What a Bad Trail Recommendation Taught Me About AI and Data

by Craig Sanchez | September 15, 2026

What a Bad Trail Recommendation Taught Me About AI and Data

When your 19-year-old daughter asks if you'll take her backpacking before she goes back to school, there's really only one answer: “Yes, sweetheart. Let's go!”

I decided to use a popular AI model to help me plan the trip, giving it as much context as possible on desired geography, proximity to lakes, my aversion to permit lotteries, distance per day, elevation gain and other parameters critical to a fun and safe adventure. The results were amazing, offering multiple recommendations for routes with gorgeous views, advice on how to avoid crowds, and prescriptive difficulty levels that were within our comfort zone...until I spot-checked a few of the trails against a popular and highly trusted hiking app. To my alarm, the specs described by my favorite AI utility were woefully off. In reality, the route we intended to take was much longer than described, involved much higher elevation gain, and recent user feedback on the popular hiking app revealed details that starkly conflicted with what the AI had promised. We could have gotten into serious trouble. 

But that’s nothing new. AI hallucinates. So, what does this tale of thwarted adventure disaster have to do with the software company I work for, CData, you may ask?

Your AI product is only as good as the data it can reach

Nearly every software company is building AI into its products right now, which turns accurate, trustworthy output from a differentiator into a baseline requirement. And accuracy has less to do with the model than with what the model can reach. An AI feature reasoning over stale, partial, or disconnected data returns confident, well-worded, wrong answers. It's the software equivalent of sending a family up the wrong mountain.

CData is the market leader in data connectivity. More than 150 of the world's top software companies embed CData so their customers can reach disparate, third-party data sources from inside their applications. I'll allow myself one brag: Google, Salesforce, SAP, Palantir, Workday, and many others trust CData to power the connectivity layer inside their products.

The real cost of building connectivity yourself

When a customer asks "can your product connect to X?" and the answer is "not yet," credibility slips, deals stall, and churn risk climbs. Building the connector yourself trades that problem for two harder ones.

First, every sprint spent on connectivity is a sprint not spent on your core product. Connectivity work is a tax on your roadmap, and it never fully ends.

Second, connectors decay. APIs change, authentication standards evolve, and vendors deprecate endpoints. A connector that worked last quarter can quietly stop returning complete data this quarter, which is exactly what turns an AI feature into a confident liar. The maintenance burden isn't only an engineering cost. It's an accuracy risk you'd be signing up to carry forever.

What you get by embedding CData

Coverage on day one

CData ships hundreds of production-ready connectors across databases, SaaS applications, ERP and CRM systems, data warehouses, and more, all available the day you embed them. With immediate connectivity into these enterprise data sources, your AI features answer from live systems of record instead of filling gaps with guesses. Your customers stop asking "can it connect to X?" and start asking "what else can it do?"

Time to value in weeks

Embedding CData means extending connectivity in weeks, not months, allowing your GTM teams to say yes to more opportunities and your AI roadmap to ship on real data instead of waiting for a connector backlog to clear.

Reliability your customers can trust

The same connectivity layer runs inside products from the most demanding enterprise vendors in the world. Proven at scale, it feeds your AI features reliable data, so your customers can trust the answers your product gives them.

Maintenance you never own

When a connector breaks because of an upstream API change, CData’s team resolves it. Your product and engineering teams stay on your roadmap instead of firefighting someone else's schema change.

Let your product choose the right path by embedding CData

Why reinvent the connectivity wheel when CData already maintains a fully supported connectivity layer your product can embed?

At CData, we don't want your product leading customers down a path that's misleading, untrustworthy, or dangerous. We're here to help.

And did I mention we like hiking? Pick a trail and let's talk.

https://www.cdata.com/embed/

 

Software vendors: ready to see more?

Test out CData Embed in your tech stack.

Request A Discovery Call