Sidharth Kakkar and the trouble with being indispensable

After selling his first company, Sidharth Kakkar wrote himself a 35-page document about what he had learned. It was a substantial piece of homework for someone who had already persuaded a buyer to acquire the business. Success, apparently, was no excuse to stop marking his own mistakes.

One lesson concerned the founder himself. He had begun as the sort of manager who got into everything. In retrospect, he could see the cost. Being useful at every turn had a way of making everyone else wait at the same corner.

At Subscript, the finance software company he co-founded with Michelle Lee, he tried a different arrangement: people would have the information and authority to make decisions without waiting for him. Required internal meetings went. Geography became less important. The founder could still have opinions. They would have to compete with other people's judgment.

The experiment is a useful way into Kakkar's career. He has built classroom software, subscription finance tools and, more recently, Woodrow's AI agents. Across those changes, one question keeps resurfacing: how much of a person's day is being spent getting around the machinery that was supposed to help?

A classroom before a boardroom

Freckle began as Front Row Education in 2013. In its early development, a teacher let Kakkar and his colleagues into a classroom in Baltimore. They watched children use the software, noticed where they got stuck, and went home to program changes at night. The next encounter with the classroom could test the previous evening's confidence.

That is a fairly unforgiving product review. A child struggling with an exercise has little reason to admire the elegance of the code underneath it. The useful question is whether tomorrow's version makes learning easier.

Kakkar kept himself close to customer support as the company grew. In 2016, he described answering teachers himself and treating those exchanges as a particularly valuable use of his time. The person setting priorities would hear directly when the priorities had missed something.

The educational premise was practical: teachers needed help providing suitable work to children at different levels. Freckle offered differentiated practice across subjects, giving teachers tools to respond to those differences. It belonged in the working classroom, with the teacher still involved.

When Renaissance announced the acquisition in May 2019, Freckle was serving more than 700,000 teachers and 10 million students across 75,000 schools. Those figures describe the scale of the classroom experiment. They also explain why a founder might leave with plenty to reconsider about the organization behind it.

The spreadsheet came along for the ride

Building an education business had given Kakkar another education: understanding a growing company's numbers could become a job in itself. He later described a sprawling Google spreadsheet, with multiple tabs and people editing it, as the place where his previous business kept track of itself.

The irritation was uncertainty. Had someone changed a number? Had the analysis allowed for every exception? A spreadsheet can produce a very handsome answer while leaving those questions politely unresolved.

Subscript addressed that problem for subscription businesses. Its work connected the numbers finance leaders wanted to examine with the underlying financial processes. Billing, revenue recognition and analytics gradually became parts of the same product offering.

For Kakkar, this was familiar territory approached from the other side of the desk. He had been the founder trying to understand the business. Now he was building software for the people expected to explain it.

In February 2025, Subscript announced a $15 million Series A led by Pelion, taking its announced total funding to $21 million. The company named customers including Pave, AcuityMD and EliseAI. Its pitch centered on bringing billing, revenue recognition and financial analysis together, reducing the reconciliation work between them.

The fundraising figure is one milestone. The more revealing ambition is the unglamorous one underneath it: a finance team should be able to spend less time reconstructing what happened and more time deciding what to do next.

The boss learns to get out of the doorway

The organizational experiment began with a simple complication. Kakkar and Lee lived across the country from each other and preferred different working hours. A globally distributed team would multiply those differences. A convenient meeting for one person could be an alarm clock for another.

In his description of Subscript's early culture, employees could work when they were productive. One engineer, Brandon, liked dancing after lunch. Kakkar saw little value in keeping someone at a laptop during an unproductive stretch simply to preserve the appearance of a working day.

Social gatherings survived, including games. Customer conversations survived too. The change concerned internal coordination: work would need to move without routinely assembling everyone at the same time.

That required more than deleting calendar invitations. If decisions still needed the founder's approval, the company would merely exchange meetings for a queue.

“Everyone who works for me is better at their thing than I am at their thing.”

Sidharth Kakkar, on delegating at Subscript

His later account of an engineer tidying a troublesome part of the codebase makes the idea concrete. The engineer documented the problem, identified priorities and carried out changes without being assigned the project. Kakkar valued the judgment involved: someone had noticed useful work and taken responsibility for it.

He was also willing to consider that an employee's unexpected decision might expose missing information from him. The founder knew things other people did not. Sharing that context was part of his job. Intervening in every choice could conceal the more basic failure to explain where the company was going.

It is an awkward adjustment for someone promoted, or self-appointed, because they are good at details. The same habit that once made a person effective can make a team dependent. Kakkar's account has room for that embarrassment. He includes himself among the managers who had to learn again.

Writing down what the room forgets

Subscript's alternative put a considerable burden on writing. Weekly context arrived through documents and recorded video. A shared wiki preserved the reasoning behind decisions. A running journal traced the company's search for product-market fit, making earlier assumptions available to later employees.

Writing gives an absent colleague a chance to catch up. It also makes a decision available for inspection after the confidence of the moment has worn off. A meeting can end with everyone remembering slightly different versions of the same agreement. A document has the modest advantage of continuing to say what it said.

Kakkar did not present the arrangement as a universal answer. People who wanted more synchronous interaction might reasonably prefer another company. An unusual workplace needed employees who actively wanted its particular freedoms and responsibilities.

That qualification matters. Freedom to choose the work includes the obligation to work out what matters. Without that appetite, an empty calendar could feel less like liberation and more like being left alone with a map whose destination someone forgot to circle.

A ball machine, then a finance workflow

By late 2025, Kakkar was describing another learning project: tennis. He enlisted friends, hired coaches and practiced with a ball machine. Each offered something different, from correction to repetition. He connected that mix of instruction and practice to his efforts to understand AI.

He used the tools himself and watched other people use them. His year-end reflections acknowledged both the progress of coding agents and the limitations of language models, including their tendency to lose context. The useful knowledge came from observing actual behavior, including the failures.

His advice to finance leaders had a similar scale. Try a specific task. Compare results. Learn enough to judge the next tool sensibly. An August 2025 essay encouraged small experiments with suitable data rather than waiting for a perfect system or a grand rollout.

The approach sounds familiar after Baltimore: put something in use, observe what happens, and revise the idea. The users and the consequences have changed. The willingness to have an assumption corrected remains visible.

In 2026, Kakkar publicly launched Woodrow for enterprise finance and operations teams. His announcement described agents carrying out repetitive tasks across approved systems, including reconciliation and accounts payable or receivable work, while recording their actions.

That is the company's proposed role. It also raises the bar on his longstanding interest in removing busywork: software taking action needs to earn confidence through what it does and the record it leaves behind.

Instant coffee, considered carefully

Kakkar's recent writing wanders beyond finance. An August 2026 essay begins with his affection for instant coffee and the judgment it attracts. He follows the subject into an argument about how inherited choices shape preferences. Even a cup of coffee becomes an occasion to question whether the standard arrangement deserves its status.

In another essay that month, the experiment was a poker practice tool. He had initially chosen Heroku for Woodrow because he wanted to avoid infrastructure work. Later, building the hobby project, he asked a coding agent to deploy it on a Hetzner server. The experience made him reconsider what he had been paying to avoid learning.

“My fear of devops was a tax on ignorance, and ignorance just got cheap.”

Sidharth Kakkar, August 2026

He distinguished a small project for friends from mistakes that could expose customer data or lose money. The pleasure was in discovering that an avoided task had become approachable, while recognizing that the consequences still mattered.

There is something pleasantly consistent here. The founder who once wanted to inspect the details now keeps testing which details require his attention. Sometimes the answer is a colleague. Sometimes it is software. Sometimes it is a ball machine patiently returning another serve.

His 35 pages of lessons were an early installment in that process. The current writing suggests the document could use an appendix. Preferably one that nobody has to schedule a meeting to discuss.