With 17 years of experience building large-scale distributed systems at Choice Hotels, American Express, and BILL, Sathiesh Veera draws on hands-on enterprise transformation work to argue that AI failure is almost never a technology problem, it is a readiness, security, and architecture problem.
-- Sathiesh Veera, AI Solutions Architect and enterprise software engineer with over 17 years of experience building large-scale distributed systems across fintech and hospitality, has joined Xraised for a conversation addressing one of the most consistent and costly challenges facing organisations adopting artificial intelligence: the gap between AI ambition and AI results.
Veera's position, grounded in direct enterprise experience and supported by McKinsey research, is direct. 95% of AI experiments fail to produce meaningful impact, and fewer than 10% of companies genuinely benefit from AI adoption. In his assessment, it’s because most organisations rush into adoption without the foundational readiness, data infrastructure, and governance architecture required to make AI work reliably at scale.
A Career Built on the Infrastructure That Makes Large-Scale Systems Work
Veera's perspective on AI adoption is shaped by a career spent building the systems that power personalised customer experiences at some of America's largest companies.
At Choice Hotels, he built the modern data platform that enabled the company to create a 360-degree view of travellers and power the redesigned Choice Privileges loyalty programme, launched in 2016, which introduced personalised rewards for over 25 million members across more than 5,500 hotels.
At American Express, he built the dynamic rules engine and workflow platforms that powered the Global Plan It programme, enabling expansion into international markets and giving customers flexible, instalment-based payment options across borders.
At BILL, the financial operations platform serving hundreds of thousands of small and midsize businesses, Veera built the Role-Based Access Control systems for the Identity platform that enabled the integration of accounts payable, accounts receivable, and spend and expense management into a single unified financial platform. He also created access management systems for AI agents at BILL, with a specific focus on customer data security and personal data protection as AI capabilities were embedded into the platform's workflows.
The Three Pillars Veera Identifies as Prerequisites for AI Adoption
In the Xraised conversation, Veera sets out a three-part framework he describes as foundational prerequisites rather than optional enhancements.
The first is AI literacy: educating all stakeholders, from engineers to leadership, on what AI can and cannot do before any deployment begins. AI remains prediction-based modelling, not human-level intelligence, and organisations that treat it as a general-purpose solution without understanding its dependence on data quality consistently encounter the failure rates the research reflects.
The second is data preparation, ensuring that the data feeding AI systems is carefully structured, contextually relevant, and properly governed, with retrieval-augmented generation and metadata contextualisation among the techniques Veera recommends, alongside pre-processing to remove sensitive information before ingestion.
The third is responsible AI use: Veera operates a governance board that reviews proposed use cases for return on investment and data sensitivity before approval, with ethical guardrails around the protection of vulnerable groups embedded from the outset.
Data Security as a Non-Negotiable Requirement
A significant portion of the Xraised conversation focuses on data security, an area Veera has worked on throughout his career and which he describes as a non-negotiable requirement in any AI workflow.
His approach involves multiple layered protections: data sanitisation in ingestion pipelines to remove personal information before AI processing, contractual agreements with vendors to prevent data misuse or training on sensitive client data, and sandbox environments with strict access controls to prevent accidental exposure or breaches.
Veera also addresses the broader cybersecurity dimension of AI adoption. AI both creates and detects vulnerabilities, making security a dynamic and continuously evolving challenge. Organisations that focus solely on the speed gains AI can deliver, without accounting for the data risk exposure that comes with it, are, in his assessment, making a significant strategic error. His guidance is to start with low-risk use cases, build security maturity incrementally, and avoid exposing customer data to AI systems until the governance infrastructure to protect it is fully in place.
His independent research on enterprise security, published through Al-Kindi Publisher, extends this work significantly beyond any single project or organisation. The research examines how the current pace and scale of AI adoption is systematically compromising security across industries, and proposes a reference architecture designed to address that challenge at an industry-wide level.
The reference architecture is built on years of direct enterprise experience combined with a thorough study of today's AI technology landscape, with a particular focus on the areas that are most commonly overlooked in standard security frameworks. Veera's position is that the vulnerabilities created by rapid AI adoption are not random or unpredictable. They follow patterns that can be identified, mapped, and systematically addressed through sound architectural principles applied from the outset. The result is a framework that companies across industries can draw on to build secure AI systems from the ground up, ensuring that the speed of adoption does not come at the cost of the security and data protection that enterprise environments require.
Why Internal Automation Is Not the Same as Customer Value
Veera also challenges a framing he observes consistently across enterprises adopting AI: the tendency to measure success by internal efficiency gains rather than customer impact.
Automating 30 to 50 percent of internal software work reduces costs but does not create competitive advantage, he argues. The organisations that will lead in an AI-driven market, in Veera's stated view, are those that use AI to rethink business processes and deliver new value to customers, not those that use it primarily to reduce headcount or streamline internal operations.
Architectural Flexibility and the Vendor Lock-In Risk
Veera also sets out a structural principle he describes as essential for any organisation building AI capabilities for the long term. The rapid evolution of AI providers means that tight integration with any single vendor is a significant operational risk. A regulatory change, a service disruption, or the emergence of a better-performing model can render tightly coupled systems fragile or inoperable.
His recommendation is to build abstracted adapter layers that allow organisations to switch between AI tools without disrupting core business workflows. Systems should be designed so that AI outages do not create single points of failure.
Veera and host Myles concluded the conversation with a point both agreed on. AI tools are more accessible than ever. But the true value of AI does not come from the tools themselves. It comes from how humans use data. Keeping humans in the loop, maintaining critical thinking alongside AI outputs, and designing systems that support rather than replace human judgement are, in Veera's view, the conditions that allow AI to generate sustainable business value.
About Sathiesh Veera
Sathiesh Veera is an AI Solutions Architect and enterprise software engineer with over 17 years of experience building large-scale distributed systems across fintech and hospitality. His enterprise work includes building the modern data platform at Choice Hotels, the dynamic rules engine powering American Express's Global Plan It programme, and the RBAC Identity platform and AI agent access management systems at BILL. He conducts independent research on AI security, RAG pipeline security analysis, and enterprise security architecture and has published research through Al-Kindi Publisher. He currently builds RAG pipelines, performs security analysis on AI workflows, and develops open-source libraries focused on customer data protection from AI exfiltration and attacks.
Contact Info:
Name: Gianmarco
Email: Send Email
Organization: Xraised
Website: http://Xraised.com
Release ID: 89194559
In the event of any inaccuracies, problems, or queries arising from the content shared in this press release, we encourage you to notify us immediately at error@releasecontact.com (it is important to note that this email is the authorized channel for such matters, sending multiple emails to multiple addresses does not necessarily help expedite your request). Our diligent team will be readily available to respond and take swift action within 8 hours to rectify any identified issues or assist with removal requests. Ensuring the provision of high-quality and precise information is paramount to us.
