McKinsey’s 2025 report, Superagency in the workplace: Empowering people to unlock AI’s full potential, argues that AI’s real value comes when employees are amplified, not replaced by AI. Leaders are investing more, but maturity is rare, speed feels slow, ROI is uneven, and the biggest unlock is building the conditions where people, processes, data, and guardrails create compounding productivity (“superagency”). This blog breaks down the report’s findings, stress tests them against what we’re seeing in the market.
The report tackles this question head-on. It doesn’t focus on the next breakthrough model or billion-parameter architecture. Instead, it zooms in on a more practical, and urgent, topic:
How do we empower people to actually use AI – meaningfully, safely, and at scale?
The answer? A concept McKinsey calls Superagency.
Superagency isn’t a product. It’s not a feature, or even a technology. It’s a state of work.
McKinsey defines it as a condition where employees, supported by AI, can dramatically amplify their creativity, efficiency, and impact. It’s not about automation replacing people, but about augmentation empowering them.
Imagine a business analyst who can generate BRDs, user stories, and impact assessments in minutes. A developer who can understand legacy code instantly and write unit tests without leaving their IDE. A customer service rep who can access personalized, accurate answers without toggling between 12 systems.
That’s Superagency in action.
It goes beyond just “using a tool”, it’s redesigning work so AI and humans compound each other’s strengths.
Despite over 18 months of surge in GenAI adoption, only 1% of business leaders say their companies have reached AI maturity. Most organizations are still stitching together pilots, proofs, and scattered automations.
Roughly half of C-suites think their organizations are releasing GenAI tools too slowly. Skill gaps, process friction, and legacy complexity are key barriers.
While 92% of executives plan to boost AI spend, many haven’t seen meaningful returns yet. The real challenge is turning prototypes into scalable, integrated workflows.
Employee optimism is highest in some functions that aren’t necessarily the largest near term value pools, while industry AI spend doesn’t always match sectoral economic potential. That misalignment leads to “feelgood” deployments that underperform on P&L.
Sales & marketing, software engineering, and customer operations surface repeatedly as high potential domains. But leaders often fund scattered tools rather than end-to-end workflow redesign, so benefits get stranded.
McKinsey shows sectors that spend heavily aren’t always the sectors with the largest modeled AI upside. This mismatch plus legacy IT constraints and complex approvals helps explain the speed/ROI tension.
About a quarter of executives report a defined GenAI roadmap and just over half have drafts. The issue isn’t having a roadmap; it’s whether it prioritizes valuemapped, dataready, guardrailed use cases with a clear path to deployment and measurement.
You can’t build intelligent systems on messy data. And most enterprises are still struggling with fragmented, siloed, and unstructured data. GenAI needs clean, structured, and retrievable information to work well. That requires foundational investments in data engineering, retrieval pipelines (like RAG), vector databases, and access governance.
Curate a stage gated use case portfolio tied to revenue/cost levers, not novelty. Each use case should have: owner, data/architecture readiness, security & compliance plan, deployment path, and success metrics.
Standing up copilots without fixing data access, lineage, and quality creates “impressive demos” that don’t scale. Establish governed connectors, RAG patterns with evaluation, and data policies before scale.
Trust grows when safety is built-in. Put the right policies, monitoring tools, and review workflows in place. Trust builds confidence and confidence drives adoption.
Balance executive-led redesigns with grassroots adoption. McKinsey recommends pairing leadership-driven initiatives with employee hackathons and training.
This two-way approach drives both scale and skill development.
Shift from generic training to role-based progression models. Prompt writing for BAs. Validation skills for QA. Context mapping for architects. Make learning relevant and useful.
One of the most valuable things you can do right now?
Decide where not to use AI.
Sounds counterintuitive but McKinsey’s report makes it clear:
If you try to “AI everything,” you’ll burn out your teams and dilute your wins.
Instead, focus where AI:
Superagency isn’t about full automation. It’s about wise orchestration, knowing what AI should do, what humans should own, and how they support each other.
McKinsey’s report doesn’t just show where the world is, it gives us a blueprint for where it’s heading.
The companies that embrace Superagency today are building something bigger than just productivity tools. They’re laying the foundation for adaptive, AI-native organizations, where people and technology scale each other, not compete.
At Prodapt, this is exactly where we operate.
We help enterprises reimagine how work happens with AI embedded at every layer of decision-making, collaboration, and execution.
Deliver measurable outcomes for your business with #PracticalAI. Let’s talk!
