TL;DR: Enzo from Accounting Prose built an automated system using AI to transform standard financial statements into personalized, actionable insights for clients. The system uses multiple AI models for accuracy, maintains human oversight, and delivers insights through existing client portals like Fathom and Xero. This approach democratizes high-level financial analysis for all clients, not just the highest-paying ones.
Picture this: Small business owners receive their monthly financial statements and... what happens next? If you're being honest, those carefully prepared reports probably sit unopened in email inboxes or get a quick glance before disappearing into some digital filing cabinet. Even those gorgeous reports with charts and graphs can feel like trying to read a foreign language when business owners don't know what the numbers actually mean for their day-to-day operations.
Enzo, founder of Accounting Prose, got tired of watching this happen and decided to do something about it. Their solution? An AI-powered system that takes those traditional financial statements and transforms them into personalized, actionable insights that clients actually get excited about reading.
|
Table of Contents |
Let's be real – traditional financial reporting, even when it looks pretty, often misses the mark on what small business owners actually need. Enzo puts it perfectly: "Our clients really need more than black and white financials. Even if we give them really nice Fathom reports with beautiful charts and all kinds of detail, sometimes it still feels a little bit overwhelming."
But the challenge goes way beyond just making things look nice. Small business owners are drowning in questions:
How did this month stack up against last month?
Are these numbers good or bad compared to other businesses like mine?
What should I actually do with this information?
What patterns should I be watching for?
Here's the kicker – most accounting firms save the really valuable analysis for their biggest clients, leaving everyone else with basic reporting that doesn't actually help them make better business decisions. That doesn't sit right with most of us who got into this business to help people succeed.
Enzo's approach centers on three principles that just make sense:
Instead of hoarding the good stuff for premium clients, why not give everyone valuable analysis? "I wanted to make sure that this wasn't something that we only offered our highest paying clients," Enzo explains. Sure, bigger engagements can go deeper, but every client deserves to understand what their numbers mean.
Nobody wants to feel like they're talking to a robot that says the same thing every month. By looking at what they've told clients before and what's happening now, the AI creates fresh insights that actually matter. It's like having a conversation that builds on itself rather than starting from scratch every time.
Even with all this fancy automation, real people make the final calls. "Even though this is gonna be automated, there's always gonna be a human in the loop as part of this process," Enzo emphasizes. Because at the end of the day, business advice needs to make sense in the real world, not just on paper.
Here's where things get interesting. Instead of putting all their eggs in one AI basket, Enzo set up a system using three different AI models: ChatGPT, Claude, and Google Gemini. Think of it like how airplanes work – they don't rely on just one computer to tell them if everything's okay. Multiple systems check the same information independently, and if two agree while one doesn't, well, you probably trust the majority.
In aviation, this redundancy literally keeps people alive. If one computer freaks out and says the engine is failing while the other two say everything's fine, the pilot knows not to panic over what's probably a false alarm.
If two models agree and one differs significantly, the outlier is likely incorrect. For example, if ChatGPT and Claude both identify declining gross margins as the primary concern while Gemini focuses on cash flow, the system recognizes the margin issue as the priority. This cross-validation catches instances where a single model might fabricate data points or misinterpret financial figures.
Each model has different strengths and training data. ChatGPT excels at structured financial analysis and industry comparisons, Claude provides nuanced contextual insights and natural language explanations, while Gemini often identifies patterns across different business metrics. By leveraging each model's unique capabilities, the final analysis incorporates multiple analytical perspectives that no single AI could provide.
If one service experiences issues, others can continue operating. During high-traffic periods or service outages, the system maintains functionality rather than completely failing. This reliability is crucial for meeting client deadlines and maintaining consistent service delivery.
When models disagree significantly, a fourth AI step acts as the synthesizer, weighing the different perspectives against company-specific context and historical data to produce balanced, actionable recommendations. This ensures clients receive coherent guidance rather than conflicting advice.
Getting to know how this system works shows off both its sophistication and Enzo's thoughtful design choices. The seven-step process smartly balances automation with human judgment, creating a workflow that scales expertise while keeping that personal touch small business clients value.
Here's how financial statements become insights that clients actually want to read:
Everything kicks off when controllers finish a client's financials and fill out a simple form. Enzo deliberately kept this part simple because nobody wants to add more work to an already crazy month-end process. The form just asks for the essentials: who's submitting it, which client it's for (from a dropdown that talks to all your other systems), what period you're analyzing, and the financial statements as a PDF.
Originally, Enzo thought about requiring both current and previous period statements, but then they had one of those "wait, why am I making this harder?" moments. The system works great with just current statements, so why create extra work? Sometimes the best feature is the one you don't build.
This is where things get clever. The system automatically grabs everything it knows about the client from HubSpot – and we're talking about details that actually matter for giving good advice.
It doesn't just see "marketing agency" and move on. It knows whether they're a specialized SEO shop or a full-service creative agency. It understands if they're a scrappy startup with three people or an established firm with 50 employees. This context makes all the difference because advice for a $500K business looks nothing like what you'd tell a $5M operation. Nobody wants enterprise-level strategic planning when they're still trying to figure out payroll.
This might be the smartest part of the whole system. Before creating any new insights, the AI reviews the last six pieces of advice given to this client. Why? Because nothing screams "this is clearly automated and no one cares enough to fact check" like getting identical recommendations three months running.
The system tracks what it's told each client across three areas: big-picture insights, specific action items, and next steps. If it recommended improving cash flow last month, it won't just repeat that – it'll build on it or focus on different areas. And when a controller tweaks the final advice before sending it out, they can update the record so future analyses stay consistent.
Now for the really cool part, and there's actually a fascinating story behind why Enzo set this up this way. Instead of relying on just one AI, they deliberately built a system using three different models working completely independently: ChatGPT, Claude, and Google Gemini.
This approach was directly inspired by how computers work on airplanes – and once you understand that connection, you'll see why it's brilliant for financial analysis too.
Here's how it works in aviation: Modern aircraft don't trust their lives to a single computer. Instead, they use multiple independent flight computers that all monitor the same data – engine performance, altitude, speed, weather conditions – but process that information separately. If one computer suddenly declares "ENGINE FAILURE! WE'RE GOING DOWN!" while the other two calmly report "everything's normal," the pilot knows to trust the majority consensus rather than panic over what's probably a malfunctioning sensor or corrupted data.
Enzo had the same "aha" moment about AI and financial analysis. What if one AI model has a bad day and starts seeing problems that aren't there? What if it misinterprets a number or gets fixated on the wrong metric? In aviation, that kind of error could be catastrophic. In financial advice, it could lead clients to make terrible business decisions based on flawed analysis.
So Enzo borrowed this proven safety concept and applied it to accounting. Each AI gets the same information – financial statements, company background, and previous advice – but processes it through their own unique "lens" based on different training data and strengths. ChatGPT might excel at industry comparisons and structured analysis, Claude could provide more nuanced contextual insights, and Gemini might spot patterns across different business metrics that the others miss.
The crucial part is they work completely separately, with no knowledge of what the others are thinking. This prevents the "groupthink" problem where one AI's interpretation influences the others, potentially amplifying errors instead of catching them.
After all three AIs finish their independent analysis, a fourth AI steps in to play the role of an experienced air traffic controller or senior pilot – someone who can look at multiple expert opinions and synthesize them into clear, actionable guidance.
This "consensus builder" examines all three recommendations with a critical eye, looking for patterns and agreements while also weighing disagreements against what it knows about the specific client and their industry context. The process mirrors how experienced pilots evaluate conflicting information from multiple systems to make safe decisions.
For example, if ChatGPT and Claude both flag declining gross margins as the primary concern while Gemini focuses on cash flow issues, the consensus builder doesn't just pick randomly or split the difference. It considers factors like the client's industry (where margin pressure might be more critical), their business stage (where cash flow might be more urgent), and their historical patterns to determine which concern deserves priority attention.
When there's significant disagreement, the system doesn't just flip a coin – it applies the same kind of critical thinking a pilot uses when instruments disagree. It weighs each recommendation against company-specific factors, industry knowledge, and logical consistency. If two AIs identify cash flow concerns as urgent while one suggests focusing on growth initiatives, the system recognizes that you can't grow your way out of a cash crisis and prioritizes accordingly.
This aviation-inspired approach gives Enzo's clients something remarkable: financial insights with built-in error checking and multiple expert perspectives, all synthesized by a system designed to catch and correct the kind of mistakes that could derail a business. It's like having a whole team of financial analysts review each client's situation, but with the reliability safeguards that keep airplanes safely in the sky.
The beauty of borrowing from aviation is that these redundancy systems have been tested and refined over decades in life-or-death situations. If multiple independent systems can keep a 400-ton aircraft safely airborne, the same principles can certainly help small business owners make better financial decisions.
Here's where human expertise comes back into the picture. Controllers get the AI-generated insights formatted as overview, action items, and next steps, but they have the final say on everything that reaches clients.
This isn't just rubber-stamping either. Controllers can modify recommendations, add context that only comes from knowing the client relationship, or completely rewrite sections that don't feel right. Maybe the AI suggests aggressive growth when the controller knows the client is dealing with personal challenges, or it misses industry-specific details that need human judgment.
This review step also creates a feedback loop that improves the system over time. When controllers consistently change certain types of recommendations, it signals that the AI prompts might need tweaking.
The final step delivers insights directly into platforms clients already use. Fathom users see insights right at the top of their monthly reports – the first thing they notice when they get that notification. Xero users get everything integrated into clean management reports alongside traditional statements.
Behind the scenes, the system also creates notes in HubSpot tickets so the team has a complete record of what was shared and when. It's like having an incredibly organized assistant who never forgets to document important conversations.
What makes this work is that clients don't need to learn anything new or check another platform – insights just appear where they expect to find their financial information, making it much more likely they'll actually read and act on them.
When team members asked about uploading sensitive client data to AI models, Enzo didn't just brush off their concerns – they took immediate action to lock everything down:
No Training Data: They configured all AI services to explicitly exclude client data from model training. Your client's information won't accidentally show up in someone else's chat.
Account-Level Protection: All data stays within the firm's controlled accounts – no mysterious third-party access.
Triple-Checked Settings: They went through every privacy setting multiple times to make sure nothing was missed.
"I've already gone through all those steps to make sure that all that is totally safe," Enzo confirms. These concerns are completely valid, and any firm considering this approach needs to address them head-on.
The system doesn't just spit out generic advice and call it a day. It considers multiple factors to make sure recommendations actually fit each business – because giving a startup the same advice you'd give an established enterprise is like prescribing the same medication regardless of age, weight, or medical history.
Revenue Size: A $500K business needs fundamentally different advice than a $10M company. When the AI spots declining profit margins, it might recommend cost-cutting measures and efficiency improvements for the smaller business, while suggesting strategic investments in automation or market expansion for the larger one. The $500K business probably can't afford a dedicated CFO, but the $10M company might need one yesterday.
Team Size: Recommendations dramatically shift based on whether they have 5 or 50 employees. A company with 5 people getting advice about implementing complex HR systems would be ridiculous, while a 50-person company ignoring employee productivity metrics could be missing huge opportunities. Revenue per employee calculations mean something completely different for a lean tech startup versus a labor-intensive service business.
Industry Details: Here's where the system gets really smart. Instead of just seeing "marketing agency," it knows whether they're an "SEO-focused digital marketing agency" versus a "full-service creative agency with video production capabilities." The SEO shop might get advice about recurring revenue optimization and client retention, while the creative agency might hear about project profitability and equipment depreciation strategies.
Business Structure: C-corps get different advice than sole proprietorships because they face entirely different tax implications and growth constraints. The system won't recommend aggressive reinvestment strategies to a sole proprietor who needs to manage personal tax liability, but it might push a C-corp toward strategic spending that reduces corporate tax burden.
Ownership Structure: Single owners can make quick decisions but bear all the risk, while multiple partners need consensus but share the burden. The AI adjusts its recommendations accordingly – suggesting bold moves for decisive single owners while recommending collaborative planning processes for partnership structures. It also considers how owner draws and distributions affect cash flow differently across various ownership models.
Location: Regional factors and local market conditions create significant context differences. A restaurant in Manhattan faces completely different cost structures and customer expectations than one in rural Montana. The system factors in regional economic conditions, local competition levels, labor market dynamics, and even seasonal variations that affect different geographic areas uniquely.
Growth Stage Awareness: The system also recognizes whether a business is in startup mode, scaling rapidly, hitting maturity, or potentially declining. A rapidly growing company might get advice about managing cash flow during expansion, while a mature business might hear about diversification strategies or efficiency improvements.
All these factors work together to create a comprehensive picture that ensures advice isn't just accurate – it's actually useful and actionable for that specific business in their specific situation
This approach represents a real shift in how accounting firms can serve small businesses, and honestly, it's about time. For too long, the industry has operated on a tiered system where your budget determines the quality of insights you receive. Enzo's system flips that model on its head.
By making sophisticated analysis available to everyone, firms can fundamentally change their value proposition:
Give all clients the same high-level treatment – No more awkward conversations about why the $2K/month client gets detailed analysis while the $500/month client gets basic reporting. Every client deserves to understand what their numbers mean and what actions they should take.
Keep clients longer through better value – When clients feel like they're getting premium insights rather than just compliance work, they stick around. They also refer more business because they're genuinely excited about what their accountant provides.
Stand out in a crowded market – While other firms compete on price or promise faster turnaround times, you're offering something that actually transforms how clients run their businesses. That's differentiation that matters.
Help more clients without burning out the team – Instead of senior staff spending hours crafting individual client insights, they're reviewing and refining AI-generated analysis. This scales expertise without scaling headaches or endless overtime.
Actually help businesses succeed during tough times – Economic uncertainty makes business owners desperate for guidance. Providing actionable insights during challenging periods doesn't just help clients survive – it positions your firm as an indispensable strategic partner rather than a necessary evil.
The ripple effects go beyond individual firms too. When small businesses get better financial guidance, they make smarter decisions, create more jobs, and contribute more to their communities. It's the kind of positive cycle that benefits everyone.
As Enzo puts it: "If there's something that we can do to help our clients sleep better at night, I'm all for it. Even if it means I've gotta learn a whole new tool, a whole new process."
Integrating AI into accounting isn't about replacing human expertise – it's about making it go further. By automating the grunt work of analyzing data and generating initial insights, accountants can focus on the good stuff: strategic planning, solving complex problems, and building relationships that actually matter.
For small businesses, this means getting access to the kind of sophisticated financial analysis that used to be reserved for companies with dedicated finance teams. When done right, with proper security and human oversight, AI-powered insights can completely change how small businesses understand and use their financial data.
The key is finding accounting partners who embrace these tools while keeping that personal touch and professional judgment that small businesses depend on. As this technology keeps evolving, the firms that successfully blend AI capabilities with human expertise will deliver the most value to their clients.