AWS report delivers CxO view of AI projects

Amazon Web Services released a report on delivering value from AI projects based on interviews with 150 C-level executives and the biggest takeaway is that business outcomes need to lead and these efforts are about organization design and change management as much as the technology.

The other notable point from AWS’ report, "Reimagine Turning AI into Value," is that the massive report is based on qualitative interviews that are then sliced into topics.

What emerges are scenes from various projects and lessons learned from CxOs. A few common themes:

• Efficiency gains from AI are just a starting point.
• The capacity freed by AI should be focused on growth preferably in the same quarter.
• Business outcomes are the priority not adoption rates, output volume or other metric traps.
• Efficiency metrics should be combined with quality scores.
• The most critical human skills are judgment, problem framing and systems thinking.
• Proprietary data is the mote.
• Governance controls should be in the system architecture instead of human approval chains.
• AI costs should reside in the unit that is deploying the project.

The limitations of the report are fairly obvious: The viewpoints are decidedly AWS-focused and many case studies are Amazon programs. The report also highlights structural problems without proven solutions, but that's often because AI is developing quickly. It's unclear whether anyone has solutions for the junior talent gap or fleet governance of AI agents yet.

More from Esteban Kolsky:

• Workforce and Organization: Experienced Judgment Is the Constraint
• Business Value and the Limits of ROI
• A New Way to Look at Enterprise Technology in 2027+
• The CEO AI Dashboard Series: All the Posts

Nevertheless, the AWS report is worth a glance if only because the cloud provider serves up ideas you can steal with each section. Here's a select list.

• "Efficiency, productivity, and value are not the same thing, and one does not automatically lead to the next."
• "Most teams celebrate when AI makes a process faster or produce more. But that only matters if speed and quantity are linked to value." In other words, combine time, support cases and code to quality metrics.
• Amazon's adoption-to-value stages where you move users from access and exploration to discrete tasks to iteration and context, workflows that are reusable and process redesign.

• Hire for skills such as unlearning agility, problem framing, judgment over knowledge, system thinking and curiosity.
• Governance for agentic AI has four parts: Identity, access and encryption basics are priority then set up security boundaries outside the agent and start with human approval and once reliable let agents be autonomous. And finally, you'll need continuous testing.
• Leverage FinOps and make costs transparent to everyone, make a unit responsible for the costs and hold people who define AI uses cases accountable for the costs.
• Resdesign work by defining intent, constraints and business decision rules, making recurring work repeatable, using smaller units, leaving intent and judgment to humans, building what you can verify and treating every redesign as a hypothesis.