How Agentlink Is Turning CPAs Into Advisors at Scale

Nilanjan Raychaudhuri has spent his career in software. He worked alongside Martin Odersky at Typesafe on the Scala language, wrote a book about it, contributed to Apache Spark, and later founded Tublian, a social network that grew to around 30,000 developers across India and Africa. When AI arrived, he wanted to put it to work on a problem that could actually be measured, something where every claim eventually met the bank account. That led him to accounting, and to a bet that might surprise people watching the space.

Most AI finance tools are built to go around the accountant. Agentlink, the company Nilanjan is building now from Columbus, Ohio, is an AI controller designed to sit inside CPA and fractional CFO firms, not replace them. The pitch is simple: as AI commoditizes reports and answers, the firms that survive will be the ones with real relationships, and Agentlink exists to help them extend that attention to every client on their book.

In the conversation below, Nilanjan walks through how the product found its market, why he treats senior CPAs as teachers rather than prospects, and what he’s learned building a trustworthy system on top of an unpredictable technology.

Tell us about yourself and your business.

I’ve been building software my whole career. I worked at Typesafe with Martin Odersky and the team behind the Scala language, wrote Scala in Action, spoke at conferences around the world, and contributed to open source projects like Apache Spark.

Then I founded Tublian, a social network for software developers. Our belief was simple: great talent is everywhere, but opportunity isn’t. Resumes work fine if you’ve got years of experience, but recent graduates don’t have much to put on one. So we built something like a credit score for developers, a way to show what you can actually do. Tublian grew to around 30,000 developers, mostly in India and Africa.

Today I run Agentlink. As AI takes over the routine parts of accounting, firms that only deliver reports and answers become easy to swap out. What clients will pay for is a relationship: a firm that knows their business and sees what’s coming. Agentlink helps CPA and fractional CFO firms build that with every client, not just the few they have time for. It’s an AI controller that keeps the firm close to each client’s numbers between calls, so the firm can offer real advisory work across its whole book and stay ahead of the shift.

How does your business make money?

Firms are our customers, not the businesses they serve. We charge a monthly subscription priced per client, with lower rates as a firm connects more of its book. The product runs under the firm’s own brand, so the relationship stays theirs.

What was your inspiration for starting the business?

Honestly, Tublian helped a lot of developers, but I never figured out how to make money from it. When AI arrived, I wanted to build something with a real business model, and I wanted to bring AI to small and medium businesses, not just large companies with dedicated AI teams.

I chose accounting and finance because it’s hard, and because it can be checked. AI is unpredictable by nature, but in finance, every claim eventually meets the bank account. You find out whether you were right. Building trustworthy systems on top of an unpredictable technology was the challenge that pulled me in.

How and when did you launch the business?

I built the first version over Christmas and started building properly in January 2026. At first I was building for business owners, a forecasting tool they’d use directly. But talking to owners, I learned most of them don’t want to do finance at all. They hand it to a fractional CPA or CFO.

So I started talking to those firms, and that’s where the real opportunity was. What began as a forecasting tool for owners grew into an AI controller for the firms that serve them.

How is the business funded? 

Agentlink is bootstrapped. We haven’t raised outside money. JumpStart Ventures in Cleveland and Bounce Innovation Hub have backed us with support and connections, which has mattered a lot for a solo founder in Ohio.

How did you find your first few clients or customers?

I’m not an accountant, so I didn’t go looking for customers first. I went looking for teachers. I wanted design partners, firms willing to work closely with me so I could understand how they actually operate.

The first came through my network and JumpStart. Then I went to my first accounting conference, which was a big step for me. I went where CPAs and fractional CFOs spend time, built relationships, and found my next two design partners there. All three still shape what we build every week.

What was your first year in business like?

The hours are crazy. I’m a solo founder with a small team in Africa, people I met through the Tublian network, working on a very big problem. So from day one my priority was using AI and AI agents to do the work of a much bigger team.

The biggest surprise came from chasing context. AI is only as good as the data it sees, and I realized the books and the bank only tell half the story. The other half lives in conversations, email threads, and people’s heads, and it walks out the door when someone leaves. That’s ironic, because this whole profession is built on the ledger, one of the oldest ways of recording things in a structured way. But nobody records the decisions.

So we built a ledger for decisions: what was decided, what we expected to happen, and whether it did, checked against the bank. I wasn’t looking for a moat, but that turned out to be one. It’s data QuickBooks and the banks don’t have, and it’s what makes the product get better over time.

The market moves incredibly fast, and plenty of AI tools are trying to automate accountants away. But the more time I spent with CPAs and fractional CFOs, the more I saw what they really give a founder: judgment about the business. So we decided early to strengthen that judgment, not compete with it. That became the most valuable thing in the product.

What strategies did you use to grow the business?

With Tublian, growth was easy because I knew where developers hang out: GitHub. CPAs and fractional CFOs have no GitHub, and it took me a while to learn that they gather in professional bodies like the AICPA and IMA, at industry conferences, and on LinkedIn. We now run several channels in parallel.

The first is content written for AI search. Competing for Google rankings in accounting is nearly impossible against the big players, so we write question-shaped guides that answer exactly what practitioners ask ChatGPT and Claude. We’re now seeing traffic arrive from those assistants. I think this is the biggest opportunity any startup has right now, because these tools don’t care about old-school domain authority. They care whether you answered the question.

The second is LinkedIn, relationship first. Cold asks for fifteen minutes and a demo didn’t work for me, and paid LinkedIn is too expensive, so I use automation to stay close to the right people and build relationships slowly.

The third is education. I’ve been lucky to have a group of finance leaders around me who are connected to the IMA, and we’re working on sessions that earn CPE credit. That puts us in a room of CPAs and CFOs as teachers rather than vendors.

Email is the newest channel, so I don’t have results yet. With so much AI-generated content out there, I’ve learned that the only thing that cuts through is being authentic and specific. My perspective is unusual: I’m a software engineer who came into finance from the outside, and I’ve watched AI change my own industry first. So instead of asking for a demo, our emails offer something useful, like a playbook on how firms can adapt to AI and a white paper on why client advisory doesn’t scale and how it can. If it helps, people reply, and the relationship starts there.

What was the biggest challenge you had to overcome?

Distribution has been the hardest part. When I get in front of a firm and show the product, people see the value. Getting in front of them in the first place is the challenge, especially as an outsider to the profession.

The second challenge is technical: building something reliable and trustworthy on top of AI, which is unpredictable by nature. It took me a while to find where AI belongs and where it doesn’t. We call Agentlink an AI controller, but it isn’t an AI wrapper. The core is plain code that does the arithmetic and keeps the record, and we use AI only where it truly adds something, like understanding language. Making those calls is hard when the models get better every month and the line keeps moving.

What have been the most significant keys to your business’s success?

Without question, the people. A handful of CPAs, fractional CFOs, and finance leaders have been incredibly generous, teaching an engineer how their profession really works and being honest about their problems. Their willingness to educate me has shaped everything we’ve built. I wouldn’t be here without them.

The other key has been building trust from the start. Take the client channel, a private assistant each firm offers its own clients, where they can ask about their business and get an answer anytime. It sounds simple, but it’s a huge trust problem, because it carries the firm’s name. Two principles from our design partners shaped it. First, it’s fine to fail, but never assume. Second, don’t just give the answer, show how you got there. So every answer traces back to the data it came from, and anything that needs judgment goes straight to the firm instead of getting a confident guess. That was much harder to build than I expected, and it’s what I’m proudest of. Knowing where its limits are is what makes a firm willing to put its name on it.

Tell us about your team.

We’re tiny: me, a part-time designer, a part-time developer, and hundreds of AI agents. I’m in Columbus, and the team works remotely from Africa. I found them through the Tublian network.

Around us is a group of senior CPAs, fractional CFOs, and finance leaders who act as design partners and mentors. They aren’t on payroll, but they shape the product as much as anyone.

What is the most important lesson you’ve learned growing the business?

Building is getting cheaper every month. With AI agents, anyone can brainstorm and ship a feature in a day. What you can’t automate is understanding the market. One conversation with a practitioner teaches me more than a hundred hours of building with AI.

I’m an introvert engineer. My natural state is headphones on, building. But every customer conversation, even the awkward ones, the no’s, the calls that never turn into a sale, has made the product better. Talking to customers is the work.

What separates your business from your competitors?

Most AI finance products sell straight to business owners and go around their accountant. We think that’s the wrong bet, and it comes from how we see the industry changing.

A model on its own won’t take over an accounting firm’s work. It answers when asked, holds nothing between questions, and has never watched the business run. But a product built on a model will try, and it will win every client who only ever got reports and answers from their firm. Clients already ask ChatGPT for a second opinion on everything. The firms at risk are the ones whose relationships are purely transactional.

So the question for a firm isn’t whether to use AI. It’s how to move every client from getting reports to having a firm that knows their business and what’s coming next. That’s what Agentlink is built for. Agentlink keeps the firm in the loop: the firm’s name is on it, the firm reviews and stands behind what matters, and every client gets that level of attention, not just the few a firm has time for.

What are your future plans for the business?

We believe that by 2028 there will be far more small, independently run businesses than there are today, because AI makes it so much easier to start and run one. More businesses means more demand for good finance and accounting. Our near-term goal is to help CPA and fractional CFO firms meet that demand: to be the AI controller that lets a small firm double or triple its client base without adding staff, while giving every client high-quality advisory attention.

Longer term, something bigger is forming. Every decision we record and check against the bank builds a verified history of how a business really runs. Today a small business needs years of financial statements to prove it’s creditworthy. A verified record could do it in months. We think that becomes a better way for small businesses to access capital, with their firm standing behind them.

What are some of your favorite books, blogs, podcasts, or YouTube channels?

• Books: Conscious Business by Fred Kofman and Inference Engineering by Philip Kiely
• Blogs: Packy McCormick’s Not Boring and the a16z blog
• Podcasts: I host my own, the 10x Builder podcast, where I talk with builders about what they’re shipping