I Gave an AI My Acquisition Model. It Gave Me My Week Back.

I Used to Spend 2 Days on a DCF. Here's How I Do It in 2.5 Hours Now. TL;DR: Startup valuation isn't a formula — it's a judgment call that changes depending on your company's stage, market, and what a buyer is actually willing to pay. On a representative acquisition model I used to build from scratch, the work took about 14 hours. Using the same inputs, methodology, and judgment calls but automating the mechanical steps, I got the same deal to the same answer in 2.5 hours. The only thing that changed is what I stopped doing manually. Let me say the quiet part out loud: most of the time a finance team spends on a valuation model isn't thinking time. It's formatting time. Comp-pulling time. Table-building time. The stuff that feels like work but doesn't actually require a human brain. I've spent probably hundreds of hours of my career doing exactly that. And I'm done pretending it was necessary. First, let's clear up the valuation confusion Before I get into the workflow, I want to address something that trips people up constantly — especially founders who've just been handed a term sheet and suddenly need to know if the number on it is reasonable. Valuation isn't a fixed answer. It's an output that depends entirely on which method you use, and which method you use depends on where your company actually is. Early-stage? Revenue multiples don't mean much when you barely have revenue. You're being valued on market size, team, traction signals, and what comparable deals looked like at similar stages. A few named frameworks are built specifically for this: The Berkus Method , developed by angel investor Dave Berkus in the mid-1990s, assigns a dollar value (originally up to $500,000 each) to five qualitative risk factors — idea, team, prototype, strategic relationships, and early sales — rather than relying on revenue projections [1]. The Scorecard Method , developed by angel investor Bill Payne, values a startup by comparing it against the average pre-money valuation of similar startups in the same region and stage, then adjusting up or down across weighted factors like team quality and market size [2]. The traditional VC Method works backward from a projected exit value and the investor's required return to arrive at a present-day valuation. Any of these will get you closer to reality than a DCF built on five years of projected numbers that nobody believes anyway. Later stage, profitable, or acquisition target? Now financials actually matter. EBITDA multiples, DCF, comparable transactions — these work because there's real data underneath them. The mistake I see most often: founders applying the wrong framework to their stage. A pre-revenue SaaS startup running a 20-year DCF isn't being rigorous — it's just making up numbers with extra steps. At the end of the day, a startup is worth what an investor or acquirer will actually pay. Full stop. No model changes that. The deal that made me rethink everything Here's a composite example based on the kind of deal I see regularly, with the numbers adjusted for illustration: a SaaS acquisition target with $8M ARR, 35% growth, and a $40M offer on the table. The kind of deal where the CFO wants a tight model and a clean deck before anyone gets on a plane. (Note: the specific figures below — the comp set, growth path, WACC inputs, and IRR — are an illustrative composite, not a disclosed real transaction. The workflow and time savings described reflect my actual practice.) The old way looked like this. I'd pull six comps manually from Capital IQ — reconciling fiscal years, debating outliers, deciding who stays in the set. Then revenue projections row by row ($8M → $13.5M → $19M), WACC by hand (4.3% risk-free + 1.2β × 5.5% + 2.1% = 13.2%), terminal value at 4× ARR, a hand-built 5×5 sensitivity table, and two hours of reformatting so the deck didn't look like a panic attack in Excel. Total: about 14 hours. Most of it mechanical. The new way started with the same inputs and ended with the same outputs. But the middle looked completely different. Comps were summarized in 12 minutes, with two outliers flagged automatically. Revenue bridge with documented assumptions: 8 minutes. WACC tested across three different capital structures: 5 minutes. A 7×7 sensitivity table (IRR vs. entry multiple vs. growth rate): 3 minutes. CFO-ready narrative for the slide: 4 minutes. Total: 2.5 hours. Same 24% IRR. Same assumptions. Same judgment calls about what the business is worth and why. Just 11.5 hours back in my week. What actually changed — and what didn't This is the part people get wrong about AI in finance. The model didn't change. The thinking didn't change. What changed is that I stopped doing the parts of the job that never needed my brain in the first place. Pulling comps from a database isn't analysis. Building a sensitivity table cell by cell isn't insight. Reformatting a spreadsheet for a deck isn't strategy. These tasks exist because someone had to do them — not because a human was the right tool for the job. What I still own completely: the entry multiple assumptions, the growth rate views, the terminal value logic, the judgment call about whether this particular business deserves to trade at a premium or a discount to its peers. That's the job. That's where being wrong costs real money. The rest? I'm happy to hand it off. The broader point about valuation Here's what I've come to believe after doing this for a while: valuation is less about the model and more about the story the model tells. Investors don't approve acquisitions because the DCF said yes. They approve them because the assumptions underlying the DCF make sense, the sensitivity analysis shows resilience under realistic stress scenarios, and the person presenting it clearly understands the business well enough to defend every input. The model is just the proof of work. The thinking is the actual product. If I'm spending 14 hours on proof of work, I have less time for the thinking. That's not a trade-off I'm willing to make anymore. AI-augmented FP&A doesn't replace the analyst. It removes the ceiling on how much an analyst can actually analyze. And honestly? That's the only version of this technology worth caring about. Sources Berkus, Dave. "After 20 Years: Updating the Berkus Method of Valuation." Angel Capital Association. The method was originally published in Winning Angels by Amis and Stevenson (Harvard Business School Press, 2001), with Berkus's permission. Payne, Bill. The Scorecard Valuation Method, as described in industry summaries including Allied Venture Partners, "Berkus Method vs. Other Valuation Models." Disclosure: I used an AI assistant for research synthesis and copyediting. The analysis, examples, and final wording are my own. I take full responsibility for the claims made here.