Knocking down four key barriers to achieve agentic success

Scale agentic AI by fixing governance, skills, data and use-case focus—and moving faster from pilots to execution.

In brief

• Agentic AI stalls when leaders lack confidence in governance, risk and compliance.
• People-first adoption and upskilling help close skills gaps and turn agentic AI into measurable outcomes.
• Centralize AI governance with a hub-and-spoke model to reduce duplication and scale safely.

Evolving into a truly unified agentic AI enterprise can help Canadian organizations achieve scalable results and greater ROI. But leaders tend to get stuck on four key challenges that stall agentic progress and hold their businesses back. To overcome those hurdles, organizations must change strategies and accelerate the process from innovation to execution.

Across Canada, organizations are piloting and implementing point solutions for targeted AI use cases. These specialized tools address one specific, isolated problem or business need. That said, these pilots or standalone implementations tend to create silos within organizations. Leaders then wrestle with the question of how and when to centralize implementation.

• Prioritize people as you become an agentic enterprise.
  
  At EY, our Work Reimagined 2025 study revealed nearly 9 out of 10 employees now use AI at work, but only 28% of organizations are positioned to turn AI deployment into high-value outcomes.
  
  What separates the 28% from everyone else? Of the 1,500 organizations surveyed, those achieving transformational results had a stronger approach to people-centric factors: recruiting and retaining talent, driving AI adoption at scale, building continuous learning into daily operations, reshaping culture and workplace norms, and aligning rewards with new behaviours and outcomes.
  
  That’s something we’ve long believed to be true. In Canada, we built our EY AI Academy to close those gaps. Our unique delivery approach embeds AI learning right in the flow of work. People can put upskilling to good use immediately as they master new capabilities. This helps entire workforces scale AI use, shift behaviours sustainably and push towards agentic enterprise status.
  
  This way of working and learning also embeds human-centred thinking within the culture itself — making it only natural to focus on people when implementing any kind of tech transformation.
  
  Research carried out jointly by EY and the Saïd Business School at Oxford University found that 96% of transformation programs experience challenges that generate a turning point when the initiative has gone, or will go, off track. Placing humans at the centre when navigating those crucial moments can double the likelihood that a transformation overperforms its KPIs. In fact, the human element plays a crucial role in successfully manoeuvring these turning points and increases the chance of a transformation significantly, improving performance by up to 12 times.
  
  We apply these principles in practice with Microsoft every day. By pairing leading technology with a deep understanding of people advisory and change management techniques, we can help organizations bridge skill, capability and confidence gaps and reinforce the business foundations that make agentic AI so powerful over the long term.

• Pick the right time to centralize governance with a hub-and-spoke model.
  
  Like most innovation, organizations can often cross a tipping point where decentralized innovation can lead to diminishing returns. In AI practice, that might mean in the first wave, where teams are using a tool like Microsoft Copilot, no governance is required. Then, as agents and humans work together to do tasks, governance at the centre becomes essential. That carries into the next wave, as agents take on entire workflows, as well. Embracing a hub-and-spoke model from the beginning can help manage standards, provisioning, enablement training and more.
  
  We’ve worked with many financial institutions and other organizations that have struggled with that tipping point, unsure of when to centralize platforms. How do you know when it’s time? An effective current-state analysis identifies where innovation is happening, and how duplication is occurring. For example, a bank might develop 10 distinct agents, each doing a remarkable job in a different business unit — only to find at the analysis stage that this collective innovation could be centralized with a single “vanilla” agent ready to achieve all of those different capabilities at once. Shifting in this direction then frees up teams to focus differently and innovate strategically around common goals and with greater clarity across business units.
  
  Striking the balance between a culture of experimentation, where pilots and development occur often, creates possibilities. Strategically determining how to elevate that dispersed approach through a centralized hub-and-spoke model takes progress further, unleashing the potential that a fully agentic enterprise can achieve.
  
  This balance point enables organizations to pick the right use cases, integrate high-quality data at the right time and confidently manage governance, risk and compliance at the centre of the process. From there, scalability becomes feasible without risking a slowdown for the lines of business.
  
  
• Streamline and speed up the process from ideation to execution.
  
  As you centralize innovation with humans at the centre, the business will need a coherent framework to generate long-term success. At EY, we’ve condensed that approach into a 10-step process that links ideating, evaluating and prioritizing, funding, building and ultimately going to market. We’ve built key accelerators that make this process possible in days, not months. This means you can move strategically and quickly.

What’s the bottom line?

Enterprise AI has reached a crucial turning point. Leaders worry about "silent failure," where companies spend on AI solutions that seem to work but don't provide real value. This happens when organizations focus on standalone successes without connecting the overall approach.

As AI tools proliferate, risks increase, leading to problems like wasted efforts, security issues and maintenance challenges. To overcome these barriers, evolve your strategy and move more quickly from innovation to execution. Doing so can help turn AI possibility into sustainable AI value.

Summary

Agentic AI can deliver enterprise value, but only if organizations address governance, skills, data integration and use-case focus. Put people at the centre, centralize governance at the right time, and speed the path from ideation to execution to avoid “silent failure” and scale ROI.