AI isn’t just accelerating work — it’s saturating it. The risk isn’t simply that AI gets things wrong. It’s that humans slowly stop thinking independently.

I recently ran into Jamie Bono of EmergenceTek Group Inc., and he mentioned a term I hadn’t heard: cognitive sovereignty, which he defined as:

“The ability to maintain independent thought and agency in environments saturated with AI‑generated outputs. It’s more important than ever for your humans in the loop to develop and protect this.”

This isn’t philosophical. It’s operational. We must protect the ability to think, decide, and intervene.

Similarly, an educator (Kelsey Pomeroy, @kelsewhatelse) recently released a short video reframing how adults should talk to kids about AI. Instead of leading with morality — “AI is cheating,” “plagiarism is wrong” — she argues that we should lead with brain protection. Her metaphor is simple and sticky: Imagine your cognitive skills as a set of marker colors. When you’re learning, you’re supposed to practice using those colors. If AI does that work for you, you don’t just save time — you lose access to the colors themselves. Meanwhile, others keep developing a full palette — an entire sunset.

Her message to students is blunt: You have to protect your sunset at all costs.

The framing strips away moral judgment and focuses on capability. The reason we don’t rely on AI during learning isn’t because it’s “bad.” It’s because it quietly weakens the very cognitive muscles we’ll depend on later.

The silent failure mode of AI adoption

Most AI strategies optimize for speed, scale, and efficiency. Very few account for what happens when humans begin to outsource judgment:

  • When AI gives you an answer but you can’t explain the assumptions behind it
  • When a summary becomes the decision — without the underlying data, definitions, or edge cases
  • When “the model said so” turns into the default owner of risk

Over time, this erodes cognitive sovereignty — the human capacity to question, contextualize, intervene, and ultimately own outcomes. AI adoption doesn’t fail loudly. It fails quietly, when people stop challenging outputs that sound right.

Data literacy is the foundation of cognitive sovereignty

Cognitive sovereignty doesn’t begin with prompt engineering. It begins with data literacy.

Forrester defines data literacy as the ability to recognize, analyze, work with, communicate, and apply data in the context of decisions and outcomes. The ACES model — awareness, comprehension, expertise, and scale — provides a practical framework for building that capability across the enterprise.

In an AI‑saturated environment, leaders must build:

  • Awareness: Ask “What went in?” before you trust what came out.
  • Comprehension: Pressure-test logic with quick checks — look for missing context.
  • Expertise: Know how to dispute the output, fix the inputs, and leave a trail someone else can audit.
  • Scale: Make these moves business as usual for everyone.

Without these capabilities, AI doesn’t augment judgment — it quietly replaces it.

Humans in the loop defend judgment when AI breaks down

Data literacy alone isn’t enough. Organizations also need intentional human roles embedded within AI workflows — what Forrester calls humans in the loop (HITL).

In practice, humans in the loop don’t rubber-stamp outputs; they catch category mistakes, name what the model can’t see, and hit the pause button before automation becomes an incident. This matters most when:

  • The output sounds confident — but it collapses under one good follow-up question.
  • The recommendation optimizes a metric while breaking a rule, norm, or promise.
  • The workflow is fast enough to outrun accountability.

My colleague Fiona Mark’s work on AI’s impact on jobs reinforces this reality: While AI may automate large portions of tasks, human judgment, exception handling, and accountability do not disappear — they become more critical. Organizations that fail to invest in these capabilities create systems that are efficient but brittle.

Data literacy is how people become effective humans in the loop

People don’t magically become good HITL once AI is deployed. They grow into that role by developing the confidence, curiosity, and competence to challenge both data and machines. Cognitive sovereignty emerges when organizations expect humans to question AI — not defer to it.

Ask your team this month:

  • Who has the job of saying “Hold on — show me how we got that” (and when do they step in)?
  • Where do humans have to intervene — and where are they already checking out?
  • Can people explain the why behind a recommendation, or are we just copy/pasting the what?
  • Are we scaling independent thinking — or scaling “Sure, whatever the bot said”?

NOW WHAT?

If your AI strategy assumes that humans will “just keep up,” it’s already fragile.

Schedule an inquiry with me (inquiry@forrester.com) to explore how humans in the loop and data literacy programs grounded in the ACES model can help your organization protect judgment, agency, and trust in an AI‑driven world.