Why 95% of AI Investments Fail—and How Leaders Can Rethink Productivity

The GenAI Divide: State of AI in Business 2025, a new report published by MIT’s NANDA initiative reveals a critical challenge: 95% of generative AI investments yield no measurable return.  Despite billions invested, most organizations see little impact on revenue or efficiency. Our hypothesis at Ilja Stucken | Transformation Partners is that this divide arises from asking the wrong questions about AI implementation. Instead of focusing on technology alone, leaders should consider a shift: productivity comes from designing human-AI teams with clear roles. Here’s how you can lead this transformation in your organization.

Reframing the AI Question

The current narrative often asks, Will AI replace human jobs? We believe this is the wrong question. Leaders should instead ask: What tasks should AI handle, what tasks require human expertise, and how do we design teams to combine both effectively? Productivity doesn’t come from AI replacing humans—it comes from human-AI teams where roles are clearly defined, with humans engineering the context that makes AI effective.

The Role of Context in Human-AI Teaming

AI excels at processing data, but its value depends on the context humans provide. For example, before a doctor’s visit, I use AI to prepare by inputting my symptoms, medical history, and test results, asking for likely causes, questions to discuss, and treatment options. The output depends on the quality and context of my input. If I provide test results from a flu recovery, the AI might misinterpret them. If I don’t specify my objectives—like training for a marathon next week—it might suggest everything is fine, leading to flawed recommendations. We now have a new player in the team (AI), but it does not replace the doctor—it enables better decisions, faster, with proper setup. This principle of context applies equally in business settings, as our cost control process shows.

Case Study: Our Human-AI Team in Cost Control

At Ilja Stucken | Transformation Partners, we transformed our monthly cost control process by building a human-AI team, aligning with our mission to make technology transformation work. Previously, we spent hours verifying invoices, tracking payments, forecasting liquidity, and identifying savings—tedious work prone to errors. Now, we’ve assigned AI a specific role. It’s customized with our vendor contracts, historical forecasts, and cash flow patterns. It checks invoices for accuracy, matches them to agreements, and confirms payments.

One month, it flagged a $500 uncontracted setup fee on a rental invoice. Another time, it reminded us to chase a $1,600 reimbursement we’d overlooked. It uses deep research to benchmark expenses, such as subscriptions or advisor fees, warning us before auto-renewals and suggesting consolidation options. It builds a liquidity forecast to balance cash reserves, avoiding shortfalls or idle funds, and reconciles bank statements, updating projections with new insights. This saved us 10 hours monthly and started to reduce our expenses after only 2 months of piloting the concept.

Humans remain essential: we define the AI’s role, select relevant data, and set expectations for accuracy and optimization. This human-AI team boosts productivity by context engineering—selecting relevant vendor data to define which tasks AI should handle and which require human judgment. This ensures AI handles repetitive tasks while we focus on strategic decisions.

Addressing the Skeptics: Unlocking New Questions

Some leaders ask, “Won’t AI eventually do everything humans do now?” We believe this assumes productivity is finite—but it’s not. Each task AI automates unlocks capacity for humans to ask more insightful questions, driving a cycle of improvement. In our firm, AI shifted our cost control focus:

•  Fixing: Correcting errors, like a wrong invoice.

•  Preventing: Creating firm forecasts to avoid issues.

•  Using to Advantage: Optimizing subscriptions through deep research.

This cycle—from fixing to preventing to leveraging, each step requiring greater context—is a key insight. If designed effectively, human-AI teaming elevates work, empowering teams to address higher-value challenges.

The Leadership Shift: Designing Human-AI Teams

To unlock AI’s productivity potential, leaders must build human-AI teams with clear roles. AI excels at data-heavy, repetitive tasks; humans excel at strategic judgment and context-setting. However, individuals in your workforce often lack a full view of processes, tools, AI capabilities, and context needed for productivity. Drawing from best practices in cloud implementations, where a Centre of Cloud Excellence (CCoE) drives transformation, we propose a Centre of AI and Data Enablement (AIDE). This team shadows existing teams, such as our accounting team in our cost control process to analyze invoice workflows and identify tasks like invoice verification for AI, enabling effective integration.

To guide this process, the AIDE team asks:

•  Which workflows benefit most from AI automation?

•  What tasks require human creativity or oversight?

•  What data is relevant, and how should it be structured?

•  What accuracy is needed, and how is it measured?

•  What optimizations should AI prioritize?

•  Who in your organization defines these roles?

In most companies, this knowledge is implicit—scattered across habits and spreadsheets. It is a critical responsibility of AI leadership and the AIDE team to make it explicit, driving context engineering that enables AI to amplify human work for productivity enablement. This is the essence of human-AI teaming.

Your Path to Productivity

The GenAI Divide: State of AI in Business 2025 report from MIT reveals that only 5% of companies achieve ROI with AI, leaving the reasons for success unclear. At Ilja Stucken | Transformation Partners, we propose a critical shift: instead of focusing on technology alone, leaders must design human-AI teams to unlock productivity. With a Centre of AI and Data Enablement (AIDE) to assess workflows and define roles, and through context engineering—as shown in our cost control process—teams can drive cycles of fixing, preventing, leveraging for transformative outcomes.

How will you lead this transformation?