Alex Vratskides and the arithmetic of a better sentence

A text message is an unpromising place to look for a life's work. It arrives, interrupts, asks for something and hopes you will oblige. Alex Vratskides noticed that changing a few words could change what happened next. At Upstream, the mobile marketing business he co-founded, he and his future Persado co-founder Assaf Baciu became interested in those differences. The offer could remain much the same. The response would move. For a mathematician, this was an invitation to investigate.

The attraction lay partly in the restriction. An SMS had little room for literary ambition. Within that small space, however, there were many possible ways to make an offer. Vratskides began to see a problem that could be defined, explored and tested. A sentence might sound like a creative decision to its author; to him, it also presented a collection of alternatives. The customer, by responding or ignoring it, supplied the useful part of the argument.

Persado, founded in 2012, grew out of that observation. Its CEO had arrived at language through mathematics and mobile marketing, rather than through the usual doors of advertising. That route matters. It helps explain why his public remarks keep returning to evidence, why he talks about the business value of words, and why a pleasing piece of copy is only the beginning of his interest. Somewhere beyond the sentence, someone has to do something.

The words he had to learn twice

Long before he was selling software that chooses language, Vratskides had a language problem of his own. Moving from Anatolia College in Greece to Columbia University, he found that familiar mathematics came with unfamiliar English terminology. The underlying ideas travelled more easily than their names. He credited his school's mathematics teaching with preparing him well, while remembering the small inconvenience of learning what to call things again.

He graduated from Anatolia in 1989. His father had encouraged him to aim for a demanding university, and classmates also heading to America helped make the move feel manageable. Recalling school in 2016, he spoke warmly of lasting friendships, athletics and teachers whose demands he appreciated in retrospect. The recollection has the texture of an ordinary education: people and expectations, alongside the formal curriculum.

At Columbia he earned a BS in applied mathematics, followed by a Master of Engineering in operations research at Cornell. His early career included the United Airlines research and development group and management consulting at OC&C/McKenna Group. These were settings for working through practical decisions. Operations research itself is concerned with choices under constraints, which makes his later fascination with the cramped possibilities of a mobile message less surprising.

The path also gave him experience inside established organizations before he tried building one. His later advice to prospective enterprise founders emphasized understanding how large businesses work: their sales processes, projects and management habits. There is a decidedly practical note here. Being good at an equation and getting a company to buy the result are separate accomplishments. His career would require repeated practice at both.

Forty countries, one small screen

Vratskides co-founded Upstream in 2001 with Marco Veremis and served as CEO from 2004 to 2012. The business worked with mobile operators across more than 40 countries during his tenure. His responsibilities included international sales expansion, product innovation and strategy. Before Persado could become a business about language, Upstream supplied the experience of selling technology across borders and seeing how customers responded in different markets.

By March 2011, he was due to speak at CTIA Wireless in Orlando about mobile marketing. The concerns sound familiar now: how to measure results, how to make an offer relevant and how often a company should interrupt somebody. His warning was plain. Mobile devices were personal; marketers risked annoying their owners. A small screen and a short attention span imposed a discipline that even the most enthusiastic advertiser could not wish away.

“Marketers should pick their battles very carefully,” he said. It is a useful sentence to keep beside the later talk of optimization. Better language does not give a business an unlimited allowance of attention. The recipient still has a life to get on with. A company can measure a response and still need judgment about when to ask for one.

Inside Upstream, a team began working on automated message generation in 2010. After roughly two years of research, Persado continued independently in 2012 with a US base. The origin is less a flash of inspiration than a problem carried through an existing business until it could support another one. The first company provided the setting, the customers and the awkwardly short messages. The second would concentrate on choosing their words.

“Marketers should pick their battles very carefully.”

Alex Vratskides, on mobile marketing, 2011

A sentence gets a balance sheet

The early proposition was fairly audacious: use software to generate language for communications intended to prompt an action. In a 2015 interview, Vratskides described marketing departments already applying software and data to targeting and timing, while leaving the sales pitch largely to creative instinct. Persado was his attempt to apply a more systematic method to that remaining decision. It put the wording itself on the experimental agenda.

The company raised a $15 million Series A led by Bain Capital Ventures in 2013. A $21 million Series B followed in 2015, led by StarVest Partners with Citi Ventures, American Express Ventures and Bain participating. The financial institutions were an interesting presence around a company working on persuasion: institutions accustomed to examining the numbers were backing software intended to improve them. Those financing rounds gave the small-message experiment room to become an enterprise business.

American Express was Persado's first US customer, according to Baciu. The relationship offers a useful glimpse of the founders' chosen terrain. They were taking an unfamiliar approach into a large organization, where an intriguing demonstration had to become something people could use repeatedly. Vratskides' concern with enterprise experience makes sense against that background. A mathematical idea still needs a working commercial home.

As the technology developed, the company accumulated the results of language experiments. In 2024, Baciu described a dataset built from a million A/B tests and a system combining generation, prediction and feedback. The distinction is consequential for Vratskides' story. Producing an alternative sentence is one task. Learning whether that sentence changes behavior is another, and that second task had been present since the Upstream days.

“Performance will always trump utility,” Vratskides wrote on LinkedIn. The phrase is brief enough to fit comfortably in one of his old messages. It also explains his impatience with enterprise AI that cannot be connected to value. A business can admire a tool, use it enthusiastically and still ask what it contributes. His answer has consistently involved a measurable result beyond the act of generating text.

The people inside the experiment

For all the mathematics in the pitch, this has been a career conducted with other people. Veremis was his Upstream co-founder; Baciu became his Persado co-founder and is now its president. The company brings together language and technical work, an arrangement that suits a problem sitting between the two. Grammar alone cannot establish a commercial result, and a result alone does not tell a team how a sentence will sound.

Vratskides has also participated in businesses as an angel investor. His public biography describes a focus on enterprise software and business applications, and earlier biographies identify him as the first angel investor in Luluvise. His Founders Pledge membership adds another connection beyond Persado. These activities broaden the picture of someone whose working life has included backing and advising, as well as running companies.

His hiring principle, described in a Greek entrepreneurship profile, is to bring in people smarter than himself. It is an appealingly economical way to describe the limits of a founder's own expertise. A company devoted to testing assumptions ought to have room for colleagues who can challenge the person in charge. The same willingness to encounter unfamiliar thinking appears in the advice he has given younger Greeks about studying and working abroad.

That advice stresses exploration, skills and experience rather than following an expected course. It leaves room for a student to change direction. Read alongside his own movement from applied mathematics into marketing, it feels less like a ceremonial encouragement than a description of how careers can actually develop. The subject on the diploma may establish a starting point without deciding every subsequent destination.

“Performance will always trump utility.”

Alex Vratskides, LinkedIn

Now the sentence has to pass review

In May 2025, Persado introduced Marketing Compliance AI for financial services. The product combined regulatory analysis, performance scoring and brand considerations. Its stated purpose was to help marketing and legal teams identify and resolve problems in communications sooner. The company reported reductions in review time, but the more revealing development for this story is the expanded definition of a successful message.

A bank needs language that people respond to. It also needs language that can clear internal review and meet the rules applying to its business. Persado's newer work addresses that combination. Its account of the product describes different AI agents examining regulation, marketing performance and oversight. This is a further set of constraints around the sentence, arriving years after the original constraint of the SMS character limit.

The company's current leadership page still lists Vratskides as CEO and co-founder. It says eight of the ten largest US banks use Persado. Those are company-reported adoption figures, rather than a promise about what any particular campaign will achieve. They nevertheless show how far the setting has shifted from the original mobile-marketing question: the audience for his experiment now includes institutions with substantial demands on both language and process.

His stated ambition remains connected to decisions. In his Panathenea speaker biography, the principle is that data should guide choices rather than merely justify choices already made. Applied to words, that means keeping the alternative sentence in play long enough to learn something from it. Applied to a company, it is a fairly demanding instruction to its founder.

The appealing detail in this story is still the small one. Vratskides once had to learn English names for mathematical ideas he already understood. Years later, he helped build a company asking how different words could produce different outcomes. There is room in that journey for algorithms, investors and bank review teams. At its center sits a familiar human activity: trying another way of saying it, then paying attention to the reply.