SAN ANTONIO — A growing disconnect between boardroom directives and enterprise trust in artificial intelligence is creating a governance crisis, according to Perry Robinson, CEO of RocketDocs, a San Antonio-based provider of knowledge management and secure AI solutions. In the latest episode of The Building Texas Show, published June 6, 2026, Robinson told host Justin McKenzie that roughly 80% of boards are pushing their companies to adopt AI, yet only about 20% of those companies actually trust the tools enough to deploy them. The episode, titled Why 80% of Companies Don't Trust AI (And They're Right), explores why that gap is widening and what regulated industries are doing to address it.
Robinson, who joined the 30-year-old RocketDocs three years ago, framed the company's philosophy around a single line: "Policy is a promise, architecture is a guarantee." He warned that contractual language alone cannot protect corporate IP once employees start routing sensitive data through public models. The conversation highlighted the risks of 'shadow AI,' where employees paste proprietary data into free ChatGPT, Claude, or Gemini sessions, often without realizing that those interactions can be used to train the very models their companies compete against. "If you're not paying for the product, you are the product," Robinson said, referencing the free tiers of OpenAI, Anthropic, and Google products.
The discussion also covered several key trends shaping enterprise AI adoption. Robinson noted that RFP management and knowledge management are critical for life sciences, healthcare, insurance, and financial services customers, who face heightened regulatory scrutiny. He pointed to Atlassian's recent policy shift on training customer Jira and Confluence data, which signals a broader movement among SaaS vendors to clarify data usage rights. Additionally, the EU AI Act, coming into force this summer, introduces revenue-based fines for non-compliance, putting additional pressure on companies to ensure their AI tools are trustworthy and auditable.
To address these challenges, RocketDocs offers Luma, a secure generative AI layer designed to run entirely on a customer's own knowledge base inside their VPC, audited against ISO 27001 and SOC 2 Type 2 standards. Robinson described Luma as deliberately 'limited on purpose,' refusing to crawl the open internet so that answers stay grounded in approved, subject-matter-expert-signed content. The company also recently introduced a secure file transfer capability built for defense, law enforcement, and product launch scenarios where large, sensitive files cannot move by email. Buyers, Robinson noted, increasingly include AI governance committees, chief compliance officers, and general counsel negotiating AI addenda.
Salesforce's headless data moves also came up as a reflection of mounting pressure to feed AI pipelines. Robinson argued that companies must architect their systems to guarantee data safety, rather than relying solely on policies. The episode is available now on YouTube and wherever podcasts are heard.
