Why Waiting for AI to Mature Is the Most Expensive Choice CPA Firm Can Make

\ I spent over a decade in Big Four doing audit and financial due diligence (FDD) of enterprises, commercial diligence and value creation for private equity across over 20 industries, from coal mines to tobacco factories to financial services. Since then, technologies have made an immense leap forward, but in finance, time seems to have stopped in the 1990s. The quality of earnings analysis, once industrialized around Excel-based workflows , has ==never fundamentally changed==. This is what I saw at KPMG from day one: the smartest people in the room spent weeks cleaning data, running reconciliations, and applying standard checklists. These are strong experienced analysts, who could create made-to-measure financial strategies for businesses. Yet here they sat, ==wasting potential on a mechanical job.== More than ten years have passed. Today we have more associates and better laptops. But just like thirty years ago, FDD still takes six to eight weeks, with half the time spent on spreadsheets, rather than real analysis - the very thing clients are paying for. What has changed is the volume of data inside a typical mid-market target -and it has changed in every direction. Some companies are drowning in it: volumes that have grown by orders of magnitude across systems that were never designed to talk to each other. Others have almost nothing structured to work with at all. And nearly all of them run on legacy architectures that require a custom integration approach before AI can do anything useful. ==The good news==, we have located the efficiency gap and we know how to close it with AI. So why does the problem remain? In this article, I'll look at why CPA firms prefer to play safe rather than optimize the process with automation, and what waiting too long for the right moment is going to cost them. The Bottleneck Let’s look at where those weeks actually go. The first three are usually spent pulling the general ledger out of a legacy accounting systems, tying the trial balance to the financials, bridging management EBITDA to reported EBITDA, and identifying non-recurring items by reading through journal entries one transaction at a time. This is structurally repetitive work, which does not require senior judgment. The interpretation part that follows is a whole another level. What does a 4% revenue decline in Q2 mean for this specific buyer's thesis? Is customer concentration a risk or a feature? Does the working capital normalization the seller proposed survive contact with the actual cash conversion cycle? These are the questions that senior practitioners were trained for and the work clients are actually paying for. AI handles the first part better than humans do. Faster, more consistent, and at a scale no analyst team can match, processing thousands of transactions, flagging anomalies, surfacing correlations that manual review would miss entirely. When the technology is built specifically for financial due diligence rather than repurposed from a general-purpose model, that's what it demonstrably delivers. The interpretation layer stays human. Context, judgment, ethics, the ability to read what a number means for a specific deal and a specific buyer are not automatable, and shouldn't be. What automation does is give senior practitioners the time and the clean data to actually do that work, instead of spending the first three weeks creating the conditions for it. So ==here emerges a bottleneck==, and it’s not really in the technology, which has been available for at least two years. It’s the willingness of CPA executives to redraw the workflow. Most are waiting for the technology to mature. Waiting for someone else to make the first mistakes. Waiting for a clearer signal from the market. As people who are professionally used to evaluating risks, the logic feels prudent, but in reality waiting is the most expensive position a firm can take right now. Where The Gap Lies Senior practitioners typically raise two objections to automation. The first is that AI can introduce errors into deliverables that have to withstand legal scrutiny, and eventually feed directly into deal terms, purchase price mechanics, and post-closing disputes. The second is genuine uncertainty about what automation does to the profession over time, the fate of junior staff training, and whether the long-term consequences are manageable. As for the first objection, purpose-built financial ==AI surpasses 99% accuracy== on reconciliation, data extraction, and transaction matching - the very rule-based, formula-driven work that defines the first weeks of any engagement. Every output is fully traceable and audit-ready. It’s like replacing an abacus with a calculator. The second objection is trickier - because it’s personal. ==Nobody wants to believe they can be replaced==, and that fear is exactly what makes avoiding automation dangerous. Many worry automation will hollow out junior roles, but automation allows them to focus on analysis, judgment, and client exposure, which are a better training than transaction matching. Data safety is auditable and controlled, no different from the standards firms already apply to client data. And clients watching this space will reward structured, demonstrable automation. Automation is already here, so the firms that move now are shaping how it lands inside their practice. The firms that choose to wait inherit whatever the market decides for them. There is no neutral position. Standing still is a choice, and the market will price it accordingly. Sequenced Adoption There’s no real need to wait until the technology gets perfect and update everything at one go. Start where your team is bleeding the most time right now - it’s a chance to see where automation delivers the fastest return and where the case for change becomes impossible to argue with. ==Start with back-testing== - run the automation in parallel with your existing manual process on completed deals. This builds institutional familiarity, surfaces edge cases in a controlled environment, and gives your team the evidence they need to trust the output. Once back-testing is consistent, move to live deals on the lower-risk work streams first. Change management is important here. The staff needs to get comfortable with these tools now, not when a client demands it. If you haven't started that conversation with your team yet, you're already behind. Financial Value Once a firm automates that data cleaning layer, the economics of the engagement change. The same work gets delivered faster, senior professionals spend their time on analysis and strategic judgment rather than reconciliation, and the capacity that was locked inside manual workflows becomes available for higher-value work. In this conversation about the financial gains it’s important to mention a widespread concern that automation will shrink billable hours. Billing by the hour for work a machine now does in minutes is a dead end for sure, but that’s not a problem of automation. Clients have always been questioning whether the hourly billing model is reasonable at all. The smarter move is to shift toward fixed-fee and value-based engagements. For the teams, automation creates real space for upskilling - senior practitioners developing deeper analytical capabilities instead of spending their careers in Excel. For the firm, it opens the door to a more comprehensive service offering: from deliverable providers to genuine transaction advisors covering the full deal lifecycle. Who Will Hold A Competitive Edge Twelve to fourteen months. That's how long it takes for a more tech-enabled competitor to start winning engagements away from firms still running manual workflows. Every quarter spent deliberating is a quarter a competitor spends building institutional capability. The firms that move now get to shape how automation lands inside their practice with time to iterate. By 2030, FDD will look ==nothing like it does today==. The Excel-based workflows industrialized in the 1990s will be gone. The engagements will be faster and the value delivered to clients fundamentally different from what a manual team can produce. The question is how you use the time to create dramatically more value for the client - and they will follow the ones who deliver more, faster, with better insight. This supplies CPA firms with contract renewals and major engagements, often before the incumbent realizes the conversation has started. In our fast-paced world, the ==luxury of moving on your own timeline== doesn't last long. \n \