The modern back office has a peculiar talent for making expensive people perform cheap magic tricks. Copy this number. Chase that invoice. Check the portal. Reconcile the payment. Update the spreadsheet that exists because the other spreadsheet cannot be trusted. Saurabh Chauhan has spent much of his career close enough to this machinery to hear every squeak. Peakflo, the company he co-founded in 2021, is his attempt to make the machine do more of its own turning. The ambition sounds technical; the irritation that powers it is familiar to anyone who has watched a routine task bounce between three systems and four calendars.
His route to the problem was neither direct nor especially tidy. It ran from a business degree in Delhi to McKinsey, then through Southeast Asia’s Rocket Internet ecosystem, where growth plans met the less cinematic facts of cash collection and supplier payment. He managed Daraz in Sri Lanka, later led the beauty-services marketplace Vaniday, and saw a recurring contradiction: companies could sell online at speed while their finance teams still carried the administrative burden by hand.
The observation was not that spreadsheets were primitive. Spreadsheets are durable precisely because they are flexible. The problem was the labor surrounding them: reminders, approvals, document checks, portal logins and reconciliations. In 2022, Chauhan estimated that 99 percent of Peakflo’s customers had arrived from Excel-like environments. Each new customer was a small vote for software that understood the workflow, not merely the cells.
The idea that survived the interviews
At Entrepreneur First in Singapore in 2020, Chauhan arrived with two possible directions. One concerned customer acquisition in social commerce, then enjoying the sort of attention that makes sensible people feel late. The other concerned finance automation. He interviewed 30 to 40 operators. The fashionable idea failed the interrogation: customer acquisition costs were not falling in the promised way, while quality problems and leakage kept appearing. The finance problem remained stubbornly, almost impolitely real.
That is the first useful piece of his story. Conviction is not an instruction to ignore evidence. Chauhan had enough of it to ask uncomfortable questions and enough distance to discard the answer he wanted. Finance automation won by being boring in public and painful in private.
“The finance function is not a monolith.”Saurabh Chauhan
Entrepreneur First also solved the co-founder question. Chauhan wanted a program with a dense pool of technical founders. There he met Dmitry Vedenyapin, a computer scientist and former AirAsia data executive. Chauhan brought the scar tissue of consulting and operating companies. Vedenyapin brought deep technical experience. The fit was less a meet-cute than an architectural decision: pair someone who knew where the work jammed with someone equipped to rebuild the track.
Build the plumbing before the intelligence
Peakflo began with accounts receivable. The team built collections automation, dispute management, branded customer portals and payment reconciliation. It took more than a year, from late 2020 into 2022, to assemble the full stack. There was no instant leap to an all-knowing finance brain. There were modules, edge cases and the grim dignity of invoices that refused to match.
Then customers redrew the map. They did not want separate systems for money coming in and money going out, so Peakflo added accounts payable: invoice capture, purchase-order matching and supplier payments. That work took another year and reached the market by late 2023. Product strategy, in this version, was not a founder revealing the future from a mountaintop. It was customers declining to maintain a second login.
The product sequence / workflow before wizardry
Y Combinator’s Winter 2022 batch enlarged the ambition. Peakflo announced a $4.1 million seed round that year, with backing that included Y Combinator, Rebel Fund, Soma Capital and Amino Capital. At the time, the company was adding customers and hiring into a market defined by the slow migration away from Excel. By 2026, Y Combinator described Peakflo as serving more than 100 businesses across receivables, payables, expenses and operations.
When the workflow starts talking
The company’s next turn came through AI. Participation in Google’s accelerator helped move Peakflo from conventional software automation toward agentic workflows. The distinction matters. Traditional software offers a place where a person performs a process. An agent is asked to carry the process forward: log into a client portal, submit an invoice, compare records across an ERP, make a contextual collection call, record the result and decide what should happen next.
Consider the collection call, the sort of business interaction nobody frames for the lobby. Peakflo’s voice agent can enter an existing sequence of email, SMS and WhatsApp reminders. It knows which invoices are late, whether a dispute is open and whether a promised payment date has passed. The point is not a voice that sounds dazzling in a demo. The point is a call that knows why it exists.
Where human attention goes as the workflow matures / conceptual
Chauhan’s larger thesis is organizational. In his telling, people spend too little of the day on the judgment that makes them valuable and too much on repetitive execution. Agents change the manager’s job from distributing every click to supervising outcomes, exceptions and risk. It is an appealing idea, but Peakflo’s earlier history supplies the necessary caution: an agent is only as useful as the workflow beneath it is understood.
That history is why the sequence matters. Collections before payables. Payables before general orchestration. Customer interviews before conviction. The company did not begin with a synthetic colleague and search for employment. It began with work and gradually taught software how to complete more of it.
The orchestrator leaves the office
In 2026, Peakflo open-sourced 20x, a desktop application its teams had used to connect task systems with AI agents. A Linear ticket can be triaged, assigned to a coding agent, developed in an isolated worktree, paused for approval and turned into a pull request. Operations tasks can follow the same orchestration pattern through systems such as HubSpot and Peakflo. The repository is MIT-licensed and local-first, with human approval built around sensitive actions.
The important feature is not that an agent can perform one impressive trick. It is that work arrives with context, moves through a repeatable route and leaves an audit trail. Peakflo calls its reusable instruction templates “skills”; they can be updated after tasks and accumulate confidence over time. The humble runbook, long sentenced to a folder nobody opens, begins to write notes to itself.
Do not begin with the agent. Begin with the queue: where work arrives, what context it needs, when a person must approve it and what counts as finished.
This is also where Chauhan’s biography becomes unexpectedly coherent. McKinsey showed him dysfunction at enterprise scale, though largely from the adviser’s chair. Daraz and Vaniday put him inside operating systems where recommendations had to survive contact with payroll, suppliers and cash flow. Entrepreneur First helped him locate a technical counterpart. Y Combinator pushed the scale of the company’s ambition. Each stop added a different view of how work becomes stuck.
Mint ice cream, medieval campaigns
Founders are frequently flattened into their companies, as though a cap table consumes hobbies by statute. Chauhan has offered a few useful acts of resistance. During a six-month aviation engagement in Alaska, he worked weekdays and hiked on weekends. He remembers the stretch warmly. His television comfort is Modern Family. His ice cream order is mint.
Fittingly, he has been obsessed with Age of Empires since 1998. The game rewards resource allocation, careful sequencing and the ability to notice that one villager has been standing beside a tree for twenty minutes. It would be silly to claim that a childhood strategy game produced a finance-software company. It would be equally silly not to enjoy the resemblance.
Peakflo’s story so far is a campaign against idle work: an invoice waiting for a nudge, a payment waiting to be reconciled, a task waiting for context. Chauhan’s aspiration is to extend the engine beyond finance into back-office functions where the same pattern recurs. The destination is not a world without people. It is a workplace where fewer people spend the afternoon copying a number from one box into another, then apologizing to a third box for the delay.
There is no romance in three-way matching, which may be its commercial advantage. Glamour attracts tourists. Operational pain attracts return customers. Chauhan chose the latter, built patiently from one side of the ledger to the other, and now wants the workflow itself to pick up the work. The spreadsheet can stay. It may finally get an evening off.