Failure to scale AI in the workplace isn’t a technical issue, it’s a cultural one 

AI adoption continues to pick up pace, with a new announcement about larger models, better benchmarks and expanded reasoning capabilities seemingly being released every day. Gallup, in fact, indicated that AI use by employed professionals has spiked by 46% since Q4 2025. 

However, the consulting firm also found that the rate of adoption has not been consistent across all roles, with employees outside of management roles in particular less likely to use the technology. Many leaders, as per the report, were discouraged from using AI tools due to concerns about its usefulness and how it would impact established work habits. 

While leaders may be keen to push forward with AI initiatives, the data on the ground highlights a growing disconnect between decision makers and broader employee bases.

The WRITER’s 2025 enterprise AI adoption report, for instance, also recently highlighted growing friction associated with AI adoption: of the 1,600 knowledge workers actively using AI in the workplace who were surveyed, 68% of C-suite executives reported that this rapid integration has sparked division within their organizations, with 42% stating it’s tearing their companies apart. 

This disconnect could undermine the progress of AI in the workplace and limit the use of tools that have been invested in by leadership. Even worse, it could put company cohesion at risk. 

While AI models grow increasingly sophisticated by the day, these systems are not failing because they lack computational power; they are increasingly likely to fail because they don’t communicate in a way that reflects our reality. 

That distinction matters more than many industry insiders and leaders are willing to admit. 

Why nuance causes AI initiatives to fail 

The industry narrative suggests that once AI becomes sufficiently intelligent, it will naturally become more useful and trusted.

However, in practice many organizations are discovering something very different, with Gartner research saying that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk management. 

The challenge is no longer simply whether a model can generate language, but whether it can communicate in ways users perceive as contextually appropriate.

This gap between capability and communication is emerging as one of the most important risks in enterprise AI deployment, and the growing disconnect isn’t because AI lacks intelligence or the technology is fundamentally flawed. Instead, natural human interactions typically consider a range of subconscious contextual indicators that allow for adjustment in approaches based on cultural differences or situational nuance. 

When considering these differences in the context of real-world AI applications, why suboptimal AI communications undermine the success of adoption in the workplace comes to light. 

Current performance frameworks used to measure the performance of AI often disregard the cultural element: an AI tool built to handle customer queries will be measured by its ability to produce correct responses consistently as its KPI, for example. 

However, this same tool could result in negative customer sentiment if the tone feels dismissive, overly formal, culturally unaware, or emotionally disconnected.

According to Sundial Media & Technology Group CEO Kirk McDonald, the tension is palpable, and this disconnect is far from a theoretical talking point. In a recent interview, he explained the crux of the issue at hand: AI is adept at scaling content, but it can’t replicate cultural credibility. 

It’s not just customer relations and brand authenticity at risk. AI miscommunication also poses serious problems for employee engagement, which is already being tested by the rapid adoption of AI in the workplace. 

If these systems fail to support their daily activities, it only adds insult to injury.  These are not edge cases, but rather systemic issues that become more visible as organizations move AI into customer-facing and operationally sensitive environments.

Making the case for bespoke cultural AI 

Solving the disconnect between AI and the way that people actually interact demands a closer look at how models are actually being built and the metrics that are rewarded. 

People interpret meaning through social context and lived experience. The same phrase can carry entirely different implications depending on the situational context, and humans navigate these variables instinctively because communication is deeply social.

AI systems, however, are often optimized primarily for generalized output generation.

That works reasonably well in broad applications. Once systems are deployed into environments where trust matters, however, generalized communication becomes insufficient.

Traditional AI-testing environments focus heavily on performance metrics such as accuracy and hallucination rates. Those measurements are valuable, but they rarely capture whether systems behave appropriately across different cultural or operational contexts.

To address this, cultural and contextual alignment must become part of the system architecture itself. 

An emerging term known as “Bespoke Cultural AI” is the deliberate design of AI systems that reflects how specific organizations actually interpret information. It recognizes that reliability in AI is not just about factual correctness, but also includes whether users consistently interpret outputs as intended. 

How to achieve bespoke cultural AI at scale 

When it comes to bespoke cultural AI, companies need to first evaluate whether their AI systems understand the communication environments in which they operate: a healthcare interaction differs fundamentally from a retail interaction, or enterprise internal workflows differ from consumer-facing support, for example. 

Context can therefore not be treated as secondary. Organizations must recognize that trust is culturally constructed; users evaluate credibility differently depending on industry norms. AI systems that ignore those dynamics will struggle to gain adoption regardless of technical sophistication.

Third, businesses should rethink how they test AI reliability. Evaluation cannot focus solely on technical outputs. It must also include human interpretation.

And finally, governance structures that account for operational communication risk are equally critical. Many organizations already assess cybersecurity, privacy, and compliance risks before deployment. 

AI communication failures deserve similar attention because they directly affect customer experience, employee productivity, and brand trust.

Culture is not a dataset 

In today’s marketplace, AI adoption remains focused on metrics like acceleration and outputs. However, in the real world, the success of AI hinges on AI that interacts through culture, nuance and relationships.

AI systems that fail to account for those realities will continue to encounter resistance, no matter how technically sophisticated the model is. It’s in every decision leader’s best interest to take steps to address this.

Article authored by Larry Adams, CEO of Chromatics AI