The product that would become Reflect began with a small collision. Alex Bilmes was working on a front-end component and called Brad Heller for feedback. Heller, independently, was building a backend for essentially the same thing. Two former Cloudability colleagues had approached one problem from opposite directions and met in the middle. "We were trying to solve the same problem in different ways," Bilmes recalled. "We knew we had something."
That origin story contains the pattern that runs through Bilmes's career. Find information trapped in a system. Make it legible. Put it where someone can act on it. His products have moved from cloud-cost dashboards to embedded charts to enterprise sales intelligence, but the design problem underneath has remained remarkably stable: people have plenty of data and not enough useful context.
Bilmes is now the co-founder and CEO of Endgame, a San Francisco software company that describes itself as a knowledge system for enterprise sales. Its current product connects records from CRM software, emails, calls and company materials with outside information, then gives revenue teams a shared place to research accounts and prepare for conversations. The pitch is deliberately human. Endgame says it does not want to create more automated prospecting or add to the flood of unwanted sales messages. It wants a seller to walk into a meeting understanding the buyer.
A designer in the data room
Before Bilmes became a repeat CEO, he worked through the visual side of software. He owned Bilmes Media, joined Cloudability as its first employee and ran design there, then became a design partner at Lucid Labs. Cloudability helped companies understand what they were spending across cloud infrastructure. That put Bilmes close to a recurring enterprise problem: valuable information existed, but turning it into an interface people could understand required specialized work.
Other companies began asking how Cloudability had built its visualizations. The repetition looked like a market. With Heller, Bilmes created Reflect as "visualization as a service," a tool meant to let developers add charts and dashboards to web or mobile products without rebuilding the plumbing every time. The company entered Techstars Seattle in 2016 and raised a $2.5 million seed round. Its ambition sat neatly between infrastructure and interface: connect a database, organize its fields, and produce something usable without forcing every customer to assemble a visualization team.
“Getting closer to our customers. We spend just as much time meeting with clients as we do work on the product. It's paying off.”Alex Bilmes on a formative Reflect decision, 2016
A revealing comment from that period concerned failure. Bilmes called an expensive technical pivot a mistake that could have been avoided by talking to people earlier. His advice to new entrepreneurs was compact: "Listen! Read between the lines." The phrasing matters. Customers rarely arrive with a clean product specification. They bring complaints, habits and awkward workarounds. Listening is only the first operation. The founder still has to find the pattern.
Reflect was acquired by Puppet in 2018. Bilmes went on to serve as Puppet's vice president of growth, adding distribution to a resume already shaped by design and product. By the time he started Endgame with Graham Murphy, he had worked on the same boundary from several seats: how a company collects information, how a product presents it and how a business turns it into action.
The good company inside the wrong question
Endgame's first version was built for the product-led software boom. Free trials and self-service products created a trail of behavioral data: who signed up, which features they used, whether activity was spreading through a company. Endgame helped sales teams spot the accounts worth contacting. Bilmes defined product-led sales with characteristic economy: "It's just selling to people that are already using your product." Internally, the team joked that anyone who signed up was already a customer who had not paid yet.
The idea attracted customers including Figma, Retool, LaunchDarkly, Loom and Airbyte. Endgame announced more than $17 million in funding in 2021, followed by a $30 million Series B in February 2022. The team grew from two people to 15 during its first year. On paper, the category, capital and customer list all pointed in the same direction.
Then the direction changed. As software companies pushed upmarket, Endgame's customers cared less about converting a high-volume self-service funnel and more about navigating complicated enterprise accounts. At the same time, generative AI filled sales software with tools for writing emails and automating activity. Bilmes and his team concluded that their existing product was not aimed at the central problem their customers faced. They spent roughly two years building a different one.
Traction can prove that a product works without proving that it addresses the right problem. Bilmes kept the customer knowledge and changed the product built around it.
Endgame 2.0 launched in November 2024 as an AI-native product for enterprise sellers. Instead of deciding whom to contact from product signals, it assembled the information needed to understand a buyer: CRM history, email, call recordings, company strategy, sales methodology and public research. The shift was substantial, but it did not discard the company's original insight. In both versions, the valuable work happened between a pile of signals and a human decision.
AI that helps people think
Endgame's 2026 behavioral report makes the new thesis measurable. The company examined more than 31,000 AI interactions from hundreds of go-to-market professionals at enterprise organizations. Among the interactions it classified, teams spent more than twice as much AI time on contextual understanding, including account intelligence and conversation readiness, as they did on producing deliverables. The demand was not simply "write this for me." It was "help me understand what is happening."
Based on roughly 25,000 classified interactions in Endgame's 2026 report. Bars share one scale; percentages are labeled directly.
The finding gives Bilmes a case against the dominant sales-AI reflex. Generating a deck is relatively easy. Knowing what belongs in the deck, whether it reflects the customer's history and whether it follows the company's actual method is harder. A fast answer built on incomplete context can move work in the wrong direction. In a podcast conversation, he described the emerging problem as agent sprawl: teams deploy many specialized agents, each operating from different information and definitions. The same question can return incompatible answers.
His proposed fix is architectural and organizational. Create a central context layer so people and agents work from the same definitions, account history and operating method. Endgame calls the broader result revenue superintelligence. The grand name matters less than the practical picture: a chief executive, sales engineer or account manager should be able to examine the same customer without first reconstructing the company memory from scattered tabs and private conversations.
“AI is in the room where decisions get made.”Alex Bilmes, writing about Endgame's 2026 usage research
The continuity underneath the pivot
Bilmes once called himself contrarian by nature, with a wink. His present contrarian position is more grounded: use AI to make a seller more thoughtful rather than merely more prolific. It is also consistent with his design background. Good design is not decoration applied after the information is assembled. It is the discipline of deciding what belongs together, what deserves attention and what someone should understand next.
That frame makes his career look less like a series of jumps. Cloudability organized cloud costs. Reflect organized raw data into visual interfaces. The first Endgame organized product behavior into sales signals. The current Endgame organizes institutional knowledge into a working view of the buyer. Each company sits one step closer to the decision.
The ambition has grown with the systems. Endgame's public mission is to help humans think and work better. Its more specific aspiration is for sellers to become trusted advisors: informed enough to understand the buyer's priorities, prepared enough to ask useful questions and restrained enough not to treat automation as a substitute for judgment. Bilmes is not promising to remove the person from the loop. He is trying to improve what the person sees before acting.
Years ago, two builders realized they were making opposite halves of the same product. Today Bilmes is still joining halves: internal and external information, software and judgment, what a company knows and what its people can actually use. The interface is now conversational instead of graphical. The underlying work is familiar. Make the hidden structure visible. Then help someone decide.