but it's a significant acceleration in development time.
"The development speed for a piece of code like an SQL extract or a DAX query is significantly shortened through using AI," he notes.
The domain knowledge requirement
Phil is particularly vocal about the importance of foundational expertise when using AI.
"You've got to have the domain knowledge to challenge what you've seen and whether it's correct. It's almost like coaching a part-qualified member of the team."
He's seen this firsthand with reconciliation work.
"I've tried asking GPT to do simple maths, and it will get it wrong but present you with the answer in such a compelling way that you would believe it if you didn't already know the simple maths was wrong."
This is especially critical in finance, where reconciliations must balance and accuracy is non-negotiable.
"The models want to please, they want to provide a solution, so they will happily provide you a reconciliation that is fundamentally flawed but looks okay on face value."
A measured approach to client delivery
While Phil uses AI within his practice, he's more cautious about client-facing applications. He's aware of several peers in the industry attempting to use AI for writing client commentary and variance analysis, but notes they're "all running into problems."
"When it comes to real client delivery like management report production and variance analysis, I'm yet to hear of anybody who's really succeeding with AI," he admits. “Yes, there are apps and automations available which are undoubtedly improving efficiency, but I’m cautious to label these as true ‘AI.’”
This isn't pessimism, it's professional responsibility. Phil understands that in a fractional capacity, trust is everything, and he won't compromise client relationships by pushing technology that isn't ready. At the moment, delivering client value supersedes investing significantly in a development effort for their own practice “ though it’s undoubtedly on the road map”.
The practical truth about AI in finance
Phil's experience offers a blueprint for practical AI adoption in finance: use it to enhance your existing skills, not replace them. Focus on areas where it can remove technical barriers and accelerate development, but always maintain the domain expertise to validate outputs.
"If you're already good at what you do, AI helps you move faster," he concludes. "But you've got to have the experience and education to validate whether the results it's produced are accurate and valid or not."
His approach proves that the most successful AI implementations aren't about revolutionary changes, they're about thoughtful applications that solve real problems while respecting the fundamental requirements of professional finance work.
For Phil, AI isn't the future of finance, it's a powerful tool that's helping him deliver better value to his clients today.
Join The Stack Exchange: Our free Slack forum where in-house finance teams share tools, test ideas, and compare notes on what’s working (and what isn’t) when it comes to AI in the finance function.
.png?upscale=true&width=1196&upscale=true&name=Untitled%20design%20(42).png)
