Despite $30–40 billion in enterprise investment into AI 95% Get Zero Return. Here are 3 changes that actually work.

​

A healthcare SVP called me last month. Let's call her Samantha Jones. Her company was about to blow their AI budget on a CRM feature that would fail.

The plan sounded reasonable: mine CRM data to find and rate the best leads for marketing and sales. Sales reps were already using GenAI individually for their accounts, so why not build it natively into the CRM?

Here's why it would fail.

What We Found

My team reviewed their current state before they started building. We identified five critical problems:

1. Nobody knows how to use this stuff

Significant skill gaps across the organization. Most senior leaders lacked the AI knowledge to lead the effort effectively.

2. People think AI will take their jobs

Sales staff worried about job displacement. Others raised concerns about bias in training data and the "black box" nature of AI decision-making.

3. ROI mismatch

70% of the AI budget went to visible, "board-friendly" functions like sales and marketing. Almost nothing allocated to backend systems and workflows where the real work happens.

4. Missing the feedback loop

The selected LLM was static—it lacked persistent memory of client preferences or previous interactions. Without this "agentic" ability to learn, AI stays relegated to simple tasks like searching for leads and drafting emails rather than getting smarter over time.

5. Misalignment with backend workflows

The CRM system was treated as an isolated "science project." Nobody had examined the data for quality and completeness. Integration with existing tools like Salesforce would be complex.

The first three challenges are structural, not functional. You can't fix them with better technology. You need changes in leadership behavior and organizational structure.

​

3 Changes That Actually Work

We went back to Samantha with three specific changes.

Change 1: Build a Real Product Team

Currently, Sam has a project manager coordinating vendors, IT managing implementation, and sales giving requirements. That's the problem.

She needs one dedicated product team that owns this CRM AI feature from conception through iteration. Not a project team that dissolves after launch. A permanent squad.

The core trio:

Product Manager - Owns the "why" and "what." Talks to sales reps daily. Understands which leads actually convert. Decides what gets built next based on real usage data, not what leadership thinks would be cool.

Product Designer - Focuses on how sales reps actually interact with AI-generated lead scores. Are they trusting the recommendations? Ignoring them? This person observes real behavior.

Technical Lead - Makes the technical calls on which LLM to use, how to handle data quality issues, and whether the system can actually learn from rep feedback.

This trio meets daily. They make decisions together about what to test next. When a sales rep says the AI is surfacing terrible leads, they talk to more reps that day, look at the data, and test a fix by Friday.

I told Sam: "Give them a clear outcome: increase qualified leads identified by 20% while reducing rep time spent on research by 30%. Payback period under six months. Here's your budget. Go learn what works."

How to implement this week:

1. Pick the trio. Take your best product-minded person from marketing, your sharpest designer who understands sales workflows, and a tech lead who's opinionated about AI quality.
2. Give them a room (physical or virtual). They need constant contact, not coordination across departments.
3. Set one clear success metric. One thing that matters: qualified leads identified, or rep productivity, or forecast accuracy. Pick one.
4. Fund them for 90 days minimum. They need runway to test, fail, learn, and iterate.

According to research, companies with distributed decision-making are 2X more likely to outperform. This is how Google, Amazon, and Spotify structure their product teams. Marty Cagan covers this extensively in Inspired.

Change 2: Stop Being the Hero

Bill Joiner spent years studying over 600 managers across different industries. His research in Leadership Agility revealed something uncomfortable: only 10% of leaders have developed the agility needed to handle today's complexity and rapid change.

The other 90%? They're stuck using leadership approaches that worked when business moved slower, hierarchies were clearer, and answers came from expertise rather than experimentation.

Most successful managers land at what Joiner calls the "Achiever" level. They're strategic, delegate effectively, and drive results. But Achievers still operate from a "heroic" mindset—they see themselves as ultimately responsible for having the answers.

Catalyst leaders operate from a "post-heroic" mindset. They don't see their job as having all the answers. They create the conditions for their teams to develop innovative solutions collaboratively. Only about 5% of leaders operate at this level.

Here's what we did with Sam's leadership team:

We shadowed them for a week and watched three moments: when her tech lead proposed switching LLMs, when two team members disagreed on priorities, and when sales pushed back on timelines.

In each case, leadership jumped in with an answer. Classic Achiever behavior great at driving results, terrible for AI experimentation where nobody has the answers yet.

We gave the leadership team one practice: pick one meeting per week where you only ask questions. No answers. No resolving tensions. Just surface what the team is grappling with.

The first week was painful. Their teams kept looking at leaders to decide. Leaders asked: "What are three ways we could solve this?"

By week three, teams stopped waiting for answers.

The bigger change was stakeholder agility. Sam used to run monthly steering committees where she presented progress and managed concerns. We asked her to stop.

Instead: bring the VP of Sales to user interviews. Let him hear directly from reps why the AI recommendations felt untrustworthy. No mediation. No sanitized findings. Raw feedback.

Sales went from skeptical stakeholder to active partner in two sessions.

The hardest question came next. The leadership team asked: "If we're not making decisions, what's our job?"

I told them: "Your job is building the conditions where your team makes better decisions than you could alone. Clear purpose. Real authority. Direct user access. Protection from organizational chaos. You're not the hero anymore. You're the obstacle-remover and the coach."

What coaching looks like in practice:

Your product manager comes to you: "Should we prioritize the mobile app redesign or the API integration?"

Old you: "Do the API integration first. Mobile can wait until Q2."

Coach you: "Walk me through how you're thinking about this. What's the impact of each? What did customers tell you? What happens if we delay each one by a quarter?"

The first response is fast. The second response builds capability.

When someone makes a mistake:

Your tech lead chose the wrong database architecture. The migration is taking three times longer than planned.

Don't reassign the work. Ask: "What would you do differently if you could rewind three months? What did you learn about evaluating technical tradeoffs?"

When someone succeeds:

Your product manager just shipped a feature driving 20% increase in conversion.

Ask: "What made this successful? How did you know this would work? How can you apply this approach to your next feature?"

You're reinforcing the thinking process, not just celebrating the outcome.

Change 3: Use AI as Your Thought Partner

Geoff Woods' book The AI-Driven Leader helped me work through this change with Samantha. The core idea: stop asking "How can I solve this?" and start asking "How might AI help me solve this?"

Here's the prompt Sam used for her overall challenge:

AI became three personas for her:

The Interviewer asks questions that pull insights she didn't know she had. "What evidence do you have that sales resistance is slowing adoption versus skill gaps? Which challenge, if solved, would make the others easier?"

The Communicator helps her craft messages for different audiences. When she needed to present to the board, it structured her complex technical challenges into clear business language they'd understand.

The Challenger pressure-tests her assumptions before stakeholder meetings. "You're assuming the LLM's lack of memory is a technical limitation. Have you validated this with your vendor? Could the real issue be how you're structuring the prompts?"

How Sam used AI to handle a difficult conversation:

The sales VP was blocking the project. He worried AI would eliminate sales jobs but wasn't saying it directly in meetings.

Sam used AI to prepare:

AI asked about the VP's background, his team's current challenges, what he'd said publicly versus privately.

After ten minutes, it suggested: "Focus the conversation on how AI can make his team more strategic, not more efficient. Show him how lead scoring frees his reps from research grunt work so they can spend more time with customers. Frame it as elevating his team, not reducing headcount."

Sam took it further. She asked AI to role-play the conversation:

The practice session revealed she was leading with technology benefits instead of his team's pain points. She adjusted. The real conversation went completely differently.

How leaders can develop coaching mindset with AI:

You know the mindset shift is happening when:

• You feel uncomfortable giving answers even when you know them
• You catch yourself mid-sentence: "What you should do is—actually, what are you thinking?"
• You get more energized by your team's insights than your own solutions
• You measure success by how rarely your team needs you, not how often they seek you out

What Happened Next

Three months in, Sam's team stopped waiting for her to decide. They brought her well-reasoned proposals instead of half-baked problems. Two-week analysis cycles became two-day decision sprints.

Her VP of Sales, the former skeptic, came to her: "Can you teach my team this AI thought partner approach? My reps are spending 40% of their time on research. If they could use AI the way your team does, we'd transform their productivity."

The biggest change wasn't the AI features they built. It was how her team thinks.

Want to learn these approaches hands-on?

I'm launching an AI-Driven Leadership class in Q2 2026. Fill out this interest survey to get early access.

Additional Resources:

Real-world examples of leadership agility in practice:

• ​Agile Leadership at Rabobank​
• ​How Komerční Bank Achieved Competitive Success with Business Agility​

​

​

​

Anil Jaising, ​CST®​ On a mission to help Entrepreneurs and Product Leaders THRIVE, Unpack Product Innovation with AI Trainer, Product Consultant and International Speaker Follow me for real life case studies and learning videos.

Upcoming Workshops:

​