Ask an AI system for last quarter's revenue and it may return a number with the serene confidence of a maître d' pointing toward the best table in the room. The trouble begins when Finance meant recognized revenue, Sales meant bookings, and the model meant whichever column looked persuasive. Mark Palmer has spent much of his career in that treacherous distance between data and meaning. His latest job gives the distance a name, a product and a deadline.
In April 2026, Palmer became Chief Marketing and Strategy Officer at AtScale, the Boston company that builds a universal semantic layer for analytics and AI. The title sounds like a pivot toward messaging. In Palmer's case, it is more like an engineer returning to an old bug with better tools. One of his first jobs after earning a computer science degree involved managing the syntax and semantics of data at Fidelity. Decades later, he is still asking what a customer code, product hierarchy or financial measure means after the raw characters have been stored.
The career between those two points is a tour through successive anxieties about data. First it was too hard to model. Then it moved too fast. Then there was too much of it. Now machines can inspect it, summarize it and answer questions about it, but they can also produce a tidy fiction at remarkable speed. Palmer's argument is admirably unfashionable for an AI executive: intelligence is not an excuse to skip the definitions.
Chapter one
Before the answer had meaning, it had to arrive in time
Palmer started as a developer. His early résumé includes Object Design, YouthStream Media Networks, Iona Technologies and Progress Software. He was a principal consultant, a co-founder, a CTO, a CIO, a product director and a general manager. The alphabet soup matters less than the progression. He learned the machine from inside the code, then learned how to assemble teams and markets around it.
The defining chapter came at StreamBase Systems, a company founded around research associated with database pioneer Michael Stonebraker. StreamBase worked on complex event processing: software that could watch torrents of live information, recognize a pattern and respond while the event was still useful. Financial markets offered a vivid proving ground. Prices jumped, trades accumulated and risk changed faster than a traditional database report could politely clear its throat.
Palmer had already seen the seduction of live data on a Bankers Trust trading floor in London. He recalled stock prices jumping and diving in milliseconds, with risk and reward made visible as they moved. At StreamBase, that velocity became a business. The company helped systems analyze streams rather than waiting to query a settled pile of records. Under his leadership it was selected as a World Economic Forum Technology Pioneer in 2010. Institutional Investor repeatedly included him among its financial technology executives and innovators.
TIBCO acquired StreamBase in 2013. Palmer stayed, first carrying the streaming business into its new home and later leading a much broader collection of analytics, data science and data-virtualization products. He ultimately became Chief Product Officer. The portfolio included Spotfire, whose visual analytics team in Gothenburg gave him one of his favorite product experiences. Their craft confirmed something he had begun to suspect: a serious software product is also an act of design.
“As a CEO, I view company building as a creative act.”Mark Palmer
The creative turn
A product manager goes on an anthropological expedition
Palmer's definition of creativity is generous enough to include pricing. It includes the inspiration required to rally a team, the persuasion required to activate a board and the sensitivity required to notice a customer need before it becomes a feature request. He has called product management an “anthropological expedition.” The phrase is better than most job descriptions. A product manager observes a tribe of users, notices their rituals and contradictions, and returns carrying a roadmap.
At Spotfire's headquarters in 2004, he picked up Edward Tufte's book on quantitative information and found a history of visual thinking. Palmer had worked on a program that visualized the silicon layout of a VAX 9000, which he likened to exploring the bones of a computer. Tufte supplied the intellectual frame: a clear visualization is evidence of clear thought. The idea migrated into Palmer's view of leadership and language.
His own transition from engineer to storyteller did not require abandoning technical detail. It required discovering that detail alone is a poor traveling companion. On his Techno Sapien site, Palmer puts the lesson plainly: the best products do not always win, while the best stories often do. His wife coined the site's name, a play on Homo sapiens and a reminder that technology ought to amplify humanity. It is branding with a built-in governor.
Palmer studied behavioral economics at the University of Chicago Booth School of Business and developed an advisory practice around AI, data infrastructure, product strategy and messaging. He has worked with Warburg Pincus' data-infrastructure investment team and served in board or advisory roles connected with Talkmap, Correlation One, the Data Visualization Society and other technology organizations. The engineer who once worried about latency began worrying just as much about whether a buyer understood the stakes.
- Read every day.
- Connect the ideas in a slip-box note system.
- Share one useful thought in public each week.
He calls that routine his “Learning OS.” Morning time before ten is reserved for reading, research and writing. He once published for 100 consecutive days, discovered that a streak can become its own tyrant, and settled into a weekly cadence. The test is conversational: explain one important, funny or surprising idea as though talking around the dinner table. If he can turn the idea into a post, he knows he understands it.
The practice also created opportunities. Palmer has said that public writing led to board invitations, new relationships and inbound work. His aphorism is characteristically tidy: “The quality of your writing equals the quality of your thinking and your leadership.” He has co-authored an O'Reilly report on ModelOps, hosted the Data Humanized podcast and interviewed leaders from companies including Best Buy, Bentley, Wayfair, Jaguar Land Rover and ADP. Writing became both a thinking tool and a way of finding the next conversation.
The new problem
The dictionary must be attached to the calculator
AtScale brought Palmer back into an operating role. He joined its board in 2024 and spent roughly two years watching the product, company and market. Then he moved from adviser to executive. His mandate covers global marketing, category strategy, analyst relations, ecosystem storytelling and strategic positioning. A board member ordinarily offers altitude. Palmer chose the engine room.
The product sits between an organization's data platforms and the tools people use to ask questions. A catalog can point toward a table. A language model can write a plausible query. The semantic layer encodes what the business has agreed a metric means, how tables relate and which rules apply. AtScale's claim is that the same governed logic can serve BI tools, AI agents and applications across warehouses rather than becoming captive to one vendor.
Palmer draws the distinction between discovery and computation. Knowing where the number lives is useful. Knowing how the organization calculates it is decisive. In his formulation, context must reach the point where the machine computes the answer. Otherwise the AI is invited to improvise business policy one prompt at a time, a management technique unlikely to survive its first audit.
“The rare opportunity here is simple: AtScale is closing the gap between semantic context and a trustworthy, computed answer.”Mark Palmer
This is category work, which suits him. Complex event processing once needed a vocabulary before customers could budget for it. ModelOps needed a crisp account of the last mile between a trained model and a decision. A universal semantic layer faces the same commercial puzzle: explain an architectural component by making the absence of it feel concrete. Palmer reaches for everyday nouns such as customer, product and revenue, then shows how quickly they fracture inside a large company.
His current writing can be combative, playful and dense with analogies. He compares AI without business understanding to Mr. Magoo walking into furniture. He borrows glasses as a metaphor for the definitions that bring a company into focus. Under the jokes sits an operator's suspicion of hand-waving. A generated answer is useful only when somebody is willing to put it in front of the chief executive and defend the arithmetic.
The human measure
Stamina, curiosity and the sentence after the demo
Palmer's career has its own repeated rhythm: learn a technical system, find the human decision inside it, and construct language that allows a larger group to act. It appears in his mentoring through Data Science for All, where he worked with people entering data careers. It appears in his fondness for humanists who ask what is missing rather than treating the available dataset as the whole world. It appears in a podcast deliberately called Data Humanized.
There is also the stamina of a runner. Palmer has completed ten marathons, with a reported best of 2:51, and participated in soccer, baseball, basketball, skiing and rowing. Running supplies an unfussy metaphor for his work, but the numbers are good enough without decoration. Categories are built through repetition. So are products, writing habits and fast marathons. The glamorous part is the finish line; most of the work is another morning on the schedule.
AtScale is now asking Palmer to make a technical layer visible at the moment the AI industry is enamored with what sits above it. Models command attention because they talk. Semantic architecture is quieter. It carries definitions, permissions and computational rules, the institutional memory nobody applauds during a demo. Palmer's assignment is to turn that quiet infrastructure into a story without pretending the story is the infrastructure.
The old StreamBase problem was unforgiving: an insight that arrived after the market moved was merely a historical curiosity. The new problem is equally stern. An answer that arrives instantly but means the wrong thing is a beautifully formatted liability. Between speed and meaning sits Mark Palmer's career, still translating, still writing in the morning, still asking the machine to show its work.