Jonathan Anderson and the Trouble with a Simple Button

Jonathan Anderson has a fairly inconvenient qualification for someone running a software company: he says he cannot code. He has made the admission often enough that it reads like part of the job description. In an early funding announcement, he described being close enough to customers to know what they needed, yet unable to deliver a product change without involving engineering. A useful suggestion had to acquire a ticket, a place in somebody else’s schedule, and the patience of everyone waiting for it.

There is a whole office comedy inside that arrangement. A person can be trusted to understand a customer’s business, persuade them to buy software, and help them use it. Move a button on the screen, however, and the conversation becomes rather more formal. Anderson built Candu around that small, familiar indignity. Its premise gives people who understand users a way to change the experience those users encounter.

The button is modest. The question underneath it is considerably bigger: who gets to make the thing better?

Seven people, then one

Before Candu, Anderson ran professional services at InsightSquared in Boston. His team built custom workflows and front-end experiences for the analytics company. When a funding round fell through, he had to let go of six of its seven members. It was his first experience as a manager. The problem of serving customers with fewer people became very concrete.

That episode gave the later company its practical starting point. Could the customization happen through the product itself? Could a customer-facing team make the changes that previously required bespoke development? The appeal was easy to understand. Finding a workable product would take longer.

Anderson moved to London, where Entrepreneur First brought him together with Michele Riccardo Esposito, an engineer and product manager with experience at Coursera. One founder brought the customer problem; the other brought experience building software and learning products. There is something fitting about a company concerned with teaching people to use software beginning with that combination.

An education before the editor

Anderson’s career spans consulting and software operations. He studied Earth Systems at Stanford, followed by Management Science and Engineering, and later earned an MBA at London Business School. His working history includes Bain & Company, InsightSquared and LaunchDarkly. This is a route through institutions with rather different ideas about what counts as a useful day’s work.

The thread is the movement between understanding a problem and getting an organization to act on it. Consulting can give a problem an elegant diagram. Customer operations has to deal with the person still waiting for an answer after the diagram has been admired. Candu sits close to that latter experience: the change needs to reach a real screen.

His interests have also extended beyond commercial software. Anderson co-founded wayOUT, the nonprofit supporting LGBTQ+ youth. His 2023 VentureFizz conversation included that work alongside his startup career. A founder biography usually makes room for a funding round; this one also has room for building an organization with a different purpose.

Four or five conversations a week

Candu’s early product research acquired an attractive shorthand: an interview every day for a year. Anderson supplied the less theatrical arithmetic: about four or five interviews a week. Sessions initially lasted an hour, then shortened to roughly 30 to 45 minutes. The routine emerged because early versions built on the team’s intuitions did not find the audience they expected.

He spent much of his time finding interviewees, talking to them and working with the designer and engineers on what came next. Designs went through several conversations before the next iteration. The repetition mattered because a single agreeable conversation is a flimsy foundation for a product. People can be generous to a mockup they would never use.

The story’s useful detail is its ordinariness. There was no grand original plan to become a daily interviewer. A habit formed around the work that was helping. Anderson’s contribution was to keep doing it long enough for the customer’s version of the problem to become harder to ignore.

Candu’s early research rhythm, described by Anderson in April 2020.

Mother meets the components

In February 2022, Anderson tried a different sort of user test. Candu’s Product Hunt launch video featured him explaining UI components to his mother. The company landed among the day’s top five products. The joke had a sensible target: the vocabulary of a software company can make a perfectly practical tool sound like an unusually complicated piece of plumbing.

Inviting a parent into the explanation gave the jargon somewhere to go and die politely. A component is a piece of an interface. An experience is what somebody does with it. People outside the industry have every right to ask why the first noun deserves quite so much attention.

That distinction became part of Candu’s positioning, as the company emphasized product experiences rather than isolated components. The launch had a family cameo, but its underlying problem was commercial: if people cannot understand what they can accomplish with your product, they have little reason to try it.

The pop-up gets a reprieve

Anderson has also had to revise his own convictions. In Candu’s digital adoption manifesto, he described his annoyance with pop-ups and the early ambition to let teams edit product pages directly. Customers nevertheless used Candu alongside other tools that offered product tours. They wanted guidance as well as better pages.

Candu added tours and hotspots. The concession is revealing. A preference for cleaner software has to accommodate the person encountering that software for the first time. Sometimes a short explanation in the right place is helpful. The question becomes when to intervene, and how much explanation the moment requires.

Anderson’s work allows for both: embedded content that belongs on the page, and guidance that appears when someone needs it. A founder can make a tidy argument against an entire category of interface. Customers tend to arrive with untidy requirements. In this case, the requirements won.

“I can’t code. And I find pop-ups very annoying.”Jonathan Anderson, Candu manifesto

A trial should have a destination

By May 2023, Anderson was working with Kyle Poyar on a nine-step audit of product-led growth. The exercise examined a free trial as a whole journey rather than treating sign-up as the finish line. It connected the practical questions a team can investigate with changes it can make.

A trial creates an odd bargain. The company lends someone access to its product, while the potential customer lends it attention. Both can be wasted. Every extra field or unexplained screen consumes some of that attention before the person reaches the reason they signed up.

This is where Anderson’s public writing repeatedly returns: the work between access and usefulness. “Activation” sounds satisfyingly technical. In ordinary language, it means somebody has managed to do something worth coming back for. Thinking about that destination makes the onboarding conversation more concrete. A tour that has been completed and a job that has been accomplished are different measures.

The onboarding company gets stuck

Then came an admission with rather good comic timing. After helping more than 100 SaaS companies with onboarding, Candu still had to improve its own. Anderson’s account of the rebuild credits customer success manager Deborah Ramirez with spotting a pattern: users were confused about the workflow. They could operate individual features and still publish content to the wrong audience.

The team clarified four stages: design, place, target, publish. It built eight supporting touchpoints, including a publishing checklist and short contextual tours. The aim was to help someone publish their first content, with more advanced subjects introduced later.

This is the most human episode in the company’s account of itself. Expertise does not exempt its owner from the problem. The experienced team knows what a placement means; the new user sees another mysterious label. Using their own product forced Anderson and his colleagues to confront that gap. The test was whether customers began asking better questions, beyond the immediate mechanics of clicking.

Let the prompt start; let the person finish

Candu’s shift into AI introduced another version of the same challenge. Its AI editor project ran across three months and eight sprints, including a hackathon and design sprint. The team kept the visual editor and taught AI to work within it. A user could request a layout in words, then refine it by dragging and editing.

Anderson’s account pays attention to small frictions. Pilot customers valued undo. The paperclip icon did not communicate uploading as reliably as the team expected. Translation was useful for occasional users. These are the details that survive after an impressive demonstration meets an actual working day.

The approach suits a founder whose limitation helped define the business. A generated result still has to be something a person can inspect and change. For someone who cannot read code, an editable visual result provides a different kind of control. The machine can do the first pass; the person still needs a convenient way to disagree.

A profitable year, a shorter queue

Looking back on 2025, Anderson reported that Candu had reached profitability. He also credited the people behind the product’s AI work: CTO Stuart Coope, designer Rocío de Torres and engineer Alexandra Hadjidaki, alongside product adviser Kelly O’Connell. The account describes prompting and manual editing working together, with product snapshots providing context.

Profitability gives this chapter a useful constraint. New technology has to serve a customer willing to pay for what it makes possible. A banner built quickly has little value if it is irrelevant, confusing or impossible to maintain. Anderson’s stated ambition extends toward software that understands the product well enough to build its educational experiences.

That ambition carries the original complaint into a different technical era. The early question concerned engineers coding buttons. The newer question concerns how much of the preparatory work a machine can take on before a human steps in. The distance he is trying to shorten remains recognizable.

The next version has to keep up

In his newer writing about Capture Flow, Anderson argues that tours have become another maintenance burden. Candu’s approach lets someone perform a task in a browser extension while AI drafts the corresponding tour. The company also describes using repository context to detect changes and propose updated selectors. These are Candu’s product claims, with human review still part of the process.

The aspiration is understandable: guidance should be able to keep pace with the software it explains. A helpful instruction pointing at yesterday’s interface is an excellent way to make today’s user feel foolish. The user may be entirely blameless.

Across these chapters, Anderson keeps returning to people whose knowledge exceeds their access to the tools of production. The customer specialist. The occasional editor. The new user with a perfectly sensible question. His own inability to code gave him a place among them. The company he runs keeps asking how much easier their work could become if a good idea did not have to wait in so many queues.