Andy Cook and the Trouble with Knowing Too Much

Andy Cook and Nelson Joyce had a problem that looked rather like success. Their project at HubSpot was growing. More engineers were joining. There was work to hand over, and two people who knew how everything worked. Unfortunately, most of the instructions were stored in those two people.

They had built Leadin together, handling the coding, support, and decisions that a small project requires. When their team expanded, the accumulated knowledge between them became difficult to distribute. Every explanation took time. Every unwritten decision needed a narrator. A company can hire another engineer considerably faster than it can copy somebody’s memory.

Tettra grew out of that inconvenience. Cook and Joyce founded it in 2015 to help teams share information. The attraction of the problem was personal: they had already been the people with the answers. Cook’s subsequent career at Tettra would involve a second education, in how easily a builder’s knowledge can become a builder’s blind spot.

The friendship came first. Cook and Joyce grew up in Hanover, Massachusetts, attended the same middle school, and bonded over skateboarding. By the time they were choosing a business partner, they had a history considerably longer than a pitch deck. They had also worked together inside another company. There was less need to guess how the other person behaved when work got difficult.

A rental business, with spelling difficulties

Cook’s earlier venture had begun with his older brother, Alex, who taught him web design. The brothers built an ordering website for a rental store connected to their father’s landscaping business. A practical job became Rentabilities. Cook’s own career profile lists him as its founder and chief technology officer from 2010 to 2013, when HubSpot acquired it.

The name deserves a small memorial. Over dinner with investor Dharmesh Shah, the brothers discussed its deficiencies: too many syllables, troublesome spelling, an unhelpful collection of i’s. Their defense was that their grandmother could remember both the name and what the business did. They called this the Grandma Test. It was an unusually domestic focus group for a software company.

The episode captures something useful about Cook’s early world. Business was close enough to touch: a family connection, a shop needing a better website, a brother who could teach a skill. His University of Massachusetts Amherst years, from 2006 to 2010, included the Commonwealth Honors College and teaching web programming. He also says a video game he built reached Late Night with Conan O’Brien. There were several possible ways for a programming education to escape the classroom.

He wrote, too. Dharmesh Shah published Cook’s conference notes on OnStartups, including observations from the Babson Entrepreneurship Forum and MIT Startup Bootcamp. Before Cook sold software about keeping useful information accessible, he was practicing a modest version of the same thing: attend, listen, write it down, pass it along.

The office experiment that became a company

At HubSpot, Cook and Joyce worked on Leadin, a small-business marketing project. Cook describes himself as having coded and grown its first version, then led a team rebuilding it on the HubSpot platform. His work crossed engineering and product management. He knew the satisfaction of making a tool, and the less tidy work of helping other people use it.

When they left to start Tettra, the founders treated research seriously. They interviewed potential customers and showed mockups before building. Yet the initial product failed to attract the use they expected. It was a particularly rude result for two people who had done the homework. Prospective customers had expressed interest. Interest turned out to be a poor substitute for a habit.

The opening came through Slack. Connecting the product to the place where teams already communicated gave the idea a more useful home. Rather than insist that people adopt another destination, the founders rebuilt around an existing workflow. The lesson had an uncomfortable implication for an engineer: getting the machinery right could still leave the behavior wrong.

Cook’s story makes the early product less an act of prediction than an ongoing negotiation. The customer says a thing sounds useful. The customer then has an ordinary Tuesday, with interruptions, old routines, and several other tabs open. Tuesday gets the deciding vote. Tettra had to earn a place there.

The engineer gets a headset

Cook eventually put down his code editor and took up a phone, headset, and notebook. In his account of Tettra’s early selling, about 150 phone demos produced the first 50 customers. For someone with an engineering background, these conversations were an additional job, and an education that could not be delegated to a spreadsheet.

He described overcoming his own introversion to make the calls. Then one of those hard-won customers left. Knowing the buyer personally made the departure sting. It also made product failures specific: a bug belonged to a person whose work it had disrupted, rather than to a ticket in an anonymous queue.

There is an appealingly unceremonious detail in his account. Early account setup could mean sending a prepared email with a beta link. The invitation sounded more elaborate than the procedure behind it. Cook was spending his effort finding out whether people would buy, while the polished machinery of selling waited its turn.

His notes from those conversations became material for the team. The person on the phone could learn something once, document it, and let a future colleague benefit. A software business about shared knowledge was using its own product to prevent the CEO from becoming the sole repository of customer experience.

Six months, and a great deal of email

The next difficult education concerned money. Tettra came close to running out. Cook and Joyce stopped taking salaries, and teammates accepted temporary pay cuts. Cook later talked through the possibility that the company might fail and that he would have to sell more HubSpot stock to cover his own expenses. Survival required smaller bills and decisions whose benefits would take longer to arrive.

Freemium was one of those decisions: let teams start using the product before asking them to pay for more. The move required patience at a moment when patience was expensive. The route toward a sustainable business involved both changing the offer and cutting the cost of keeping the company alive. There was no single afternoon when everything became easy.

In 2018, Cook published the arithmetic behind a subsequent fundraise. Raising $590,000 took 193 days and 1,482 emails. Of 75 investors approached, 25 committed. The figures are a useful corrective to the ceremonial funding announcement, which tends to begin at the point when the money has already arrived.

Even that arrival had its comic complications. One paper check went to the wrong office because Cook had forgotten to tell the investor they had moved. Another was too difficult to read for remote deposit. Software founders can automate a surprising amount of commerce and still find themselves negotiating with handwriting.

Cook made the process public. Sharing the awkward details gave other founders something more useful than a triumphant number. It showed the waiting, repetition, and administrative effort inside the result. Transparency, in this case, included the mail.

A better answer starts with a better question

Cook’s curiosity could also be inconvenient for the people working with him. In 2018, Jay Acunzo described bringing him listener feedback about Org Uncharted, a show they were making together. Cook welcomed the feedback, then asked whether the listeners might be biased because they already knew Acunzo’s work so well.

The question was small enough to fit into a conversation and large enough to unsettle the method behind it. Asking customers is sensible. Asking which customers, and why those people might be unusually encouraging, is more demanding. Cook had learned that collecting answers and testing an assumption were different amounts of work.

There is a gentler version of that attention in NanaGram, a side project he and Alex started for their grandparents. Family members texted photographs to a dedicated number; prints arrived by mail. Alex recalled returning to find pictures covering the refrigerator and walls. Their grandmother began recommending the service to visitors. The original Grandma Test had acquired a distribution department.

The family project and the workplace product solve different problems, but both concern getting something valuable to the person who would otherwise miss it. A photograph buried in a phone and an instruction buried in chat are stored perfectly well. Delivery is another matter.

“Do you think these people are biased in our favor because they know your work so well?”Andy Cook, recalled by Jay Acunzo, 2018

Less hunting, less blank-page work

On August 30, 2023, Cook introduced Kai, Tettra’s AI knowledge assistant. The announcement emphasized using information a company already had: answering questions, tagging pages, and making existing material easier to retrieve. When an answer was missing, the workflow could route a question to a human expert and preserve the response for later use.

That choice fits Tettra’s origins. More text alone would scarcely help the colleague already hunting through too much of it. The product’s job was to shorten the distance between a question and a dependable answer. In writing about the launch, Cook was explicit about the risk of adding low-quality material to an already difficult knowledge system.

In November 2024, he announced Tettra’s acquisition of Supportman, which brought Intercom customer-satisfaction alerts into Slack. He described plans to explore AI-assisted feedback and coaching for support teams. Those were ambitions in the announcement, with the emphasis again on bringing information into an existing working day.

Guidemaker addressed an earlier obstacle. In its 2025 Product Hunt launch, Cook explained that prospective Tettra users often got stuck creating documentation in the first place. The free tool records browser actions and turns them into guides. The question had moved upstream: before someone can find the instructions, someone has to make them.

Across these changes, Cook keeps returning to ordinary work: a question, a demonstration, a missing explanation. The problems are small enough to be dismissed individually and persistent enough to support a company. There is something fitting about a founder who sells organizational memory having learned so much by keeping track of the times his own first answer needed revising.