As we’ve all seen in 2026, politics and economics together have triggered a grassroots, anti-AI movement. Business, AI and data leaders working together need to devise a response to that movement.

In this post, I’ll try to describe the nature of the current situation and how to think beyond first impressions to reposition your own organization to be able to respond effectively.

The backlash against generative AI

By relentlessly hawking a vector-only, purely probabilistic, half-baked, all-agentic, no-human-in-the-loop version of generative AI, the dominant gen AI startups and hyperscaling incumbents have unwittingly polluted how the general public conceives of AI in general. For some strange reason, these vendors have placed autonomous-only agents as a higher priority than agents designed and intended to assist first.

It’s been puzzling to hear how the tech sector’s most prominent gen AI promoters couch their arguments in terms of warnings. Senior Editor at Futurism.com’s Victor Tangermann described the situation succinctly in a May 2026 article:

“For years now, tech leaders have warned that AI will usher in a technological revolution on an unprecedented scale, wiping out countless jobs. If you’re lucky enough to survive sweeping layoffs continuously roiling the tech industry, bosses say their employees will have to adopt the tech to keep their jobs — whether they like it or not.”

The unintended consequence has been that large swaths of the general public now assume that all enterprises who are trying to make agentic AI work are anti-human too – not just their tech sector suppliers.

The backlash from the burgeoning political movement against AI and AI data centers is palpable, particularly when it comes to Gen Z. Tangermann cited a recent poll result: “One recent Gallup poll showed that only 18 percent of Gen Zers said they felt ‘hopeful” about AI, a drop of nine percent compared to 2025.”

A PR Imperative: Shifting direction to a human-first approach

From a public relations perspective, enterprises need to make a firm and public commitment to an alternative to mainstream generative AI. Here are two elements of a solution to consider together: Agentic networks and semantic graphs.

Agentic networks with humans at the helm: If we step back from the unfavorable political and economic morass and the bad taste Gen Z has in its mouth for AI in its current incarnation, we can all agree that agents themselves aren’t the problem. Agents led by humans can  collectively create beneficial network effects. Their effectiveness, efficiency, reliability and trustworthiness hinges on the environment they’re interacting with and the inputs they’re receiving from that environment.

As I’ve said before, statistical machine learning has a garbage in/garbage out dynamic to it. Such machine learning needs to be complemented with deterministic facts, rules and conditions of a logically consistent, topically coherent knowledge graph to be trustworthy. Such a graph with an ontological, standards-based backbone in one department can be a source and a model of coherence and consistency for other departments to follow.

Humans are at the helm of the best knowledge graph development. The most capable humans can also harness the power of agents in the right way for the best outputs.

Semantic graphs as a network effects amplifier: Metcalfe’s Law as it stands says the value or influence of a network is proportional to the square of the number of connected nodes (n²). Let’s qualify that law by pointing out that all nodes and connections aren’t created equal. Just adding nodes and connections haphazardly won’t produce the promised network effects.

The more a company’s data becomes contextualized, logically linked, knowledge-enabled and FAIR upstream, the more agents and humans can discover and implement new uses for that data downstream. Optimally, these self-describing graphs need to expose and enable both deterministic and probabilistic capabilities.

How to help agents help us: Add meaning context by context to their input

Agents can interpret and act based on that interpretation. Forget the debate on whether or not agents “understand”. That’s not relevant. Agents don’t need what humans conceive of as the ability to “understand”. They just need to find the right abstractions and mapped specifics in a self-describing graph to do their work.

Today we can build layers of explicit, dynamic, contextualized, interlinked, machine-readable knowledge to help some agents do their work precisely, while other agents are checking after them and humans are overseeing the whole effort. It’s a proven capability in the life sciences, financial services, manufacturing and other knowledge-intensive industries.

That knowledge can be mapped to and live with the instance data (specific observations in datasets) and help make those datasets that on their own are underdescribed and created for a single context findable, accessible, interoperable and reusable (FAIR) in multiple contexts.

That way, agents can travel across boundaries from one context to another and assist humans in their vocational and avocational endeavors, across departments, fields of endeavor, and areas of learning.

In the field of life sciences, various organizations are working on creating FAIR context and enabling the sharing of FAIR datasets so that one scientist is aware of what other scientists are working on.

The goal of FAIR² Data Management, according to Frontiers scientific publishing organization, is “an intelligent way to organize, share, and publish data – making it AI-ready, reusable and impactful.”

Today’s AI is a black box. The effectiveness of AI depends entirely on the quality of the information inputted into that black box. The people behind FAIR data are focused on creating data layer ecosystems that are AI ready.

Here’s how life sciences organizations do this advanced version of knowledge-based data management:

The path to actual AI readiness

Agents can accurately interpret and reason about what they can read if the input is good enough. We can define AI-ready data so that the input can be good enough.

Building human-first AI requires a fundamental shift in how to treat information. AI-ready data isn’t a format; it’s a commitment to design that is findable, accessible, interoperable, and reusable.

 True AI-readiness earns trust through documented provenance, standardized structures, and complete transparency—ensuring that both agents and humans know exactly what they are working with. 

This is not a feature you simply switch on. It is built bit by bit, by tracing every source and acknowledging every bias. 

By prioritizing data that machines can actually read, trust, and use, organizations can move upstream from the black box and build systems that genuinely serve human needs.

For more information:

Graphwise. “AI-Ready Graph Environments: The Key to Scaling AI with Graphwise’s Knowledge Graphs.” Graphwise Blog, November 12, 2025. https://graphwise.ai/blog/ai-ready-graph-environments-the-key-to-scaling-ai-with-graphwises-knowledge-graphs/.

Tangermann, Victor. “Gen Z Is Turning Against AI in an Incredible Way.” Futurism, May 1, 2026. https://futurism.com/artificial-intelligence/gen-z-turning-against-ai.

Thurner, Thomas. “Thomas Thurner on What’s Missing Between Data and Intelligence.”  Graphwise Thought Leadership, originally published in CMO Thinks, June 18, 2026. https://graphwise.ai/thought-leadership/thomas-thurner-on-whats-missing-between-data-and-intelligence/.

Editorial note: Some of the information in this post appeared previously in a Quora answer of mine on a related topic.