Joel Rosenberg explains why enterprise AI adoption depends on people, trusted data, strong governance, business context and measurable outcomes.
Most enterprise AI programs do not struggle because the technology is incapable. They struggle because the organization has not changed with it.
A company can purchase an AI platform, give employees access, and announce that innovation has begun. Yet access alone does not create adoption. A capable tool can remain unused because employees do not trust it, leaders cannot connect it to business value, or the process around it was never redesigned.
This is the gap Joel Rosenberg has spent much of his career trying to close. As a data and AI leader, he has helped organizations move AI from experimentation into daily operations across senior living, healthcare, insurance and financial services.
In a conversation with The Executive Outlook, Joel put it plainly. “The technology gets the headlines,” he says. “Culture, governance and process determine whether AI actually delivers results."
Enterprise AI adoption is not simply a technology rollout. It is a change in how an organization thinks, collaborates, decides and works.
There is no single formula for introducing AI into an organization.
Some companies give employees access to enterprise AI tools and allow them to explore. Others create structured programs with training, approved use cases, governance policies and central oversight.
Neither approach succeeds automatically.
The right model depends on culture, readiness, regulation, industry and data sensitivity. Still, one principle applies across sectors. The people closest to the work must participate.
Marketing teams understand campaign friction. Financial analysts know which reports consume hours of manual effort. Operations teams see the exceptions and workarounds that rarely appear in strategy presentations.
These employees are often better positioned than a central technology team to identify where AI can create immediate value.
The goal is not uncontrolled experimentation. It is to give business users secure tools, clear boundaries, and permission to explore responsibly.
Give one curious employee a paid Claude or ChatGPT seat and watch what comes back in two weeks. It is real work, and it is genuinely impressive. It is also the ceiling. “That person can rebuild their own job,” Joel says. “They cannot reach the company’s data, they cannot secure it, and they cannot hand it to the person sitting next to them. That is where a personal win stops and an enterprise problem starts.”
Too much control can suppress experimentation. Too little control can create security, compliance and operational risk.
He is blunter about one temptation in particular. “A lot of companies were building their own internal ChatGPT, and what they are really building is a worse ChatGPT,” he says. “If your assistant is weaker than what your employees already use at home, you have not gained control. You have taught them to work around you.”
The better move is to use capable enterprise platforms, apply real guardrails, and spend internal investment on context, integrations, workflows, and governance. That is where the differentiation lives.
At one organization, 60 percent of roughly 700 employees were using AI within eight weeks of launch.
That result did not come from one training session or an executive announcement. It came from treating adoption as a shared organizational effort.
A top-down program gives direction, and employees hear a mandate. A bottom-up movement generates ideas, but it lacks priorities, governance, and any path to scale.
Sustainable enterprise AI adoption requires both.
Leaders must explain why AI matters and how it connects to business strategy. Frontline employees must have a genuine role in identifying problems, shaping solutions, and improving how the tools are introduced.
Joel worked directly with employees across business functions. They mapped the manual steps, the delays, the recurring pain, and the processes that ate whole days. Then they designed the solutions with the people who would have to live with them.
This created something more powerful than awareness. It created ownership.
Part of the adoption came from secure access. The other part came from applying AI to time-intensive processes employees already wanted to improve.
When people see AI reducing frustration and helping them contribute at a higher level, it becomes a practical tool rather than an executive initiative or a threat discussed in headlines.
Many organizations now feel both excitement and exhaustion around AI. Vendors are presenting new offerings. Business teams are experimenting independently. Security leaders are asking about data loss prevention. Legal and risk teams are reviewing acceptable use. Data leaders are trying to establish governance while the technology continues to change.
AI touches people, processes, and technology at the same time. Yet it also weakens the longstanding wall between business and technology.
Employees without a traditional technical background can create useful workflows. Engineers can use AI to understand business language and processes more quickly.
The strongest organizations will bring business expertise and technical discipline together from the beginning. Business users must follow governance and security policies. Engineers must understand the business outcome, not only the technical delivery.
They must build with the business, not simply for the business.
For years, data teams focused on dashboards, reports, and performance metrics. Those tools remain important, but expectations are changing.
Leaders increasingly want insights to reach them without requiring another login or search. They want systems that monitor the business, identify meaningful changes and surface what deserves attention through tools such as email or Microsoft Teams.
“I call it an operations advisor,” Joel says. “It understands the workflow, watches the data and tells you where to look.”
This represents a shift from passive analytics to active assistance. Promising agentic AI use cases are emerging in engineering, customer support, research, finance and back office operations. In customer support, agents can handle routine questions and direct complex issues to the right person. Operational agents can monitor data alongside documents and business knowledge, then bring important issues forward.
However, a successful demonstration is not the same as a successful deployment. This is especially clear in senior living, where a solution may perform well in a controlled setting but struggle inside the daily work of employees serving residents. Joel said, “Frontline environments are human and filled with exceptions. A tool must fit the reality of the work.”
The excitement around agentic AI can make traditional data foundations sound outdated. They are not.
Data warehouses, data lakes, integration, quality, governance, and access controls all remain essential.
If the information is incomplete or poorly governed, an advanced agent will only reach the wrong answer more efficiently.
What has changed is that structured data is no longer enough.
AI also needs context from documents, emails, meeting notes, call recordings, process descriptions, operating procedures, and the knowledge employees carry in their minds.
He calls this the context layer. “Your warehouse knows what happened,” he says. “The context layer tells AI why it happened and why it matters.”
Some companies are capturing internal conversations so they can preserve knowledge that was never formally documented. This may include how teams handle exceptions, why a process developed in a certain way, or which factors influence a decision.
The objective is not to collect information without purpose. It is to identify the knowledge required for a defined use case, organize it responsibly, protect it appropriately, and connect it with trusted enterprise data.
Without this context, an agent may sound intelligent while lacking operational relevance.
Measuring the return on AI investment remains difficult for many organizations. The first instinct is often to calculate time savings or expense reduction. Those measures matter, but they do not capture the complete value.
An AI initiative may help a company win new business, improve customer satisfaction, or allow employees to spend less time on administration and more time on judgment and problem-solving.
Employee satisfaction can also be a meaningful business measure. When people focus on higher-value work, they may feel more fulfilled and become more likely to remain with the organization. Retention matters because turnover creates financial and operational costs.
The right measurement begins with the business problem, not with a general promise of productivity.
Leaders should name the outcome before they name the tool. Faster cycle times, better service, more revenue, a stronger customer or employee experience, lower risk, quicker decisions. If a use case cannot be tied to a number the business already tracks, it is not ready to leave the pilot stage.
The number of users, prompts, agents, or experiments may show activity, but it does not automatically demonstrate value. The strongest AI portfolios connect adoption with results the business already cares about.
Resistance to AI is often described as a skills problem or a communication problem. At a deeper level, it is a human problem.
People develop pride in the work they have built. A process may appear inefficient from the outside, but it can represent years of experience, judgment, and contribution.
When leaders introduce AI without acknowledging that history, employees may hear that their knowledge is no longer valuable.
The challenge is not to remove that pride. It is to help people understand how their expertise can become more valuable when repetitive work is reduced.
Employees worry about their jobs. They also can become dejected when handed a tool they had no part in choosing.
Transparent communication is necessary, but participation is more powerful. When employees help identify the use case, design the workflow, test the solution, and define good performance, the change becomes something they are shaping rather than something happening to them.
Empathy is not a soft addition to AI strategy. It is an operating capability that helps leaders understand where resistance comes from, where trust is missing, and what employees need to move forward.
In practice, that means one thing. Ask the people who own the process what is worth keeping, before you tell them what you are changing.
For leaders who feel overwhelmed by the pace, Joel's advice is short. Start small.
The first step can begin with thirty minutes in a monthly or quarterly leadership meeting. Leaders can discuss what they are seeing, what they are learning and which questions AI is creating for the business.
The same practice can continue through department and team meetings. A standing conversation about how employees are using AI can reveal useful ideas already developing inside the organization.
When selecting formal use cases, leaders should look for processes with significant pain, repeated manual effort and manageable implementation complexity.
The ideal early project creates visible value without requiring the organization to solve every data, technology and governance challenge at once.
These successes create evidence. Employees see real problems being solved. Executives see measurable progress. Investors and stakeholders gain a clearer story about how the organization is applying AI responsibly.
Starting small does not mean thinking small. It means creating a disciplined learning cycle before complexity expands.
Enterprise AI adoption succeeds when leadership direction and frontline knowledge reinforce each other.
It requires capable technology, trusted data, governance, experimentation, context, measurement and empathy. No single element is enough on its own.
The strongest organizations will combine guardrails with freedom, strategy with participation, data with context and ambition with empathy. AI can be purchased quickly. Adoption must be earned through trust, relevance and results.
That is the real work of bringing AI into the enterprise. It is not simply about introducing a new tool. It is about helping people build a better way of working together.
Every transformation carries lessons that can help other leaders make better decisions.
The Executive Outlook features the perspectives of CIOs, CDOs, Data and AI leaders, founders, and global executives shaping business and technology.
Do you have a leadership journey, transformation story, or industry perspective to share?
Get yourself published in The Executive Outlook and bring your insight to a global executive audience.
Listen in for business strategies and interviews with business experts to help you build a business and a life that you love.
About Us
