Building the Foundation for AI: Reflections from Node Talks

I recently had the opportunity to join a panel of technology leaders at Avinode Group's Node Talks for a conversation titled Building the Tech Foundation for Tomorrow.

The discussion was moderated by Alex Hanrahan, Senior Product Manager at Avinode Group. I was joined by Dudley King, Founder and President of FlightBridge, and Andrew Morelock, CTIO at BOND.

We approached the conversation from different backgrounds, but we kept returning to the same conclusion: AI can create remarkable speed and efficiency, but it cannot compensate for weak processes, poor data, or a workforce that has not been prepared for change.

The technology may be new. The leadership challenges are not.

AI Does Not Fix a Process You Don't Understand

There is a tendency to approach AI as if it were a switch. Connect a model, add a chatbot, and efficiency will appear.

The results can feel almost magical once everything is working. But reaching that point requires a significant amount of work.

Before attempting to automate a process, a company needs to understand how that process actually operates:

• What outcome is the process supposed to produce?
• Where does the required information come from?
• Who makes decisions along the way?
• What exceptions occur?
• Which parts require human judgment?
• How will success or failure be measured?

If a workflow is poorly defined, AI will not make it well defined. It may simply perform the wrong process faster.

This is why I encourage companies to begin with the business problem rather than the technology. Identify a specific source of friction, understand it from beginning to end, and then decide where AI can create measurable value.

AI Readiness Starts With People

Andrew emphasized the importance of giving employees a baseline understanding of AI before expecting them to incorporate it into their work.

I agree.

AI literacy cannot be limited to the technology department or a small innovation team. Employees need to understand what these systems do well, where they struggle, what information is appropriate to share, and when an output requires additional scrutiny.

That does not mean everyone needs to become a machine-learning engineer. It means people need enough familiarity to use the technology thoughtfully.

Without that foundation, organizations tend to experience one of two extremes. Employees either avoid AI because they do not trust it, or they trust it too quickly because they do not understand its limitations.

Neither creates sustainable adoption.

Your Data Becomes Part of the Product

The panel also discussed the role of data in practical AI adoption.

An AI system is only as useful as the information available to it. If customer records are incomplete, operational terminology is inconsistent, or important decisions exist only in someone's memory, the model is starting from a weak foundation.

This is especially relevant in business aviation, where scheduling, aircraft availability, crew requirements, customer preferences, airport constraints, and vendor coordination can all affect a single trip.

Small data problems can become significant operational problems when they are repeated at scale.

Data quality work may not generate the same excitement as an AI demonstration, but it is often the difference between a useful system and an expensive experiment.

Companies preparing for AI should know:

• Which systems contain authoritative information
• Who owns and maintains that information
• How systems exchange data
• Where duplicate or conflicting records exist
• What information should remain restricted
• How the accuracy of an AI-generated answer will be verified

AI readiness and data governance are becoming inseparable.

Adoption Is a Change-Management Problem

Dudley offered an important reminder: even a technically excellent system can fail if people prefer the workflow they already know.

In aviation, that might mean someone continues calling an FBO because it is familiar, even when a digital system could complete the same task more efficiently.

This is not necessarily stubbornness. Existing habits often represent years of experience and trust. Leaders cannot expect those habits to disappear because a new tool has been deployed.

Adoption requires confidence.

People need to understand why a workflow is changing, how the change will help them, and what support will be available while they learn. They also need permission to provide feedback when a new process does not reflect operational reality.

Technology deployment is an event. Adoption is a process.

Experts Still Own the Standard of Quality

One of the most important questions during the panel was how companies can trust AI-generated output.

My answer was simple: check the work.

As I said during the discussion, "You don't replace the experts, but you enable the experts with technology."

Domain expertise becomes more important when AI is introduced, not less.

The model may be able to summarize information, identify patterns, prepare a recommendation, or complete a repetitive step. But an experienced professional still understands the context, recognizes unusual conditions, and knows when an answer does not make sense.

The goal should not be to remove experts from the process. It should be to remove unnecessary friction from their work.

If technology can reduce the time an experienced employee spends searching, copying, formatting, or reconciling information, that employee can spend more time applying judgment and serving the customer.

That is augmentation—not replacement.

A Practical Starting Point for Leaders

For companies trying to move from AI experimentation to practical adoption, I recommend beginning with five steps:

1. Select a specific, repeatable business problem with a measurable outcome.
2. Document the current process, including its data sources, decision points, and exceptions.
3. Train the people involved on both the capabilities and limitations of the technology.
4. Pilot the new workflow with domain experts actively reviewing the results.
5. Measure quality, time saved, user adoption, and risk before expanding it.

A successful pilot should teach the organization something even if it does not immediately become a production system.

The objective is not to generate the most demonstrations. It is to develop the organizational capability to evaluate, deploy, and improve AI responsibly.

AI Is Already Part of Today's Business

Near the end of the panel, Andrew summarized the urgency well: "AI is not the future, it's the today."

Organizations should absolutely be moving forward. But the race is not simply to deploy AI faster than everyone else.

The real advantage will belong to companies that build the processes, data practices, workforce skills, and governance needed to use it repeatedly and responsibly.

AI can accelerate a capable organization. It cannot build that organization for you.

Thank you to Alex Hanrahan, Dudley King, Andrew Morelock, and the Avinode Group team for the thoughtful conversation.

You can read Avinode's recap and find the full panel discussion in Building the Tech Foundation for Tomorrow.

If your organization is working through the transition from AI experimentation to practical adoption, I would be glad to compare notes.