AI is moving quickly, but not all progress is positive. Deloitte notes that 2025 may be a transition year for artificial intelligence because of three factors: rising energy consumption, gender disparities in AI roles, and slow enterprise adoption of autonomous agents.
So where are the gaps, and what do they mean for businesses considering AI adoption this year?
Why Is AI Creating a Sustainability Problem?
Training and running large AI models consumes vast amounts of energy. Research from the University of Massachusetts found that training a single large language model can emit the same carbon as five cars over their entire lifetimes. As models scale further in 2025, energy requirements continue to rise.
Key points:
AI training requires extensive data centre use and cooling infrastructure.
Data centres already consume 2 to 3 percent of global electricity (IEA 2024).
Governments are beginning to question the long-term environmental cost of unfettered AI growth.
How SMEs should think about this:
Look for cloud providers with published carbon-reduction commitments.
Use AI for targeted workflows instead of blanket adoption.
Audit usage regularly to avoid unnecessary compute costs and wasted energy.
Related reading: What’s the Difference Between a Cloud Backup and a Cloud Sync — and Why It Matters
What Is the Diversity Gap in AI?
Despite years of awareness, women and minority groups remain underrepresented in AI roles. The World Economic Forum’s 2024 report showed that only 22 percent of AI professionals worldwide are women. Leadership roles skew even further.
Why this matters:
Homogeneous development teams increase the risk of biased systems.
Lack of representation impacts the fairness of AI tools in recruitment, finance and legal services.
Diverse teams have been shown to build safer and more accurate AI models.
How businesses can address it:
Audit the AI tools you use for bias testing.
Push suppliers to disclose diversity data.
Invest in training programmes for underrepresented staff to grow internal capability.
Related reading: AI Tools Your SME Can Actually Use (Without Breaking the Budget)
Why Are Enterprises Slow to Adopt Autonomous AI Agents?
Autonomous agents, software systems that can make decisions and take actions without direct human oversight, are still slow to roll out across enterprises. Deloitte’s 2025 outlook suggests adoption is being held back by:
Trust issues: Companies are wary of reputational and regulatory risks.
Integration challenges: Many agents require complex setup across existing systems.
Lack of governance: Few firms have a framework for accountability if agents make mistakes.
What this means for SMEs:
It is safer to focus on assistive AI (tools that support staff, like Copilot or Duet AI) rather than autonomous systems.
SMEs can adopt low-risk workflows such as invoice processing, email triage, or forecasting without handing over full control.
Related reading: Why IT Shouldn’t Be an Afterthought: Building Tech into Your Growth Plans
How Can Businesses Balance AI Growth With Responsibility?
Practical steps include:
Energy: Prioritise cloud providers with renewable energy commitments and monitor AI usage costs.
Diversity: Push for diverse hiring in your own teams and demand it from your AI suppliers.
Adoption: Start with assistive AI to prove value and minimise risk before expanding.
Summary Table: The Gaps in AI (2025)
Gap
What It Means
Business Action
Sustainability
High energy use and carbon footprint
Choose greener providers and audit AI usage
Diversity
Lack of representation in AI workforce
Audit tools for bias, demand supplier transparency
Adoption
Slow uptake of autonomous agents
Start with assistive AI, apply governance early
Final Word
AI adoption is not only about innovation but also responsibility. 2025 is shaping up as a year where sustainability, diversity and governance questions will shape how AI grows.
SMEs have an advantage here. By starting small, asking the right questions, and building AI use around real workflows, you can benefit without falling into the traps that are already slowing larger enterprises.
