Jul 6, 2026 · 57m · product-market-fit
He raised $75M—but kept his team to 21 people & pays everyone the same salary. | Cos Nicolaescu, ... · PMF Show
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The Product Market Fit Show, Accrual co-founder Cos "Kaz" Nicolaescu shares how the enterprise AI accounting platform achieved product-market fit, raised $75 million, and scaled effectively while maintaining a lean 21-person team and an equal-salary compensation model.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Pablo holds 15.7% of the talking time here. How this is scored →
Pablo as informed peer 5.3 Guest teaching 6.2 Guest disagreement 1.7 Pablo pushing back 2.1
speaking balance: gold is Pablo, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 45:20 Forceful rejection of traditional startup headcount scaling
Kaz strongly rejects the conventional VC expectation that Accrual should have scaled to 50 or 100 people after raising large capital, arguing it would have severely crippled organizational effectiveness.
Hardest push from Pablo ▶ 6:09 Challenging customer feature requests versus true pullPablo presses Kaz to clearly separate genuine customer product-market fit signals from the common enterprise trap where prospects endlessly demand 'one more feature' before buying.
Biggest teaching moment ▶ 40:42 Reframing enterprise sales champion strategyKaz debunks standard enterprise sales advice focused strictly on CEO executive sponsorship, laying out why practical deployment requires targeted multi-tier engagement across service line heads and operational managers.
Pablo holds their own ▶ 14:15 Hypothesizing full-context AI requirementsPablo demonstrates deep product insight by proposing that Generative AI uniquely revitalizes all-in-one software architectures because LLMs require end-to-end workflow context to deliver true value.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Pablo as informed peer | Guest teaching | Guest disagreement | Pablo pushing back | Why |
|---|---|---|---|---|---|---|
| Raising a $75 Million Series A in the AI Era | 5 | 5 | 1 | 2 | Pablo kicks off by discussing the ballooning size of Series A rounds in AI and defines true product-market fit. Kaz explains how Accrual moved from unproven hypotheses to enterprise validation during tax season with massive efficiency gains. | |
| Indicators of True PMF and Customer Expansion Pull | 6 | 5 | 1 | 3 | Pablo pushes Kaz to differentiate true customer pull from the common trap of prospective buyers asking for 'one more feature'. Kaz clarifies that Accrual's expansion pull comes from existing paying customers requesting broader workflow coverage like audit and engagement letters. | |
| Eliminating the Fragmented Multi-Tool Accounting Stack | 4 | 6 | 1 | 1 | Kaz details the heavily fragmented accounting stack where top firms juggle up to eight disjointed tools across client intake, document storage, and signing. Pablo listens as Kaz explains how consolidating these tools delivers major productivity gains. | |
| Selling an All-in-One Vision Before Building Every Feature | 6 | 6 | 2 | 4 | Pablo asks a sharp tactical question about how a startup convinces risk-averse enterprise buyers to replace multiple existing tools with an unbuilt all-in-one vision. Kaz outlines their high-risk approach of selling an operating system vision supported by rapid iteration. | |
| Why Generative AI Demands Full-Workflow Context | 7 | 6 | 2 | 2 | Pablo formulates an insightful hypothesis about how Generative AI fundamentally shifts the all-in-one software value proposition by requiring end-to-end workflow context. Kaz strongly agrees and explains why isolated point solutions and OCR wrappers lose critical domain context. | |
| Kaz's Career Background and Leaving Brex to Found Accrual | 5 | 5 | 1 | 2 | Kaz reflects on his engineering journey from Microsoft to early Stripe and then CTO of Brex, noting his preference for hands-on building over corporate executive administration. Pablo connects Kaz's five-year tenure philosophy to the 'tour of duty' startup framework. | |
| Ideation Strategy and Choosing the Accounting Industry | 6 | 6 | 1 | 2 | Pablo contrasts mission-driven founders like Mark Zuckerberg and Elon Musk with systematic founders who enter ideation without a predetermined problem. Kaz breaks down how he and his co-founder evaluated industries by personal impact, deciding to empower accountants with strategic leverage. | |
| Narrowing the Focus and Selecting B2B Workflow Automation | 5 | 7 | 2 | 1 | Kaz explains their deliberate elimination matrix, discarding crypto, foundational AI models, consumer apps, and hardware to focus on complex B2B workflows where they had high domain competence. He notes choosing accounting over legal simply because they preferred working with accountants. | |
| Pragmatic Founders and Landing Enterprise Launch Partners | 5 | 6 | 1 | 2 | Pablo notes the distinction between naive young founders and pragmatic experienced executives. Kaz explains how partnering early with massive launch partners like H&R Block and Armanino forced Accrual to build high-scale, robust foundational architecture from day one. | |
| Achieving a 100% Pilot-to-Production Conversion Rate | 5 | 7 | 2 | 2 | Kaz details how Accrual achieved a 100% pilot conversion rate by eliminating AI smoke-and-mirrors, offering side-by-side accuracy comparisons on real tax returns, and committing to daily product iterations rather than over-promising on vaporware. | |
| Multi-Level Champion Mapping and Embedded Change Management | 5 | 7 | 2 | 2 | Kaz rejects the conventional enterprise sales dogma of relying solely on top-down CEO buy-in, outlining Accrual's multi-level champion mapping across tax service line leaders, IT security, and mid-level practitioners driving ground-level change management. | |
| Operating Lean: Why Accrual Maintains a 21-Person Team | 4 | 7 | 3 | 2 | Pablo asks why Accrual maintains a lean 21-person team despite raising $75 million. Kaz forcefully argues that large teams generate paralyzing coordination drag and explains that AI tools allow individual '100x engineers' to outperform bloated legacy organizations. | |
| Management-Free Operations, Fluid Roles, and AI Integration | 6 | 7 | 2 | 2 | Kaz describes running an organization with no scheduled meetings, no formal one-on-ones, and fluid cross-functional roles where even sales staff utilize AI coding tools. Pablo observes how rotating engineers across modules eliminates protective domain silos. | |
| Strategic Capital Allocation and Future AI Token Expenditures | 5 | 7 | 3 | 3 | Kaz reveals Accrual's radical compensation model where every employee earns the identical flat salary while holding heavy equity stakes, filtering for high-conviction early-stage builders. He also predicts that token expenditures will soon exceed human payroll costs. |
Statements from this episode (19)
Nicolaescu: Top US accounting firms used 3 to 8 tools per workflow
“Before accrual, the best accounting firms in this country and the most innovative ones would have anywhere from three to four tools to six to seven to eight tools. That's kind of the range.”
Nicolaescu: Multi-player workflow software has far higher switching costs than single-player AI
“I think the single use scenario, like a single player scenario where I'm doing some work with some tools is very different when I'm working as part of a collection of people throughout the workflow and all of us use the same tool. And I think the latter is way…”
Nicolaescu: Software engineers should avoid job hopping every 1 to 2 years
“Unless you're the founder of a company, you shouldn't be at the same place forever. I very much encourage people not to
Jump ship every one to two years, not just because it looks bad, but because to actually see, especially as a software engineer, to see the …”
Nicolaescu: One of Brex's Earliest Hires Was Its CFO
“At Brex, it was from the beginning. One of our first hires was our CFO.”
Accrual Signed H&R Block and Armanino as First Two Customers
“Our first two customers were H&R Block and Armanino.”
Nicolaescu: Enterprise pilots should prioritize sample variety over volume
“When you go through a lot of returns, like if you run like a handful of returns or certain complexity or certain shape, you're not going to learn much more from running like dozens of those. Like you'll get the signal pretty quickly if they're good or not, but…”
Nicolaescu: Masking AI errors during enterprise pilots guarantees post-sale disappointment
“I've heard of pilots where you get something and they have AI, but then they have humans that go and correct things that AI got wrong, and then they give you the final thing. And it's like, well, that's not really the AI part. You should be upfront about that …”
Nicolaescu: Top-down CEO enterprise sales fail at accounting firms
“So the thing that you normally hear with enterprise sales is Find the CEO. They give you buy-in and then you can make decisions. That doesn't really work well because the managing partner or the CEO at an accounting firm is not going to be close enough to the …”
Nicolaescu: Top software engineers are now 100x engineers with AI
“Like there used to be the Tenex engineer before. Now I think those Tenex engineers are a hundred X engineers.”
Accrual operates without management, 1-on-1s, or scheduled meetings
“So we don't have management at this company. I think it's technically everybody reports to meet, but we don't do one-on-ones. We don't have scheduled meetings.”
Nicolaescu: Remote work is awful for new grads and hampers learning
“It's awful to be a new grad and to work in a remote company. I think you just don't learn as much.”
Nicolaescu: Accrual rotates engineers across the codebase every few months
“We move people around every few months. Like if you're an engineer, you're not like I'm an engineer on the client portal. Maybe you work on a client portal for one or two sprints or like a few weeks, maybe a couple months, and then you're going to move on some…”
Nicolaescu says AI breaks standard interviews, requiring multi-day work trials
“We do work trials, for example, where we spend like several days with people, typically like a couple of days with every hire that we haven't worked with to work together through real problems. Part of it is I don't know how to interview in the day of AI, beca…”
Nicolaescu expects AI token costs to rise due to energy constraints
“At the end of the day, tokens represent capacity and we're very constrained on capacity. We don't have enough energy. We don't have enough data centers and the demand is, is exponential. And so if I think about it from that angle, I think actually token costs …”
Nicolaescu predicts AI companies will spend more on tokens than humans
“If you project out maybe a year or two from now, do you think you're going to spend more on humans or tokens? And I don't think the answer is obviously humans. I think there's actually a good chance you spend more on tokens than you spend on humans.”
Nicolaescu: Accrual Founders Take Lower Salaries Than Employees
“Sid and I make less because I think it's only fair for founders to be in that situation, but everybody else makes the same amount of money”
Nicolaescu: Early-Stage Startup Equity Value Is Effectively Zero Today
“The other thing I tell people is the value of your equity today is zero. It doesn't matter what our valuation is. Either we go to zero or we grow significantly more. What the value is today doesn't really matter. So I don't do the math in terms of what the dol…”