Every AI deployment eventually runs into the same wall: an exception that never made it into the documentation, a workaround nobody wrote down, a judgment call that depends on knowing the client's history. That context lives with the people doing the work, and it's why AI can't simply be installed on top of an organization and expected to perform on its own. It still needs the people who hold that context to shape how it gets used. Their capabilities deserve the same deliberate design as the technical architecture built to support them.
“Technical tool tutorials drive a negligible 7% of daily use, while future role clarity holds a massive 24% impact,” according to Gartner® research.[3]
We’re already seeing job functions reimagined as this takes hold. Engineering is the clearest case, as AI takes on more of the routine work and changes what a strong day-to-day contribution looks like. But the shift isn't limited to technical teams. Anyone bringing AI into daily work is asking a version of the same question: what does this role need to be able to do now that wasn't true a year ago?
That question is the starting point.
The Five Capabilities Framework
Once a team starts asking what capabilities it needs, a pattern emerges. Across large-scale AI initiatives, the same areas of expertise show up repeatedly. Together they describe five capabilities distributed across a team, united by a shared context and supported by an agentic backbone. However many people it takes to cover them, and whatever their titles happen to be, all five need to be present somewhere on the team.
The process capability is worth a closer look, because it's easy to mistake for a single role rather than a function that gets filled in different ways. At scale, it often means embedding directly with the people whose work is being redesigned, closing the distance between a sound idea and a working outcome. Some enterprises call this a forward-deployed engineer (FDE), common among service providers and consultancies, but the function itself isn't tied to that title or that setting. A product company or a B2C brand fills the same function by embedding with an internal business unit, a customer-facing team, or the operational staff who'll use the system. The title and the setting change; the capability doesn't.
Knowing the five capabilities is the easy part. Identifying which ones a team already has today, and where the gaps will open up next, deserves a closer look.
Sense, Develop, Deploy: Match Capability to Need
Most organizations still struggle to identify, develop or evaluate talent at the pace AI now demands, and it's rarely for lack of technology. Mid-to-large companies already run substantial people systems. The problem is that these systems rarely talk to each other and use different definitions, so data gets lost as people move through the talent lifecycle. They're built to describe where someone stands today, not to catch what's coming next.
That gap shows up in a telling imbalance. Only 38 percent of organizations maintain a single skills library, while 91 percent use smart building systems to monitor their office space.[4][5] Most businesses know more about the chairs in the building than the people sitting in them.
Closing the gap starts with knowing what the business needs, who already has that knowledge and where to deploy it. That's the discipline behind the Sense, Develop, Deploy framework.
None of this runs once and stops. Sensing, developing and deploying repeat continuously, each cycle revealing new gaps as the organization evolves.
Four Moves to Make Now
Move 1: Assess
Start with your current skills data instead of a list of jobs to be filled. Skip legacy HR systems and self-reported profiles, and pull data from task execution, peer feedback and project delivery. How well do your current capabilities cover the Five Capabilities Framework? What needs to change to close the gap, and how fast can that realistically happen?
Move 2: Explore
Build a picture of what future demand will look like. Gather signals from a wide range of sources, paying particular attention to emerging areas where the evidence is still thin. Waiting for established sources like job boards to catch up puts you behind the curve.
Move 3: Educate
Build continuous development into the flow of work so it can meet new needs as they emerge. Pair learning with immediate application on the job: observing, practicing and reflecting alongside any formal training.
Move 4: Measure
Track whether the capability being built is useful. Once each cycle completes, measure its impact and iterate. Are teams finding faster routes to market? Which skills are rising, retaining or retiring? Is the current approach meeting the organization's needs for skill, speed and scale?
EXAMPLE
In a survey of Wolters Kluwer workshop participants, 98% said EPAM's AI enablement workshop helped them identify AI opportunities relevant to their own area of the business.
Capability Is the Strategy
The biggest opportunity lies in matching capability to where value gets created, then staying flexible enough to hold that match as the work keeps shifting. That's true in whatever direction headcount is moving.
The two frameworks do this together: knowing the capabilities a team needs and where they currently sit, then keeping that picture accurate as demand changes. In practice, that means flipping the old build-it-and-they-will-come instinct. Instead of creating positions to fill, often by leaning on the same handful of overstretched people, run Sense, Develop, Deploy as a standing discipline.
Technical mastery alone won't hold an edge for long once it's easy to buy. What's harder to buy is a team that understands the tools, the data behind them, the workflows they reshape, and where the business creates value. Build that, and the technology has something worth running on.
[1] IDC, Work rewired: Navigating the human-AI collaboration wave
[2] World Economic Forum, Four futures for jobs in the new economy: AI and talent in 2030
[3] Source: Gartner Report: AI-Workplace Outlook 2Q 2026: Benchmarks and Trends on AI, Work, and the Workforce, By Diana Sanchez, Julia McDowell, Emily Rose McRae, CV Viverito, August 2026. Gartner is a trademark of Gartner, Inc. and/or its affiliates
[4] Mercer, The 2025/2026 Skills Snapshot Survey report
[5] ASHB, 2025 Smart Building Trends & Technology Adoption
