Startups operating in today’s fast-paced technology ecosystem have never faced more intense pressure to achieve rapid growth and scale their operations sustainably. To secure that trajectory, many early-stage companies find themselves compelled to land large enterprise clients—the deep pockets of the global IT world that can provide the recurring revenue and market validation necessary for long-term survival. However, as any seasoned founder or enterprise sales leader will readily attest, breaking into these lucrative corporate accounts and winning high-stakes contracts is an uphill battle fraught with hidden obstacles.
One of the most critical and common mistakes made by younger, less-experienced startups is operating under the comforting illusion that exceptional technology will practically sell itself, according to Arsalan Tavakoli-Shiraji, co-founder and senior vice president of field engineering at Databricks. Today, Databricks stands as an undisputed artificial intelligence and big data storage giant, boasting a staggering valuation of $190 billion and counting more than 70% of the Fortune 500 among its active customer roster. But the company’s current status obscures its humble origins as a tiny startup, a time when Tavakoli-Shiraji was personally responsible for helping secure the organization’s foundational enterprise deals.
Reflecting on those early days during a keynote session at TechCrunch Disrupt, Tavakoli-Shiraji offered a candid assessment of the realities of enterprise sales. When founders first venture into the market to sell to massive corporations, they almost universally labor under the assumption that success is purely a function of having the superior technical product. Yet, he ultimately learned the hard way that technology is only one piece of a complex puzzle, and that human relationships and rigorous internal corporate processes ultimately matter the most.
Recalling the friction of those early endeavors, Tavakoli-Shiraji shared a formative memory of securing an eagerly anticipated meeting with a high-profile corporate prospect, only to discover upon arrival that the decision-maker was the exact same individual who had recently signed a multi-million-dollar mega-deal with Databricks’ primary market competitor. The prospect had virtually no interest in listening to the pitch because he had essentially bet his professional career on that competing solution. Beyond entrenched competitor loyalty, Tavakoli-Shiraji also vividly remembers early deals fizzling out at the final hurdle because the startup simply lacked the maturity and preparation required to navigate complex legal, contractual, and cybersecurity security reviews.
Drawing from these hard-earned lessons, Tavakoli-Shiraji outlined several strategic approaches and crucial perspectives that early-stage startups must adopt if they hope to successfully break through the enterprise barrier and capture the attention of risk-averse corporate buyers.
Sell under a bigger tech company’s umbrella
In its earliest incarnations, Databricks experienced firsthand the frustration of being a small, unknown entity trying to capture the attention of massive financial institutions. The startup tried repeatedly, and failed consistently, to score a introductory meeting with UBS, the prominent Swiss multinational investment bank. The breakthrough finally arrived not through direct outreach, but when Databricks forged a strategic partnership with Microsoft, formally becoming an Azure cloud partner.
Aligning with Microsoft fundamentally transformed the startup’s market positioning, meaning that its sales agreements and technical deployments would be covered under Microsoft’s established enterprise terms, compliance frameworks, and trusted vendor status. Crucially, the partnership also meant that Microsoft’s extensive sales force actively booked meetings and advocated for the startup’s capabilities to their existing client base.
All of a sudden, that meeting with UBS and other legacy institutions wasn’t hard to get at all, the co-founder recalled. It proved to be a massive boon for the young company, granting immediate access to distribution channels and customer relationships that would have taken years to cultivate independently. With Microsoft’s sales machinery helping pave the way, Databricks suddenly found its solutions being pulled into enterprise accounts globally. While that rapid influx created new operational challenges—such as prospective clients demanding localized features for international markets where Databricks had not yet expanded—they were undeniably the kind of good problems that every scaling startup hopes to encounter.
Over the years, Databricks has evolved to occupy the exact opposite side of that equation, now functioning as the established enterprise giant that acts as an umbrella for promising smaller startups. The company frequently integrates and sells complementary emerging products under its own enterprise terms, billing infrastructure, and rigorous cybersecurity setup. Notable examples of this ecosystem strategy include Hightouch, a company that provides advanced AI-powered marketing and customer data tools and in which Databricks has made strategic investments; Addepar, a provider of specialized data and portfolio tools for investment and wealth management professionals; and Tecton, a machine learning feature store startup that Databricks ultimately acquired to strengthen its internal capabilities.
When selling AI, address the “four Cs”
As the technology landscape has shifted dramatically toward generative artificial intelligence and machine learning applications, enterprise buyers have grown increasingly cautious about how they vet and integrate third-party tools. To help startups navigate these sophisticated purchasing committees, Tavakoli-Shiraji groups the primary concerns and objections voiced by enterprise executives into four distinct categories, which he refers to as the "four Cs": context, choice, cost, and control.
Context refers to an enterprise’s absolute requirement for AI models that can be trained and customized dynamically, functioning much like onboarding a new human employee. Enterprises demand systems that can absorb company-specific terminology, internal corporate processes, operational priorities, and proprietary workflows without leaking sensitive intellectual property.
Choice represents the vital freedom of architectural flexibility, ensuring that corporations retain the ability to switch between different underlying AI models and foundation providers rather than being locked into a single proprietary vendor ecosystem. Cost addresses the pressing financial mandate of predicting and tightly controlling how much compute-heavy AI products will actually cost to run at scale over multi-year periods. Finally, control encompasses the robust security guardrails, compliance frameworks, and administrative safeguards required to govern AI behavior, prevent hallucinations, and protect sensitive corporate and customer data from exposure.
Attack the expensive legacy systems enterprises think they’re stuck with
In evaluating the current state of enterprise adoption, Tavakoli-Shiraji employs a playful mental model, asking chief information officers to rate their organization’s integration of artificial intelligence on a scale from one to five. On this imaginary spectrum, a rating of one signifies an organization where employees merely have casual access to consumer-grade tools like ChatGPT, while a rating of five represents a company that has successfully adopted deeply embedded AI capabilities that have fundamentally transformed its core business processes and operating models.
While corporate executives across industries love to self-identify as being a three leaning toward a four, Tavakoli-Shiraji estimates that roughly 99% of the enterprise leaders he talks to actually reside firmly in buckets one and two. Most large corporations still have not figured out how to deploy autonomous AI agents in ways that create substantive business value beyond allowing internal software engineering teams to write code slightly faster. He cautions that it is going to take many, many years before the broader market can properly and securely deploy artificial intelligence in a deeply meaningful way across traditional enterprise environments.
This reality presents a massive, untapped opportunity for agile startups capable of delivering tangible value by modernizing the deeply entrenched, expensive legacy systems that enterprises often feel hopelessly stuck with. If you walk into any major commercial bank today, you will find that they are still fundamentally running core operations on legacy hardware like IBM AS/400 systems and aging mainframes. For decades, institutional leadership has been terrified to touch or modify that underlying infrastructure because the original development teams are long gone, and virtually no one left in the organization truly understands what the legacy code actually does.
However, the rapid maturation of modern AI agents has fundamentally altered that equation. These sophisticated systems can now autonomously dive into legacy codebases, analyze complex instructions, and actually decompose what the legacy software does line by line. Tavakoli-Shiraji notes that this capability finally opens the door for innovative startups to build products that achieve everything from wrapping ancient systems in modern, user-friendly interfaces to replacing the aging infrastructure altogether.
