Ask The Analyst: If shadow AI has taken root, what's actually broken, and what fixes it?

Every week for The Briefing, UNLEASH’s weekly intelligence newsletter for senior decision-makers, we put an important question to the analysts, the true HR experts.

Got a question for them? You can ask put it directly to the analysts via The Briefing - make sure you’re signed up.

This week we asked: If shadow AI has taken root, what's actually broken, and what fixes it?

Here are their insights:

David Forry, SVP, Principal Analyst, Brandon Hall Group:

Shadow AI is usually a symptom, not the root problem. Employees generally don't use unapproved AI tools because they want to break policy. They use them because they have work to do, approved tools don't meet the need, or they don't know what "good" AI use looks like. Organizations often respond by tightening policies, but policy alone rarely changes behavior.

The organizations making the most progress combine governance with enablement. They establish clear guidelines for acceptable AI use, provide secure enterprise tools, invest in AI literacy across the workforce, and maintain human oversight for decisions that affect people.

HR has an especially important role because it sits at the intersection of technology, workforce capability, ethics, and change management. Shadow AI should be viewed as useful data. It highlights where employees see value, where workflows are inefficient, and where official AI capabilities are missing.

Rather than asking how to eliminate shadow AI, leaders should ask why employees felt the need to go outside approved channels in the first place.

When organizations pair practical governance with training, trusted tools, and a culture of responsible experimentation, shadow AI naturally declines because employees no longer need to work around the system.

Kathi Enderes, SVP Research, Global Industry Analyst, The Josh Bersin Company:

When employees use Claude or Perplexity alongside your company-approved platform, they're not rebelling; they're problem-solving. The approved tool doesn't do what they need and hasn't been meaningfully embedded into workflows.

Our research shows that value is realized not when AI is available, but when it becomes part of how work is performed. Skills gaps are the biggest barrier to AI success, yet most companies still treat AI capability-building as optional. Not surprisingly, 64% have made no real progress on AI transformation, using AI as a personal productivity tool rather than changing how work actually happens.

Shadow AI is not a compliance failure or a technology problem. It's a work design, capability, and culture failure, landing squarely in the CHRO's lap.

The fix isn't stricter governance or shinier tools. Instead, it's redesigning work with AI, building real capability across the workforce, and creating a culture where people are so well equipped and empowered they don't need to go looking elsewhere.

That's what it means to superpower your workforce.

Chris Havrilla, Chief Advisor & Analyst, AI Alchemists:

Shadow AI at scale is likely an organizational health signal: trust is broken - in tools, leadership, communication, and/or the employee-employer relationship underneath them.

People self-solve when:

• They don't believe the sanctioned path works
• Nobody has been clear about expectations, talked transparently about direction, pitfalls, challenges, and what “good” looks like and why
• The culture doesn't make it safe to say the official answer isn't working.

The part most organizations miss: measuring the wrong things long before AI arrived.

Shadow AI didn't create the measurement gap, but it sure has exposed it.

So, what fixes it? Not another policy or mandatory, top-down training on top of a broken foundation.

The fix in moving from systems of record to the systems of consequence that agentic AI delivers is consequence architecture - deployed before the tools, not after consequences have been inherited. That means clarity on what outcomes AI is meant to drive, governance built around decisions not just data, and a real readiness diagnostic before deployment, not a checkbox after.

Organizations that get ahead of shadow AI are the ones where employees understand and are connected to the outcomes, the why, trust the system and leadership, and have the architecture and support to use AI well.

Jason Averbook, Co-Founder, Now to Next:

Think about shadow AI the way you'd think about a detour. If the official road somewhere is closed, or the tollbooth takes forever, people don't sit in traffic; they find another way around. That is basically what two-thirds of your employees are doing right now, according to Verizon's latest breach report, since most AI use on company devices runs through personal, unapproved accounts instead of anything HR or IT built.

That's not defiance, it's just people navigating around a route that doesn't work for them. BlackFog's research found that 69% of C-suite leaders think that detour is worth taking, nearly double the rate among admin staff.

That's the stat HR should actually worry about: the people setting the rules are the ones most comfortable ignoring them, and employees notice.

Fixing this takes three moves, not one:

• Ask people which tools they're already using and why, instead of guessing at it.
• Build a sanctioned option that's genuinely as fast and useful as what they found on their own.
• Hold leadership to the same road as everyone else, since a policy nobody at the top follows is just a suggestion for people below a certain pay grade.