Shift*Academy

Shift*Academy

Share this post

Shift*Academy
Shift*Academy
Learning in the Age of Agents

Learning in the Age of Agents

Why What, Where, When and How We Learn Needs Reimagining for Intelligent Work

Cerys Hearsey's avatar
Lee Bryant's avatar
Cerys Hearsey
and
Lee Bryant
Jun 17, 2025
∙ Paid
8

Share this post

Shift*Academy
Shift*Academy
Learning in the Age of Agents
2
Share

Our capacity to learn and adapt will be just as critical to success in the algorithmic era as AI technology itself, both for individuals and the organisations they work in. But conventional approaches to learning will not be enough — what we learn, how we learn, where and when we learn are all likely to change.

We should expect learning to become a key part of our organisational infrastructure, and as such the experience of learning will be richer in a variety of ways.

  • Our tools will help us learn: personal learning agents will provide quick answers by summarising the organisation’s own knowledge and content, personalised to our own context and needs, whilst finding and filtering new learning content that might interest us.

  • We will also teach our tools, from reinforcement learning to guiding and overseeing agent behaviours, and curating the knowledge that AI systems rely on to avoid mistakes. We will be able to create and share practical learning guides such as process recipes, techniques and checklists that can help other people as well as AI agents.

  • By combining our organisation’s guidance, documentation and knowledge with instructional learning content, we can create learning agents that go beyond generic theory and methods to provide answers or suggestions that are context-specific, wherever and whenever we need them.

  • There will still be a role for well-designed courses and learning programmes, but increasingly this content will be accessible in-the-flow-of-work at a modular level as part of task support or in the context of dialogue with our personal learning support agent.

In an optimistic scenario, we can improve everybody’s employee experience by empowering them with learning where and when they need it, rather than treating them as drones who need to sit through boring training courses to perform boring repeatable process work. But for those who really grasp the opportunity, there will be few limits to how quickly and how far they can advance using the kind of continuous learning methods that are now possible.

And this is not some imagined future. Leading companies are exploring this already.

Uber CEO Dara Khosrowshahi, speaking at Brown University last month said:

“Within Uber…not enough of my employees know how to use AI constructively. Learning to use AI agents to code is going to be an absolute necessity at Uber within a year”

The message is clear: if you're not embedding learning directly into work, your workforce will be falling behind. We won’t build agent-ready organisations by sending people off to isolated courses, but by integrating learning into the flow of work: ambient, automatic, and always-on.

That means learning needs to become a foundational organisational capability - designed, resourced, and woven through every system, process, and platform.

To stay ahead in a world defined by intelligent systems and dynamic workflows, learning must be infrastructure: a living system of intelligence that fuses tools, data, services, and people to drive continuous resilience and adaptation.

We do a lot of work on how to better prepare leaders at every level to lead and manage teams in the algorithmic era, and we will be sharing some of our techniques and insights on the future of learning over the coming weeks and months.

To kick off, in this edition of Shift*Academy, we ask what would it take to build learning-as-infrastructure? How does it diverge from traditional L&D, and what are the practical signals emerging from organisations already making this shift?

From Episodic Training to Embedded Capability

Most organisations still approach learning as a series of discrete events: onboarding, annual training, mandatory refreshers. or perhaps leadership development programmes. These are scheduled, standardised, and often abstracted from the flow of work. Success is measured in completions and certificates, not in adaptability or sustained capability.

Reframing learning as infrastructure marks a profound shift in organisational design. It asks us to start treating it as a service layer - a responsive, intelligent capability that underpins day-to-day performance, experimentation, and evolution.

What can we, as individuals, expect from this shift?

This shift redefines how we think about workforce development. It is not just about upskilling individuals. Instead, it becomes a way to:

  • Detect Learning Moments: Identify when transitions, failures, or new contexts signal a need to learn, adapt, or seek support.

  • Enable Contextual Support: Deliver guidance, feedback, and resources directly within tools, workflows, and AI environments.

  • Strengthen Collective Intelligence: Turn individual insights and problem-solving into team-wide knowledge and reusable practices.

The real shift is not about delivering more learning. It is about becoming a learning organisation by design.

Key Applications: Learning as Infrastructure in Action

Here are four high-impact applications that show what’s possible when learning becomes infrastructural, not just instructional:

  • In-flow support at the point of need: Learning becomes contextual and immediate, surfacing just-in-time guidance inside the tools people already use. For example, a customer service agent handling a new escalation type in Salesforce might receive a summarised protocol, similar past cases, and a GenAI-powered walkthrough, without leaving the interface.

  • Continuous capture and reuse of tacit knowledge: Team rituals like retros, debriefs, and incident reviews can feed directly into an evolving knowledge base, structured through promptable documentation systems. For instance, an engineering team might use a simple prompt to turn their sprint retrospective into reusable guidance for new team members or other squads.

  • AI-augmented onboarding and transitions: Rather than relying solely on static materials, onboarding becomes adaptive. A new hire in a product team could be guided by a GenAI tutor that walks them through past project logic, key decisions, and common pitfalls - tailored to their role, tools, and immediate context.

  • Capability co-ordination across teams and services: As AI agents proliferate, teams need to coordinate not only work but also learning across disciplines. A transformation team rolling out new AI tools might embed prompts and nudges into Slack or Microsoft Teams that direct users to peer walkthroughs, feedback channels, and live support options triggered by task type or common blockers.

These examples show that learning infrastructure is not about layering more content into the workflow, but rather about creating systems that sense, respond, and evolve alongside the work itself. This is what enables organisations to treat learning not as a cost centre, but as a driver of strategic agility.

Real‑World Signals

The shift toward embedded learning infrastructure is already taking shape. Here are some examples of how organisations are putting this into practice:

1. Microsoft 365 Copilot: learning embedded in tools

With Copilot now integrated across Outlook, Word, and Teams, organisations have started using it as an in‑context tutor, not just an assistant. For example, SHI - a global IT provider - reported that Copilot improves meeting summaries, boosts productivity, and supports faster onboarding right inside existing workflows.

2. Law firms gamifying AI learning

Firms like MinterEllison, Gilbert + Tobin and King & Wood Mallesons have launched creative L&D programs that embed AI learning into work. At MinterEllison, they are using internal “mintcoins” for AI training, AI-bounty innovation incentives, belt‑style certifications, and “AI investigator” placements. These in-context programmes blend learning with real contributions

3. Open‑source agents as enterprise learning pilots

Frameworks like AutoGen and LangChain are enabling internal use cases where teams employ promptable agents for documentation, troubleshooting, or orchestration. One Fortune 500 R&D team, for instance, used AutoGen-powered agents to collaboratively brainstorm, refine, and summarise technical ideas in real time

4. Voice‑AI agents supporting frontline staff

Health‑tech organisations like Cencora’s clinics are deploying voice‑AI agents (e.g., Eva by Infinitus) to manage administrative tasks - benefits verification and empathetic patient check‑ins - freeing clinicians to focus on care, while patients receive timely assistance

Building the Capability: From Events to Embedded Infrastructure

Creating learning as infrastructure requires more than platforms or content libraries. It demands a capability that spans systems, data, software, services, and skills, designed not for episodic training, but for continuous, embedded development.

This shift calls for a layered approach: one that advances through improvement loops, not ‘launches’; adapts through use, not just design; and brings learning to the flow of work, rather than pulling people away from it.

Two strategic principles can guide this build:

  • Capability Mapping: Mapping learning as a capability means surfacing the interdependencies across platforms, roles, and routines. It helps organisations move beyond LMS silos to identify where learning can show up in work. Done well, it reveals reuse opportunities, duplication risks, and system disconnects, while anchoring learning design in the actual rhythms of delivery, decision-making, and change.

  • Loops and Layers: Rather than launching a static system, organisations should treat learning infrastructure as a set of dynamic, interlinked loops. Early loops might involve in-flow nudges or prompt libraries; later loops could coordinate multi-agent support, generative knowledge capture, or role transitions. Each loop strengthens the next, turning scattered learning moments into coherent, evolving infrastructure.

Together, these components create a responsive learning environment - one where every task can be a teacher, every transition a development opportunity, and every team a node in the organisational learning loop.

Read on for our detailed guide to building this new type of learner experience.

Keep reading with a 7-day free trial

Subscribe to Shift*Academy to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2025 Shiftbase Ltd
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share