The Enterprise Procurement Trap: Why Strategic Partnerships Trump Feature Sets

In the race to scale, early-stage startups often fall into the trap of believing superior technology is the primary driver of enterprise adoption. However, as Arsalan Tavakoli-Shiraji, co-founder and senior vice president of field engineering at Databricks, told TechCrunch during a session at TechCrunch Disrupt, the reality of winning large contracts is that "people and processes matter the most."

From an operational lens, the biggest hurdle isn't a lack of features, but the friction of enterprise procurement. Tavakoli-Shiraji recalled early deals that failed not because of the product, but because the startup was unprepared to handle security, contractual, or legal inquiries. Furthermore, enterprise buyers are often locked into competitors not by technical superiority, but by professional risk; Tavakoli-Shiraji shared an instance where a prospect refused to meet because they had already "bet his career" on a competing product.

**The Partnership Moat**

To bypass these roadblocks, Tavakoli-Shiraji suggests that startups should sell under the umbrella of a larger tech company. Databricks utilized this strategy when it struggled to secure a meeting with the Swiss bank UBS. By becoming an Azure cloud partner with Microsoft, Databricks shifted the procurement burden: sales agreements were covered under Microsoft's existing terms, and Microsoft's own sales team began booking meetings and advocating for the startup.

This distribution model allows a startup to scale rapidly by leveraging the trust and infrastructure of an incumbent. Databricks now employs this same strategy for other companies, allowing startups like Tecton (which Databricks later acquired), Addepar, and Hightouch to sell their products using Databricks' own cybersecurity setup and terms.

**Solving the 'Four Cs' and Legacy Debt**

Beyond distribution, Tavakoli-Shiraji argues that winning the enterprise requires addressing four specific categories of concern, which he calls the "four Cs":

* **Context:** AI must be trained to understand a company's specific terminology and priorities. * **Choice:** Customers want the freedom to switch AI models rather than being locked into one provider. * **Cost:** Enterprises require strict control over the operational costs of the AI product. * **Control:** Robust safeguards are necessary to protect sensitive data and govern AI behavior.

Finally, there is a significant ROI opportunity in attacking legacy systems. Tavakoli-Shiraji points out that many banks still rely on mainframes and IBM AS/400 systems that are too complex for human teams to easily modify. He suggests that AI agents can now decompose this ancient code, allowing startups to build products that either modernize the interfaces of these systems or replace them entirely.