Meet the Founder Using AI Agents to Automate Billing

Manav: In this episode of Emerging Founders with Manav, I interview David Pang. He's building the AI agent for B2B billing at Timely, and he also created another company called SheetWiz, which has over 1.8 million shortcuts and counting. David is solving his own problem — when he was creating invoices at his previous company, he realized he was wasting a lot of time on the process, and with Timely, he's automating that with AI agents. In this podcast we talk about everything he's focusing on, how he raised money from investors, and how he got accepted into YC.

David, how are you doing today?

David: Hey, I'm doing great, thanks for having me on the show.

Manav: Before we get into Timely, I'm really curious about you as a person — I want to understand your prior venture, SheetWiz. Can you explain what problem you were solving there?

David: SheetWiz is a Chrome extension that brings Excel shortcuts and functionality to Google Sheets. We took shortcuts as simple as inserting a row or column, or resizing a column or row, and brought those Excel-specific shortcuts over to Sheets. We also support more complex functionality, things like trace precedents, trace dependents, auto-color, format cycling — features that exist in Excel plugins finance and consulting professionals rely on — and brought that over to Sheets, because there's a huge population of ex-consultants, ex-bankers, ex-Big Four accountants who grew up using, or started their careers using, Excel, but are now forced to use Google Sheets wherever they work now.

That was actually my own experience — I started my career in tech private equity before jumping into a startup, and when I made that jump and suddenly had to do financial modeling in Google Sheets, I lost maybe 80% of my productivity overnight. So the tool we built was really a way to solve that specific problem.

Manav: Let's move into the founding story of Timely — what problem did you see in the market, and can you explain in simpler terms what Timely does?

David: Timely builds AI agents for B2B SaaS finance teams. There's a ton of complicated, manual finance workflows that exist in basically every finance team today, and we build agents that finance teams can configure to solve their specific problems.

Why did we build this? It was a problem I faced personally at a startup I was at called OneSignal — I was there for over two years, and during that time, both my co-founder Peter and I were building and running billing for OneSignal's enterprise customers. When I say "running billing," it was basically me and a spreadsheet, doing a lot of manual calculations. Doing that work, I genuinely thought, I shouldn't have to be doing this — this isn't work a human should need to do, an AI could handle this just as well. When AI agents started becoming a real, viable technology, back in late 2023, early 2024, we saw a genuine opportunity to use this new tool to solve a problem that plagues almost every company in the world.

Manav: One problem I've personally experienced was using QuickBooks and creating invoices — it was such a pain, creating essentially the same thing over and over every day. With Timely, are you going to integrate with tools like QuickBooks, or what kind of tools are you connecting to, and how exactly are you helping people save time?

David: We launched our first AI agent back in August, and it focuses on connecting where your contracts are stored to where you're sending out invoices — exactly the flow you described. Your contracts probably aren't being generated in QuickBooks — more likely Salesforce or HubSpot, maybe signed through DocuSign, PandaDoc, Ironclad, systems like that. Our agent takes the PDF copy, or whatever metadata is available, reads and understands it, and then figures out how to create the invoice for you in your end system, whether that's Stripe, NetSuite, QuickBooks, Chargebee, Maxio, whichever billing system you use. Instead of manually clicking around in a UI, you just review the changes we're proposing, hit approve, and we make those changes for you in your billing system of choice.

We've found that, especially for B2B billing, this is actually a very complicated, time-consuming process — 20 to 30 minutes per single contract, because these are heavily negotiated, nuanced contracts that require someone with real system knowledge to read and process. That becomes especially overwhelming at the end of a quarter or month, when a bunch of deals are closing, sales is chasing quota, and suddenly, as the accountant, you're getting pinged constantly, processing everything late into the night, on top of everything else you need to do to close the books. So that's our first problem to solve — an agent that helps with invoice configuration from contracts.

Manav: Would you mind explaining what an AI agent actually means?

David: The way I think about an AI agent is something that can take information from somewhere, understand it, form a plan of action, and then execute on that plan. So, pull a contract, read it, extract the amounts and products, identify those same products and amounts in another system, and then make an API call to actually create the invoice — that would be an example of an agent acting end to end.

Manav: Would you mind talking about the pricing structure and business model at Timely?

David: It depends on the use case, but today it's more of a direct-sold motion, and we determine pricing based on the specific client we're working with. We always think about the amount we charge in terms of ROI for the customer — we don't want to overcharge relative to the value they're actually extracting. We generally aim for somewhere around 20 to 30% of the value we're generating for them, so we run something like an ROI calculation: how much time are we saving you, how much more visibility are we giving you into your own tools, are you collecting faster because invoices go out faster, are we enabling sales to sell more complex billing structures because our agents can handle that complexity instead of a finance person spending hours on it manually.

I know these things aren't always perfectly calculable, but we always try to think about pricing from that ROI lens rather than just saying "we charge X dollars per contract processed" — that doesn't really do justice to the value being created.

Manav: The biggest thing I've realized is that, essentially, if you can replace a full employee — which is anywhere from $70,000 a year minimum, more if you're in New York or California — how do you even justify replacing one person with just a subscription, right? Even if you charge $500 a month, that's still a great deal for the company. I feel like that's why these AI companies are making startups leaner and leaner — even in my own companies, we used to have 18 or 20 employees, now we're down to two or three, we've gotten so lean because of this shift.

One of the biggest things for me — I was listening to the Replit founder's podcast, and he mentioned they actually got rejected by YC four times before getting in. We also got rejected by YC before, and when they finally got in, it was this insane three months where they went all in, put in a crazy number of hours. How was your journey getting into YC, and how much work did you put in?

David: We were rejected for Summer '23 — we pitched SheetWiz at the time, not Timely. The feedback, effectively, was: it's a Chrome extension, how is this going to grow into the next Airbnb, Stripe, whatever the reference company was at the time? We'd just launched, hadn't even monetized the product yet, so it was a harder sell.

We then applied again in Summer '24 with Timely, once we felt like the problem and the underlying technology had reached a point where we could genuinely go after it. By that point we'd also grown SheetWiz quite a bit — we'd started charging for the tool, only $4.99 a month, but still, we were charging. So we could go back to YC and say: my co-founder and I have worked together, we're also co-workers from another startup, we're also roommates, we know each other extremely well, we know our working styles, we've already built and monetized a business, and now we want to go after a much bigger opportunity, using AI to do it. That was a much more compelling pitch than SheetWiz alone had been, which at the time probably had around 10 users.

Manav: What advice would you give to other entrepreneurs starting a company?

David: There's a lot of advice out there, and honestly, some of it shouldn't necessarily be taken at face value. For example, you hear a lot about ACV for SaaS businesses, and this idea of a "dead zone" —

Manav: Can you explain what ACV is?

David: Average contract value — basically the size of the deal.

Manav: Got it.

David: If you're above $20K, for example, it's harder to expense on a credit card, so you have to go through a more formal process, but that's fine because the deal's big enough to justify it. If you're below $20K, it's harder to justify hiring salespeople, since their commission on smaller deals isn't very motivating. So you end up in this "dead zone" — too big to expense on a credit card, too small to properly incentivize a sales team. You hear that advice over and over: get your ACV above 20 or 25K, push those values up.

The reality is, I think that's really a later-stage company problem — a hotter take, maybe, but: start with the customers who are actually willing to pay you, that's step one. Once they're paying, there's more value you can build on top. But if you're rejecting customers early on just because their contract value falls below some arbitrary threshold, that might be worse than just taking the deal — customers are better than no customers, and you can use what you learn from working with them to eventually build something that genuinely deserves a higher contract value. That's an example of advice that maybe shouldn't be taken too literally, especially if you're an early-stage startup.

Manav: That's great advice, in my opinion — I think a lot of people, especially early on, just want to see some traction, some kind of win right away, and that dopamine hit from landing a customer can be exactly the fuel that keeps a company going. I really love that advice.

Awesome, David, thank you — we had such a great talk about AI agents. We'll do a recap video in a year or two and see where our journeys have led us. Thank you so much for coming on the show, I really appreciate you.

David: Yeah, thanks for having me. Please make sure to click that subscribe button if you got any value from this, and share it with a friend — that's all I ask.

Manav: Thank you so much for watching, I'll see you in the next episode.