Sophie Cheng’s Next Campaign Is Trust

Sophie Cheng has a way of turning the grandest promise in technology into a plain operating question. If an AI agent failed today, she has asked, would the company know before its customers did? The question has no futuristic shimmer. It sounds more like something taped beside a support desk, which is precisely why it lands. The age of customer-facing AI will be decided in the unglamorous minutes after a system meets an impatient person, loses context, chooses the wrong tone, or simply fails to deliver a message.

Cheng became Chief Marketing Officer of Sinch in June 2026, joining the Swedish cloud communications company’s Global Leadership Team. The promotion came exactly two years after she arrived to lead global product go-to-market work. During that stretch, her remit covered product marketing, revenue enablement, market positioning, analyst relations, and execution. The title widened, but the underlying problem remained familiar: how do you translate a dense technical system into a promise a customer can understand, then make the system keep that promise?

For Sinch, the question comes with industrial-scale consequences. The company carries messages, email, and voice interactions for businesses around the world. Its machinery largely sits beneath the surface. A delivery notification, a verification code, a service conversation, or a branded message can feel like a tiny event to the recipient. Multiplied across companies and channels, those tiny events become infrastructure. Marketing this kind of company is partly an exercise in making the invisible legible.

The translation business

Cheng’s route to the C-suite runs through product marketing, a discipline built around translation. Product teams speak in capabilities. Sales teams speak in objections and deadlines. Customers speak in the language of jobs they need to finish. The product marketer has to listen to all three and find a sentence that stays true when it crosses the room.

Her education gave her an early version of that crossing. She studied business administration and marketing, services, and communication management at the University of St. Gallen in Switzerland, with an exchange program at Singapore Management University. Her career then moved through brand and product roles, including Audi, DHI Group, and Stride, before she entered the fast-moving sales technology market.

• 2019-2021Customer and product marketing leadership at Chorus.ai, the conversation-intelligence company later acquired by ZoomInfo.
• 2021-2024Successively broader roles at ZoomInfo, culminating in Vice President of Global Product and Customer Marketing.
• 2024-2026Senior Vice President of Product Go-to-Market at Sinch, connecting positioning with revenue execution.
• June 2026Chief Marketing Officer at Sinch, with a seat on the company’s Global Leadership Team.

Chorus.ai and ZoomInfo were useful places to learn how quickly software changes the behavior it is designed to observe. Conversation intelligence turns sales calls into data. Customer communications platforms turn the exchange itself into programmable infrastructure. In both cases, the marketer cannot remain far from the product. The language used to sell the system has to survive contact with how people actually use it.

“If customers can’t trust it’s you, they simply won’t engage.”Sophie Cheng

That line compresses a sprawling technical problem into a human reflex. People ignore messages they do not recognize. They abandon conversations that feel unsafe or strangely impersonal. They notice when a company asks the same question twice. Identity, context, delivery, and tone may belong to different teams and systems inside an enterprise; to a customer, they arrive as one experience.

Start before the chatbot

Cheng’s sharpest marketing lesson is about sequence. Brands often begin by deciding they need a chatbot, then search for a place to put it. Her alternative begins with the journey. What is the customer trying to accomplish? Where is the friction? Which interaction benefits from speed, and which one requires more context or a person? Only after those questions does the technology earn a role.

This is also why “omnichannel” can be a misleading trophy. Owning more channels does not guarantee a better conversation. A useful system chooses an appropriate channel for the task and carries context between touchpoints. The customer should not have to understand the company’s org chart to get a coherent answer.

The distinction becomes more important as large language models make automated conversations feel smoother. Fluency can disguise a broken journey. A bot may produce elegant sentences while routing a person in circles. Cheng’s focus on the journey keeps the scorecard close to the customer: Was the task completed? Was the channel welcome? Did the conversation preserve trust?

After pilot purgatory

By 2026, the enterprise AI story had moved beyond pilots. Sinch surveyed more than 2,500 enterprise executives and found that 62 percent had AI programs in production. Yet 74 percent said their organization had rolled back a live customer-communications agent. The reasons ranged from exposed data and hallucinations to tone problems, gaps in auditability, and other reliability failures.

A rollback is an ugly metric, but Cheng offers a useful way to read it. “More monitoring means more failures are detected,” she has said. “It does not mean more failures are created.” In other words, a team that can see its system fail has crossed one important threshold. It can diagnose, contain, and improve. The more dangerous system may be the one generating no alerts because nobody built the instrumentation.

That observation reveals an operator’s temperament. It avoids both easy panic and easy celebration. Production AI needs governance, guardrails, escalation paths, connected context, and communications infrastructure that reliably moves information across channels. These are not glamorous launch assets. They are the conditions under which a customer promise becomes repeatable.

Cheng has also described work on what Sinch calls a “Truth Layer,” a library of trusted, AI-readable marketing knowledge that can supply tools with reliable context. The name is memorable because it makes knowledge architecture concrete. Before a system can answer well, a company has to decide which information is authoritative, keep it current, and make it usable. Good judgment has to be encoded upstream if the automation is expected to reflect it downstream.

For a CMO, this changes the border of the job. Brand once seemed to live mostly in stories, design, and media. When software speaks with a customer, the system’s behavior becomes brand behavior. A broken handoff is a message. An unrecognized sender is a message. So is a graceful transfer to a person when automation reaches its limit. Marketing cannot own every component, but it has a stake in the whole chain.

A global seat with a local test

Cheng’s career has crossed Europe, Asia, and North America, and Sinch operates across more than 60 countries. The scale creates a temptation to standardize every interaction. Her public work points toward a more careful balance: build a connected global system, then respect regional habits, rules, and channel preferences. A useful communication strategy travels, but it does not pretend every market behaves the same way.

This is where her product-marketing background may matter most. Good product marketing is allergic to a promise that the product cannot keep. It asks for evidence, listens for friction, and rewrites the story when the market teaches something new. It also turns complexity into choices. At Sinch, those choices might include when to use SMS, RCS, WhatsApp, voice, or email, and how to preserve identity and context between them.

“We’re only scratching the surface of what’s possible.”Sophie Cheng

Cheng used that phrase when accepting the expanded role, alongside a goal of turning customer and market intelligence into growth, innovation, and more meaningful customer experiences. The interesting word is intelligence. It suggests a feedback loop rather than a broadcast plan: hear what customers are doing, sharpen the product story, equip revenue teams, watch the real experience, and feed what happens back into the company.

Her appointment also says something about the modern CMO path. The route can run through the product, not around it. Market positioning, analyst relations, and revenue enablement are not side assignments; together they create a detailed map of what a company believes, what outsiders hear, and what sales teams can prove. Cheng spent her first two years at Sinch inside that map.

Now comes the broader assignment. AI is making customer communication more capable and less predictable at the same time. The winners will not be the companies with the most human-sounding demo. They will be the ones that know what the customer came to do, choose the right moment and channel, keep the context intact, and notice trouble early. Trust will be earned in those small operational decisions.

That brings the story back to Cheng’s plain question. Would you know before your customers do? It is a good test for an AI agent, a communications platform, and a marketing organization. It asks whether the promise has instrumentation. It asks whether listening continues after launch. Most of all, it treats the customer’s experience as reality and the company’s presentation as a hypothesis that still has to pass.