Signal
David Boskovic is building Obvious around outcomes, not chat From CSV importer to agent-directed software work Frontier asks teams to bring 12 weeks of roadmap into a two-day room David Boskovic is building Obvious around outcomes, not chat From CSV importer to agent-directed software work Frontier asks teams to bring 12 weeks of roadmap into a two-day room

Founder profile · AI & enterprise software

David Boskovic Keeps Turning Friction Into Software

He taught himself to code from a 600-page manual, built Flatfile out of frustration, and now asks what engineers should do when agents can carry the work from specification to shipped code.

The first useful clue in David Boskovic's career is not a funding round or a product launch. It is a thick manual. As a teenager, he wanted to build a website for his father's small business, so he taught himself web development by reading roughly 600 pages on PHP and SQL. The project grew into an e-commerce system that took two years to make. By 15, he later recalled, he could code almost anything he could imagine. More important, he had learned what happens when code meets an actual business: inventory, customers, databases, and all the awkward movement between them.

That movement became his subject. Across Rainmaker, where he was a co-founder and CTO, and Envoy, where he became platform architect, Boskovic kept running into the same unglamorous assignment. A customer had information in one system. A software product needed it in another. Between them sat spreadsheets, mismatched fields, legacy exports, patient support staff, and an engineer asked to build yet another importer.

At Envoy, the cost of that gap became unusually visible. One large customer had hundreds of offices. Front-desk workers were arriving early to type expected visitors into the system because there was no good bulk-import path. Boskovic and product leader Eric Crane scoped the fix. It looked like three months of work. For two people who had seen this movie before, the estimate felt less like a project plan than a dare.

“This needs to exist! This is the last time I'm going to build it!”David Boskovic on the first version of Flatfile

A small fury with a price tag

Boskovic called what followed “rage designing.” Over two weekends in 2018, he produced the first version of the tool that became Flatfile. The phrase is funny because the output was disciplined. Instead of merely complaining about repetitive infrastructure, he isolated the repeated part, made it reusable, and put it on the internet at $1,000 a month. He and Crane had been tinkering nights and weekends for about six months when the market answered.

The early signal came from Blackbaud, a large software company serving nonprofits. It found the React component and began installing it before Flatfile had raised money or even incorporated. This was unusually clean evidence. The customer did not need to be educated into feeling the pain. Its users already had donor records to move, and the rough product solved enough of that problem to enter a real enterprise workflow.

01 · EXPORTA customer arrives with years of records.
02 · FRICTIONFormats, fields, and meanings refuse to line up.
03 · HUMAN LOOPTeams clean, explain, retry, and reconcile.
04 · PRODUCTFlatfile turns the repeated handoff into software.
The overlooked enterprise loop: the software cannot begin until the customer's history gets through the door.

Flatfile was founded in 2018 and raised a $2 million seed round the following year. By 2021 it had more than 300 customers and announced a $35 million Series A. In September 2022, Boskovic announced a $50 million Series B led by Tiger Global Management. He wrote that the company was powering data-import experiences for 500 customers. The funding mattered, but the more durable achievement was conceptual: a chore that had lived in the margins of product teams now had a name, a buyer, and a dedicated layer of software.

2Weekends behind the first “rage designed” version
500Customers reported when the Series B was announced
$50MSeries B announced in September 2022

The product behind the product

Boskovic's language around Flatfile was always larger than the upload button. “Hardware is built on silicon. Software is built on data,” he wrote when announcing the Series B. His argument was that faster, safer data exchange could unlock the work that came afterward. A CRM without customer history is an empty box. An operational system without the old records cannot produce much magic. Data onboarding was the threshold customers had to cross before the product they bought could become useful.

That framing also shaped how he sold. Boskovic described developers as advocates rather than the people Flatfile wanted to monetize at small scale. An engineer could try the product, prove it inside a narrow use case, and carry it toward an enterprise purchase. The decisive moment was often one “hero feature”: the ability to accept a 20-million-record file, handle a strange fixed-width format, or accommodate the one workflow that had blocked a buyer for years. Solve the difficult specific thing, he argued, and the internal champion could carry the rest.

His management ideas followed the same systems logic. Boskovic has said that first-principles leadership means understanding the pieces of a problem instead of imitating another company's visible success. He used the image of being stranded on an island and needing to build an engine: remembering the outline of an engine helps less than understanding heat, gas, and energy. The conclusion was concise. “You can't just copy success. You need to understand its building blocks.”

Flatfile's remote-work policies offered one practical expression. In 2021, the company paid for ergonomic setups and offered homeowners money toward purpose-built workspaces. Boskovic's point was that distributed work should have an environment designed for it, not merely an office culture translated onto video calls. Whether designing an importer or a workplace, his instinct was to examine the whole system around the outcome.

Flatfile becomes Obvious

By late 2025, the company had widened its frame. Flatfile, the company, became Obvious. Flatfile, the data-migration product, stayed in active support. The distinction was deliberate. Seven years of working on migrations had shown the team that clean data was often only the opening move. Someone still had to analyze it, build the report, prepare the presentation, connect another tool, and decide what happened next.

Obvious was built as an AI workspace for that longer chain. Its agent can make documents, sheets, slides, applications, dashboards, and code, with projects retaining context and teams collaborating in the same environment. The old company removed friction between a customer's data and a product. The new ambition is to remove friction between a person's intent and a finished piece of work.

TEENSLearns PHP and SQL while building an e-commerce system for his father's business.
2010Begins building Rainmaker as co-founder and CTO.
2016Joins Envoy as platform architect and encounters the data-import problem again.
2018Builds the first Flatfile version with Eric Crane.
2022Flatfile announces a $50 million Series B.
2025The company becomes Obvious while Flatfile remains a supported product.
2026Obvious takes its agent-directed engineering method on the road through Frontier.
Sixteen years of the same instinct: find where intent gets stuck, then redesign the handoff.

When the engineer becomes the underwriter

In 2026, Boskovic began speaking more directly about software teams that no longer write every line themselves. The provocative surface is speed. He has described Obvious moving from roughly 400 pull requests a month to 7,000 with the same team. Frontier, the company's in-person build sprint, asks participants to bring a codebase and attempt 12 weeks of roadmap in 48 hours. The workshops teach specifications, system mapping, parallel agent work, and verification.

The deeper claim is about responsibility. If an agent can plan a feature across many pull requests, respond to review, run tests, inspect security and performance, record the result, and watch a staging release, the human job changes. Boskovic uses the language of underwriting: the engineer defines the intent, steers the process, evaluates evidence, and accepts responsibility for what ships. Production becomes less sequential, but judgment does not become optional.

“You can't just copy success. You need to understand its building blocks.”David Boskovic on first-principles leadership

That position is easier to understand beside Flatfile. Data onboarding also moved work from people into software, but it could not pretend ambiguity had vanished. A file might contain several addresses in 85 columns. A “name” field could mean a person in one system and an account in another. Flatfile learned from behavior, proposed mappings, and kept humans available for uncertain edges. Obvious brings the same tension to code: make the routine autonomous, make uncertainty visible, and give the person enough evidence to decide.

There is a useful humility in choosing CSVs as the first mountain. They are old, plain, and stubbornly common. They also contain the accumulated reality of organizations: conventions nobody documented, exceptions somebody once needed, and years of decisions flattened into rows. Boskovic did not escape that mess by declaring it obsolete. He built a company around helping it move.

Now he is applying that patience to a louder technology. AI coding demos can make creation look instantaneous. Real organizations still have architectures, tests, permissions, review queues, rollout plans, and customers who notice mistakes. The unglamorous work has simply moved. Boskovic's wager is that the connective tissue can become a product again.

A career built around the missing middle

Boskovic describes himself publicly as an engineer, designer, experimenter, and humanist. The combination fits. Engineering supplies the mechanism. Design notices the painful interface. Experiment puts a rough version in front of a buyer. Humanism keeps attention on the person carrying water between systems.

His story does not require a myth about predicting the future. The pattern is more practical. As a teenager, he wanted a business to work on the web, so he learned the stack. At Envoy, he wanted customer data to enter the product without morning shifts of manual typing, so he built an importer. At Flatfile, he wanted data migration to stop consuming entire teams. At Obvious, he wants intent to survive a much longer journey through agents and tools until something finished emerges.

The aspiration is large, but the method remains specific: inspect the handoff, understand its building blocks, and replace repetition with a system people can trust. Friction, in this telling, is not merely a nuisance. It is the place where a product is waiting to be noticed.