The agent landscape two years ago looked different from today. Better models, more reliable tool use, richer orchestration. The landscape two years from now will look different again. What’s coming—and how do we prepare?

Emerging Capabilities (and Their Implications)

Longer context, better memory.
Agents that maintain context across sessions, remember past interactions, and build project-specific understanding. Implications: less prompt repetition, more continuity, agents that “know” your codebase. Change impact: Knowledge shifts from “how to prompt” to “how to curate context.” Ability shifts toward context management.

Multi-modal and tool-rich.
Agents that see (images, UI), act (APIs, terminals, browsers), and integrate more tools. Implications: agents that can do more end-to-end—research, design, implement, deploy. Change impact: the bridge from prototype to production may shorten. But orchestration complexity grows. Reinforcement needs to account for more autonomous capability.

Specialization and fine-tuning.
Agents tuned for specific domains—legal, healthcare, your codebase. Implications: better output quality in narrow domains. Change impact: the taxonomy (Augmentation, Automation, Orchestration) may need new categories. Multi-agent orchestration becomes more common as specialists proliferate.

Improved reasoning and planning.
Agents that chain steps, recover from failures, and adapt plans. Implications: more reliable autonomous workflows. Change impact: human checkpoints may be less frequent—but the stakes of “wrong” agent decisions go up. Judgment and oversight remain critical.

How to Stay Ready

1. Invest in principles over tactics.
The specifics will change. The principles—ADKAR, agentic patterns, the bridge, human value—are more stable. Build organizational muscle around those. Tactics (which tool, which prompt) will evolve.

2. Maintain the human role.
Regardless of capability gains, strategy, judgment, coherence, and learning stay human for the foreseeable future. Reinforce those roles. Don’t let them atrophy as agents get more capable.

3. Design for evolution.
Assume your agentic workflows will need to change. Build in flexibility: versioned prompts, pluggable agents, clear boundaries. Avoid lock-in to today’s tooling.

4. Keep learning.
The teams that thrive will be the ones that treat this as a learning journey—not a one-time adoption. Retros, sharing, experimentation. Stay curious.

The next wave will bring new opportunities and new complexity. The framework we’ve built—ADKAR meets agentic—is a starting point. Adapt it. Extend it. And keep asking: what only humans can do now? That answer will keep shifting. Our job is to stay ahead of it.