People and AI
The Human Premium in an Agentic Era
“Workforce anxiety about AI is aimed at the wrong variable. Capability isn't the bottleneck — orchestration is: knowing which of the three tiers a problem belongs to, and keeping human judgment as the guardrail across all of them. For Canadian scaleups, that orchestration question is inseparable from a sovereignty question.”
Session Overview
Hosted the People and AI roundtable discussions at the Calgary Summit of ScaleUP Week 2026 — two days of executives, tech leaders, and entrepreneurs from across Canada. The room was deliberately cross-functional rather than technical: operators trying to decide what to automate, what to augment, and what to leave alone. Thanks to Janet DeSousa and Dr. Simon Raby, Ph.D. for the invitation to host, and to ScaleUP and Long View Systems for the platform. Post and discussion on LinkedIn.
The Problem Statement
AI capabilities are compounding faster than org charts can absorb them, and the anxiety that produces is real. But the roundtable consensus pushed back on the framing: the scarce resource in an AI-saturated org isn’t model capability, it’s human judgment. AI handles the heavy lifting; humans supply the guardrails, the empathy, and the strategic vision. The question leaders actually face isn’t will we be replaced — it’s how do we deploy three very different classes of AI without stripping out the ingenuity that makes the company worth scaling.
The Multi-Tiered AI Stack
- Copilots: Assistive AI aimed at individual daily productivity — drafting, summarizing, retrieval. Lowest risk, fastest adoption, shallowest structural change.
- Process-Based Agents: Automation of structured, repetitive workflows with defined inputs, outputs, and exception paths. This is where most measurable ROI lives, and where process debt surfaces fastest.
- Agentic Frameworks: Autonomous, multi-step systems tackling complex problems with planning and tool use. Highest leverage, highest governance burden, and the tier most often deployed before the other two are stable.
The leadership challenge is orchestration across all three at once — matching each problem to the right tier rather than defaulting to the newest one.
Key Takeaways
- The Human Premium: Human judgment has never been more valuable. We aren’t being replaced; we’re being elevated — but only where organizations deliberately keep humans in the decision loop rather than assuming the loop will hold on its own.
- Match the tier to the problem: Copilot, process agent, and agentic framework are not a maturity ladder to climb. They’re three tools with different risk profiles, and picking the wrong one is the most common and most expensive AI mistake in the room.
- Sovereign AI is non-negotiable for Canadian tech leaders: Data residency, algorithmic sovereignty, and jurisdictional control over hardware, model weights, and applications. Safeguarding the ecosystem from foreign law and the unchecked reach of global hyperscalers is a national strategic imperative, not a compliance checkbox.
- The Canadian momentum is real: The caliber of the conversation across executives, founders, and operators made the case that the constraint here is coordination, not talent.