Practical leadership lessons from the U2030 RADIATE conversation with Jody Allison and Bashir Bseirani
This conversation began as part of the U2030 RADIATE curriculum. The insights were too valuable to keep inside the classroom.
AI is moving faster than most organizations can establish policies, redesign work or prepare their people.
Feeling uncertain is understandable. Refusing to engage is not a good idea.
Utility leaders do not need to become technologists overnight. But they do need to understand enough to ask better questions, establish boundaries, build trust and decide where human judgment must remain in control.
The central lesson from our conversation was simple: leading in an AI world is not primarily a technology challenge. It is a leadership challenge.
Here are ten ways to move from uncertainty to informed action.
1. Design for Trust—Do Not Ask for Blind Faith
When Bashir Bseirani first introduced AI inside Avertra, his greatest concern was not whether the technology would work. It was whether his own team would trust it enough to use it.
His response was to start small, keep humans involved and make each step visible. The system could not feel like a black box.
Trust is not something you can convince people of. It is something you have to design for—and show.
That is a powerful standard for any AI rollout. Explain what the system can do, what it cannot do, what information it uses, where a human reviews its work and what happens when it fails. Trust grows from transparency and evidence—not a launch announcement.
2. Remember: Confidence Is Not Correctness
AI can sound certain while being wrong. Bashir described an AI system that repeatedly claimed it had corrected an error when it had not. The lesson was not that AI is uniquely deceptive. It was that polished language and confident delivery can hide weak work.
Speed can create the illusion that the task is finished. Often, AI has only built the scaffolding.
We are becoming addicted to speed. Work that once took two weeks can appear in ten minutes, and the temptation is to treat the first output as the final product.
Do not confuse fast with finished—or fluent with true. Before using an AI-generated answer, ask:
Is it accurate, current and legal?
What assumptions is it making?
What evidence supports the conclusion?
What context, stakeholder or risk might be missing?
Would I be comfortable defending this decision?
The leader remains accountable, even when AI helped produce the answer.
3. Manage AI Like a Talented Junior Employee
One of Bashir’s most useful analogies was to treat AI like a junior employee who has extraordinary speed but limited organizational context.
A new employee would need a clear objective, the right files, defined boundaries, examples of good work, access appropriate to the role and a manager who reviews the result. AI needs the same.
Before opening a tool, write down:
What am I trying to accomplish?
Who is the audience?
What information actually matters?
What constraints, policies or risks must be respected?
What should the output look like?
What decisions must remain with a person?
This is more than prompt engineering. It is context engineering: organizing the facts, judgment and guardrails the system needs to perform responsibly.
4. Use AI as a Challenging Companion
Jody Allison encouraged leaders to stop treating AI like an answer center. Its greater value may be as a thinking partner that helps us test our reasoning.
Do not ask it only to validate your first idea. Ask it to pressure-test the idea:
Make the strongest case against this recommendation.
What could cause this plan to fail?
What would a customer advocate challenge?
What operational or workforce risks have I overlooked?
What evidence should I gather before acting?
Ask me questions that would help me think about this more deeply.
The goal is not to surrender judgment. It is to strengthen it.
5. Design the Journey Before You Add the Technology
Jody shared an early attempt to give contact-center employees easier access to information through an AI agent. The tool initially addressed a real problem, but the information became unreliable. Confidence fell, usage stopped and the organization had to rebuild trust.
The deeper lesson was that the organization had tried to place AI on a problem instead of redesigning the broader customer and employee journey.
Before asking, “Where can we add AI?” ask:
What should the future experience feel like?
Where is trust being lost?
Which steps create friction without adding value?
What should technology handle?
Where must a human remain involved?
Who will maintain the information and system after launch?
AI should help create the future state—not make a flawed process move faster.
6. Treat Adoption as a Culture Shift
An AI rollout is not an IT project with a training module attached. It changes how people learn, ask for help, make decisions and demonstrate value.
This is not a technology rollout. It is a corporate culture shift.
That means IT and cybersecurity cannot lead alone. HR, learning and development, operations, legal, communications and frontline employees need to shape the rollout from the beginning.
Create a safe place for people to ask basic questions without feeling exposed. Pilot with small groups. Watch how the work actually changes. Capture lessons. Build a network of champions who can coach colleagues privately as well as publicly.
Access without training creates risk. Training without psychological safety produces silence. Leaders need both.
7. Build Governance Into the Work—Not Around It
Governance cannot be a policy employees read once. It must show up inside the way AI is configured and used.
As AI agents gain the ability to retrieve information or take action, leaders should define:
What systems and data the AI may access
Which actions it may recommend versus perform
Whose permissions the AI must mirror
Which requests or data types should be blocked
How prompts, outputs and actions will be monitored
When a human must approve, intervene or stop the process
How knowledge, policies and procedures will stay current
A call-center employee should not gain access to information through AI that the employee could not access directly. The system’s authority should never exceed the authority of the person using it.
8. Decide Where Your Intelligence Should Live
The AI conversation is not only about which model to use. It is also about where the system runs—and where company intelligence goes.
For some work, an enterprise cloud platform with contractual protections, encryption, access controls, retention settings and audit capabilities may be appropriate. For highly sensitive or regulated uses, utilities may want to consider private-cloud, hybrid, on-premises or even air-gapped AI infrastructure.
Before selecting an environment, ask:
What information will the system receive?
Where will prompts, files and outputs be processed and stored?
Can the provider use our information to improve its models?
Who can access the data and interaction history?
Can the system honor existing permissions and retention rules?
What monitoring can detect or stop inappropriate use?
Does this use case involve critical infrastructure or regulated information?
“Not in the cloud” does not automatically mean secure, and “enterprise cloud” does not automatically mean unsafe. The right choice depends on the use case, the sensitivity of the information and the controls surrounding it.
9. Settle the Ownership Question Before It Becomes a Dispute
As people work with AI over time, they may create agents or digital personas shaped by their voice, expertise, preferences, relationships and institutional knowledge.
Who owns that intelligence when an employee changes roles or leaves?
Bashir suggested a useful distinction: a personal agent that supports someone’s broader career and likeness may belong to the individual, while a corporate agent that supports a paid role and learns from company systems may belong to the organization.
Real situations will be messier. HR, legal, IT, cybersecurity, records management and business leaders should define:
What is personal likeness or transferable skill
What is company knowledge, data or work product
What employees may retain or transfer
What must remain, be separated or be deleted
Who may create or use a person’s digital twin
What consent and disclosure are required
Do not wait for an employee departure—or misuse of a leader’s likeness—to answer these questions.
Take the 30-Day Leadership Challenge
Choose one utility process, employee experience or customer pain point that could benefit from being redesigned. Then:
Define the outcome you want.
Map the broader customer or employee journey.
Identify one small, low-risk AI use case.
Confirm the approved data, access and AI policies.
Select the right tool and provide clear context.
Keep a human in the loop.
Challenge and verify the output.
Record what required human judgment.
Share the lesson with your team.
The goal is not to become an AI expert in 30 days. It is to move from avoidance to informed action.
The Bottom Line
AI does not diminish the need for leadership. It raises the standard.
Information is abundant. Judgment is the competitive advantage.
The utility leaders who thrive will not be the ones who blindly accept AI—or reflexively reject it.
They will design for trust, set clear boundaries, challenge confident answers, prepare their people, protect organizational intelligence and use technology to create better human experiences.
Do not wait until you feel completely ready.
Start small. Learn visibly. Question relentlessly. Lead.
