Julian Rowlands

The through-line

Rowlands' whole argument starts from a problem he lived: as a CFO, the numbers he needed were scattered across a sales CRM, an accounting system, a payroll system, a budget and "that one random Google sheet that I have that that has that super important slice of data" [3], none of which he controlled and all of which labelled things differently [1][2][3]. The only fix he saw among the CFOs he admired was heavy-duty custom data infrastructure, connectors into a warehouse like Snowflake or Redshift, a BI layer like Looker, Tableau or PowerBI on top, and a team of analysts paid to read the dashboards and write narratives for the business [1][2]. His verdict on that stack is consistent: "that's an incredible destination to get to, but it's so expensive and it takes so long" [1]. Cashboard is his attempt to hand the same result to a mid-market finance team in weeks rather than quarters, with AI writing the analysis that the analysts used to write [1][2][3].

The north star he keeps returning to is "inventing the solo CFO" [1][3], and the notable thing is how carefully he polices his own claim. He calls it aspirational, says plainly "we've not done it yet," puts the company at "30 40% of the way there" [3], and reserves open contempt for the vendor habit of claiming otherwise: if he hears one more product described as "infused with AI" he says he will lose his mind [3]. The division of labour underneath the slogan has stayed fixed across everything he says. AI takes execution and data chores; humans keep judgement.

On what AI can and cannot do in finance

His distinction is between knowledge and experience. AI is "great for synthesizing data," good at executing workflows and at very small agentic decisions, but "it doesn't have wisdom" and "it has knowledge, it has data, but it doesn't have experience" [3]. From that follows the rule: "you never want AI to be your CFO" [3]. What he expects instead is that "human CFOs will be using AI to execute" [3], pulling together analysis, drafting write-ups, running reporting, producing first and second rough drafts, with a person in the loop who signs off, sends work back to the agent for another take, or exports to Excel for the final touches [3]. He draws a hard line at the board deck: he does not think Cashboard will ever be in that business, and if you are letting AI build yours, "you've got bigger problems" [3]. The same line runs through his view of budgeting, where the conversations with department leaders and the board are the point and "AI cannot and should not do that," while everything around them, consolidating templates, splitting data back out, redoing the roll-up after the board says make more money and spend less, is chores that should be automated [1]. He is equally impatient with the technical naivety behind a lot of AI product claims, arguing that building a real financial product means understanding how the tools allocate GPU and CPU tasks correctly rather than dropping data into ChatGPT [1][3].

On the single source of truth, and ignoring the P in FP&A

Cashboard's design follows from a deliberately narrow bet. He describes the choice as "a little uh heretical in terms of building FPNA software," which is "to completely ignore the P" [1]. His reasoning is that plenty of companies are moving planning and modelling into the cloud, and a CFO can already consolidate a budget in Anaplan, Adaptive, Runway or Mosaic, but those same CFOs are still separately building the pipeline into a warehouse and out to PowerBI, with something bolted on to reconcile the data labels [1]. Nobody, in his view, was replacing that infrastructure, so that is where he aimed: a full BI tool, "a full replacement for something like PowerBI, but purpose-built for CFOs" [3], with AI-powered data mapping keeping labels aligned so joins actually produce apples-to-apples comparisons [3]. The AI that writes and sends reports on autopilot, "so you don't have to be awake" [3], is the natural extension rather than the starting point. Budgeting and forecasting are on the roadmap as the next step towards the solo CFO, not the foundation [1][3].

On who this is for

He is blunt about the floor. "We solve complexity" [3]. A five-person company where the founder carries the bank balance in their head and knows every customer by name does not need the product [3]. The need appears when data has split across HubSpot or Salesforce, QuickBooks or an accounting system, payroll and a spreadsheet budget, and several decision makers across the business each need to see performance in a single pane of glass [3]. In practice that means middle-market businesses, many around 15 to 20 million in topline, plus smaller ones whose complexity is compounding fast, with enterprise deferred because of the complexity involved [3].

On what happens to the finance career ladder

This is the question he says he spends the most time on and claims the least certainty about. The traditional route ran through execution: an FP&A associate does what the manager or CFO tells them, sits in the room when the numbers are presented, learns, and climbs to director, VP, CFO [3]. If AI absorbs the execution, that apprenticeship disappears. His best guess is that the junior finance role starts to look like RevOps under a CRO, people responsible for making sure the machines are hooked together properly, the data flows in, the data model and mapping are right, and the whole system behaves as desired [3]. The skill that matters is therefore "data fluency," concretely an introductory command of SQL, understanding how joins work, and being able to read a cleaned data model built on top of the raw tables from the source systems [3]. Without that, he argues, you are guessing, and you will either run in a black box or trust the system too little to automate anything at all [3]. He notes this is not an exotic profile, since plenty of finance people already pair finance with management information systems [3].

On becoming a CFO without the résumé for it

He calls himself an "accidental CFO" [1]: four years of data systems and data warehouse consulting for hedge funds, business school at Chicago where he mocked the place as a CFO factory before immediately becoming one, an intent to do emerging markets work in sub-Saharan Africa, Myanmar and Bangladesh, and then a fit at a Southeast Asian startup that needed someone who knew data systems, would live in Jakarta and could speak both finance and VC [1]. He was explicit in the interview process that he had never worked in finance and was not a CPA, and describes what he got as "a build rather than buy experience," a bet that he was smart enough to figure it out [1]. His method for figuring it out is his most transferable advice: on day one he asked the company which CFOs in their network he could learn from, fast, and worked his own network in parallel, ending up with mentors including Cam Heiser, later the CFO who took ZoomInfo public in the first half of 2020, plus introductions through Y Combinator to the CFOs of OpenAI, Segment and Brex [1]. He still trades emails with many of them [1]. He also frames his own qualification for building Cashboard as an overlap rather than an excellence: "Not the best CFO in the world, not the best data engineer or anything in the world, but I think there's there's there's one of me" [3].

On product-market fit and how hard building is

The lesson he carries from Xendit is the oldest one in the Y Combinator canon and he says it anyway: "build things that people want" [1]. His gloss is that product-market fit is the only thing that matters, because "if people are ripping your product out of your hands, then everything else like you have the opportunity to figure all the other stuff out," whereas without it no amount of foundation-building will help [1]. He watched that happen at close range, joining at roughly 70 or 80 employees with developers working in a sometimes-flooding basement in South Jakarta and leaving eighteen months later at 250 [1][3]. Spruce rhymed with it: a complex B2B money-moving problem wrapped in a simple API, built by founders working on their own pain points [1]. What he did not anticipate is how much harder building from scratch would be than watching it, despite having been in those boardrooms, and he cites a Collison line about the world being a museum of passion projects to explain his new respect for anyone running a bakery or a coffee shop [3]. The counterweight is customer reaction: he says they "see the whites of people's eyes" during demos when someone sees something that genuinely changes their job [3].

Takeaways

• The core CFO pain is not analysis but fragmentation: data in a CRM, accounting, payroll and budget systems you do not control, with different labels in each, which makes understanding performance impossible without expensive infrastructure [1][2][3].
• Cashboard deliberately skips planning and modelling to replace the connector-warehouse-BI-analyst stack, on the view that budgeting tools like Anaplan, Adaptive, Runway and Mosaic already exist while the data layer does not [1].
• Draw the automation line at judgement: AI handles synthesis, drafts and reporting, humans keep decisions, board conversations and the board deck, because AI has knowledge and data but not experience [1][3].
• The solo CFO is a stated north star at roughly "30 40% of the way there," with a single CFO plausibly running mid-market FP&A in a year to eighteen months and enterprise later [3].
• The junior finance role is likely to become RevOps-like systems stewardship, so the skills to build are SQL, joins and the ability to read a data model [3][7].
• The product is aimed at complexity, not size: unnecessary at five people, valuable around 15 to 20 million in topline or wherever data has split across several systems [3].
• If you take a CFO job you are not yet qualified for, be explicit about the gap and immediately assemble an ad hoc board of experienced CFOs through the company's and investors' networks [1].
• Product-market fit is the only foundation that matters; if customers are ripping the product out of your hands, everything else can be fixed later [1].