> Research from Harvard Business School (Das Narayandas and Shunyuan Zhang, HBS) confirms it: at least 30% of generative AI projects in companies are abandoned — not because of technical bugs, but because employees reject them. Read the full research on HBS Working Knowledge →

How to Deploy AI in Your Company Without Creating Internal Resistance

Artificial intelligence promises spectacular productivity gains, but the reality on the ground is often more nuanced: many projects fail not because the tool doesn't work, but because teams reject it. That's exactly what Das Narayandas and Shunyuan Zhang, professors at Harvard Business School, demonstrate in research widely covered by HBS Working Knowledge: the main cause of failure for generative AI projects isn't technical — it's human.

Why Employees Resist AI — And It's Not What You Think

The Harvard researchers describe generative AI as a "self-disruptive" technology: unlike previous tools, it doesn't just automate tasks — it directly touches the professional identity of the person using it. Their research identifies three specific forms of perceived threat among employees:

• Role compression: the employee keeps their job, but AI absorbs the most interesting and valuable parts of their daily work.
• Control shift: algorithms reduce managers' decision-making latitude and their sense of authorship over their own choices.
• Span erosion: less influence over people, budgets, and processes — a quiet shrinking of perceived authority.

So it's not the technology itself that blocks adoption — it's the absence of a clear answer to a very concrete question for every employee: what do I become in this new organization?

The HBS Solution: Making AI Identity-Compatible

Based on this finding, Narayandas and Zhang propose five concrete mechanisms to turn resistance into adoption:

1. Rechart roles to highlight newly visible expertise rather than erase it. 2. Build decision guardrails that give employees real oversight power over AI-assisted decisions. 3. Create analytical overlays that reinforce human judgment rather than replace it. 4. Plan redeployment pathways with retraining, for the roles most transformed by automation. 5. Secure clear executive sponsorship of redesigned roles, driven by leadership — not just IT.

The central principle of this research is simple to remember: an AI system doesn't need to be convinced or motivated to work — but the human does need to understand exactly what they stand to gain.

What This Means Concretely for an SMB

These principles, originally designed for large organizations, apply just as much — if not more — to SMBs, where each employee often wears several hats and team trust is an even more fragile asset.

At Busony, we see this directly with our export-focused SMB clients: the AI agent projects that succeed are the ones that start from a precise business pain point (qualifying inbound leads, sales follow-ups on WhatsApp, multilingual customer support) and that explicitly leave the final word on the customer relationship to the sales team — AI prepares, qualifies, and follows up; the salesperson decides, negotiates, and closes.

This clear division of roles — AI on repetitive, time-consuming tasks, humans on judgment and relationships — is what avoids the "black box that replaces you" effect and turns an automation project into a tool teams adopt rather than endure.

Our Approach at Busony

Before deploying an AI agent for a client, we always start with the same question implicitly raised by the HBS research: what does this tool change for the people who will have to use it every day? That step — often overlooked in favor of pure technical performance — is what separates an AI project that sticks from one that fades out after three months of declining usage.

Preparing an AI rollout for your teams and want to avoid the classic adoption pitfalls? Contact Busony to talk it through.