Don’t bet the future on shiny AI—build your rails first

Boardrooms are mesmerized by “GPT,” “Gemini,” and the promise of AI magic. The hype is real, but the hidden risk is even bigger. If you want your company to lead—not just follow—you need to see past the shiny surface. The smart money isn’t just chasing algorithms. It’s buying the rails beneath them.

Why Salesforce’s $8B Informatica deal really matters

You might have seen the headlines: Salesforce is acquiring Informatica for $8 billion. Informatica isn’t a household name like OpenAI, but every AI leader should care deeply about this move.

Informatica does the dirty work. It cleans, governs, and integrates data. That’s not glamorous, but it’s absolutely foundational. As I wrote in The Intelligent Enterprise, AI is only as smart as the data it ingests. Without this backbone, your AI dreams are just that: dreams.

Most enterprise data is a disaster

Let’s be honest. Most companies run on data silos:

• Customer records in Salesforce

• Financials in SAP

• Marketing in Adobe

• Product analytics in Mixpanel

Each system speaks a different language. None connect seamlessly. Every integration is a brittle patchwork. So when you try to deploy AI on top of this mess, what happens? The predictions are unreliable. Insights are wrong. Executives lose trust. Projects stall.

It’s no wonder nearly 80% of AI and machine learning projects stall before reaching production [Gartner, 2022]. The culprit? Poor data infrastructure—not “bad AI.”

As outlined in the Data Supply Chain framework from The Intelligent Enterprise, every AI ambition stands or falls on how well you manage the journey from raw data to actionable insight.

The new railroad wars: Data pipelines as the battleground

This moment echoes America’s railroad boom in the late 1800s.

Back then, whoever controlled the rails controlled the economy. Towns flourished or vanished based on track access. The real fortunes weren’t made from the goods shipped—they were made by owning the infrastructure that moved those goods.

In today’s digital economy, your data supply chain is the infrastructure that determines whether you win or fade into irrelevance. Data pipelines are the new rails.

And right now, two competing visions are fighting for dominance:

1. Private rails (integrated stacks) Tech giants want you on their end-to-end railways:

• Microsoft: Azure + Dynamics + OpenAI

• Oracle: From database to cloud applications

• Google: BigQuery to Vertex AI

• Salesforce: Now bolstered by Informatica

They promise efficiency, integration, and simplicity. But just like the railroad monopolies, their real business is lock-in.

2. Public rails (open ecosystems)

Others are betting on open, interoperable “public tracks”:

• Databricks ($62B valuation, as of 2024): Open data standards, flexibility

• Snowflake ($50B): Multi-cloud, independent data cloud

• Elastic ($9.6B): Search and analytics everywhere

• Confluent ($7.4B): Real-time streaming across platforms

These firms offer freedom and choice. But as with public railways, you must manage more complexity yourself.

If this sounds familiar, it should. The early cloud era started open and affordable. But soon, vendor lock-in and expensive exit fees arrived. Today, smart companies run multi-cloud strategies to stay nimble.

The same hybrid reality is emerging with AI. As outlined in The Intelligent Enterprise, modular architecture isn’t optional anymore—it’s a survival trait.

The executive playbook: Four actions you need now

If you’re not already locked into a proprietary ecosystem, take these four steps—immediately:

1. Audit your data to surface silent failures before they derail your AI strategy. Broken records, missing values, stale tables, and inconsistent formats are often invisible until they show up in a hallucinated chatbot response or a corrupted model input. Rather than relying on periodic manual checks, embed automated profiling and validation into your pipelines. Focus first on high-impact domains—customer, product, and financials—and build a living data health score that’s tracked over time. Auditing isn’t a one-time project; it’s a continuous signal loop that keeps the rest of your stack honest. Assign ownership and treat data as a product, with accountable “data product managers” for each domain.

2. Implement data lineage to understand how data flows through your systems and who or what it touches along the way. When an upstream source changes or a column is renamed, downstream models, dashboards, or LLM prompts may break silently. Lineage gives you traceability and impact analysis: Which reports depend on that field? Which AI features are now at risk? Layer in ownership metadata so teams know who’s responsible—and who to call when the data goes sideways.

3. Invest in data quality to shift from reactive firefighting to proactive prevention. While auditing catches existing issues, data quality engineering helps you avoid them altogether. Define quality contracts for key tables—rules for completeness, validity, and freshness—and bake them into CI/CD. Tag and promote “gold” datasets once they’ve passed consistently for a defined period, and track quality coverage as a first-class platform metric. When quality is treated like uptime—monitored, owned, and budgeted—trust and reuse increase across AI and analytics teams. Dirty data breaks AI. Clean data makes it sing.

4. Establish data governance to align people, policy, and control. Governance turns data from a technical asset into an accountable, compliant business product. Start by assigning ownership for each critical domain, and define stewardship responsibilities beyond the data team. Apply policy-as-code to enforce row- and column-level access, redaction, and retention, especially for sensitive or regulated fields. A modern governance layer includes a shared glossary, lineage, and policy tags that travel with the data, regardless of which “rail” it runs on. Without governance, even clean and audited data can become a liability.

Don’t get distracted by model hype: Secure your advantage

In the Gilded Age, the winners weren’t those who owned the fanciest trains. They were the ones who controlled the best routes. Today, your data infrastructure is that route.

Don’t be dazzled by the latest model demo or AI headline. Invest in your rails. Your competitive edge, your future innovation, and your board’s confidence depend on it.

Ready to get serious about your data foundation? It’s the one bet you can’t afford to lose.

If you don’t know the health of your “rails,” you’re not ready for the AI future. Start your data audit—today.

• AI
• Artificial Intelligence
• Data
• AI hype
• Data rails