Beyond chat: Using AI to unlock workflows that were previously impossible

Mindset matters. Achieving value through agentic AI requires leaders and organizations to rethink the workflows of how work happens.

In brief

• AI’s true power lies in linking data, tools, and workflows to automate complex tasks, not just chat-based Q&A, enabling end-to-end business impact.
• Most AI use remains basic and siloed; scaling requires redesigning workflows, roles, and governance to unlock enterprise-wide transformational value.
• EY’s agentic AI framework integrates people, processes, data, and tech for safe, scalable AI adoption, driving productivity, risk reduction, and real ROI.

Most companies still think of AI as a chat interface: ask a question, get an answer. That’s useful, but it’s not transformational. The real unlock comes when AI is connected to your tools, systems and workflows — not when you simply chat with it.

When data, models, infrastructure and business processes are linked, agents can complete steps end to end, apply guardrails, trigger actions across systems and even perform tasks that humans simply don’t have the capacity to do. For example, scanning thousands of documents, consolidating information for executive reports in minutes, or reviewing an entire code base to identify vulnerabilities. 

While most companies know they need AI, few understand where to deploy it. And the answer isn’t another pilot. Realizing agentic AI’s full value takes a broad-based approach, one that links up business, technology, data and people. 

Human-led, agent-operated businesses that align AI to clear business objectives and metrics, scale impact across functions and do so responsibly are best positioned to tap the opportunities of a true agentic enterprise.

Speed bumps at the intersection of AI tools and agentic transformation

There’s no denying AI has already changed the world. Even so, many organizations find themselves investing in the tools without generating proportional returns. The market is full of pilots and tools, but short on business transformation. 

In the working world, AI use cases remain stubbornly basic: search, summarize, write, refine. 

Workforce adoption helps explain why progress so often stalls. Nearly 9 in 10 employees now use AI at work, yet only 28% of organizations are positioned to translate that usage into transformational business outcomes. 

Employees may be saving incremental time, but in most cases, roles, workflows and decision rights remain unchanged. Without redesigning how workflows — who does what, when agents step in and how humans supervise outcomes — AI adoption stalls at small productivity gains instead of driving enterprise-wide value.

That gap shows up clearly in how AI is being used day to day.  New EY research shows some 51% of employees are using AI to search for information. More than one-third say they lean on AI to summarize documents or draft emails. Meanwhile, far fewer are drawing AI into more complex tasks. For example, just 22% use AI to conduct deep research; coaching, mentoring and decision evaluations rank even lower. 

This use-case-led approach to AI can generate some productivity gains. Still, its impact is siloed and generally disconnected across business functions. Ad hoc or bolt-on AI adoption simply layers tools on top of existing processes and hopes for the best. Without a broader strategy or framework to build and scale AI use, clear value tends to plateau. 

What does that look from day to day?

• Use case obsession and siloed pilots spark incremental gains, but no compounding value.
• Legacy frameworks and tools sprawl, complicating governance, creating duplication and leading to fragile integrations.
• Initial productivity boosts taper off, and benefits fail to build momentum across business lines.

Embracing AI and evolving into an agentic enterprise is a generational opportunity for businesses and organizations looking to thrive in a complicated environment. Getting there will take a meaningful shift in perspective that extends beyond a single use case and repositions AI adoption as an integrated and connected way of operating.

Building an agentic AI framework for enterprise advantage

When we talk about the agentic enterprise, what do we really mean? 

An agentic enterprise is a hybrid model where humans set direction and agents operate workflows — following rules and standard operating procedures, escalating exceptions, documenting actions and completing steps across systems with precision and speed.

Pioneering organizations don’t get there by experimenting with chat interfaces. They get there by connecting models to business logic, tools and data, and layering the guardrails that make autonomy safe from the beginning.

In Canada and around the world, EY teams connect business experience with Microsoft platforms to create high-impact, tech-enabled solutions. We bridge our cross-sector insights with Microsoft’s tools to help organizations evolve towards an agentic reality. 

Microsoft research shows 81% of more than 30,000 business leaders surveyed expect agents to be moderately or extensively integrated into their company’s AI strategy in the next 12 to 18 months. Those leading the charge are progressing on the journey to the agentic frontier by taking a three-pillared approach:

Phase 1: Human with assistant

Every employee has an AI assistant that helps them work better and faster. Think Microsoft Copilot-style augmentation.

Phase 2: Human-agent teams

Agents join teams as “digital colleagues,” taking on specific tasks at human direction. At this stage, specialized agents join teams and orchestrate tasks across workflows.

Phase 3: Human led, agent operated

Humans set the direction while agents execute business processes and workflows, with humans checking in as needed. Leaders manage agent portfolios against the business’s key performance indicators (KPIs).

We know people are excited about that prospect. In fact, the EY Responsible AI Pulse survey found that AI is already paying off: 8 in 10 report efficiency and productivity gains. Still, many struggle to scale solutions more broadly — and that’s where AI’s full potential lies.

Market sentiment and enterprise value often rise together. But they’re not the same thing. Financial optimism and investment momentum can accelerate experimentation but aren’t a proxy for sustained business outcomes. 

Market enthusiasm reflects how quickly organizations are investing and piloting AI. Enterprise ROI, by contrast, shows up later in cycle-time reduction, decision quality, risk mitigation and durable performance gains. Confusing the two is one reason so many organizations feel busy with AI yet struggle to point to material results.

That’s where the need for fresh thinking comes into play. Like any change, using agentic AI to transform the way an enterprise carries out processes, makes decisions and serves markets requires a big-picture view. As organizations move in this direction and adopt AI at scale, leaders must reconsider how people, processes, data and technology interact across the business.

That insight is fundamental to creating value as an agentic enterprise. At EY, we call this stage zero – the first step in defining strategic priorities to ultimately enable innovative solutions. Every success on the path to becoming an agentic enterprise starts right there. 

Join up people, processes, data and tech to unleash AI value

We’ve designed, tested and deployed an AI Strategy Framework to help organizations implement AI and realize value in becoming agentic enterprises. We help businesses move beyond siloed use cases and pilots to create a clear and connected AI roadmap that gets to the heart of necessary change and enables sustainable progress. 

In your work, the same patterns emerge:

• Value comes from connecting systems and workflows.
• Agents thrive in structured, rule‑based, high‑volume contexts.
• Successful teams adopt, experiment and stabilize first and then scale cycles.
• Guardrails are required early.

At EY, we start by evaluating existing strategies against our five-pillar framework.

1. Strategy and value realization  

First, we establish a clear North Star vision for AI transformation, including must-win use cases, governance sessions, roles and success metrics. This helps us align AI adoption with financial goals and transformation objectives from the get-go. 

At this stage, organizations will want to ask:

• What enterprise objectives are we looking to achieve and how can AI enable us?
• What enterprise objectives are we trying to move in 6 to 12 months?
• Where are the current workflow bottlenecks (handoffs, rework, waiting time)?
• What must be true for the first agent‑led workflow to scale?

2. Platform architecture and technology strategy

By assessing existing capabilities across an organization’s platforms, we spot reusable components, enabling the design and development of scalable AI tools. This goes beyond cookie-cutter adoption, instead helping businesses use exactly what they need in precise ways to achieve maximum impact.

At this stage, organizations will want to ask:

• What systems can we reuse rather than rebuild?
• What integrations are required for agents to trigger actions?
• Where do we need an orchestration layer so agents can complete multi‑step flows safely?

3. Data foundations

Understanding where you’re starting from is crucial to setting clear future-state goals. We analyze your organization’s existing data strategy and ownership, data infrastructure, data quality and access and controls aligned with AI strategy.

At this stage, organizations will want to ask:

• Which data sources drive the target workflow?
• What quality issues or access gaps introduce risk?
• What monitoring and lineage controls are needed?

4. Workforce transformation

Agentic AI changes the shape of work itself. When implemented well, these capabilities are not introduced as standalone tools, but integrated directly into the systems people already use every day. The shift isn’t about doing the same work faster, but about changing how work gets done. 

For example, specific of assembling materials manually or coordinating multiple meetings to synthesize perspectives, individuals can spin up specialized agents — each equipped with the relevant context, data and expertise — to act as thinking partners. This helps speed up analysis, challenge assumptions and surface insights in parallel. Adapting in this way fundamentally reshapes roles, moving people away from coordination and assembly towards judgment, direction and decision‑making.

EY research carried out jointly with the Saïd Business School at the University of Oxford shows 96% of transformation programs experience challenges that generate a turning point when progress has gone or will go off track. Prioritizing the human element at those turning points can improve transformation performance by up to 12 times. That rings particularly true in the agentic context. 

We look at the high-level structure, role and skill impacts associated with must-win strategic priorities and define a workforce planning and skill development roadmap that prioritizes people. Learning models must evolve alongside agentic workflows. Only 12% of employees receive sustained AI training, yet those who do unlock more than four times the productivity gains of lightly trained peers. 

This gap reflects a broader challenge: episodic, one‑off training cannot keep pace with continuously evolving AI systems. In an agentic enterprise, workforce transformation is not about teaching people to use new tools — it's about preparing them to operate, govern and continuously improve AI‑driven workflows at scale.

This shift also changes how learning happens. In an agentic model, capability development is no longer driven primarily through episodic training programs. Instead, it’s embedded directly into the flow of work.

As employees interact with agents inside their productivity tools, the system applies using organizational leading practices through contextual prompts and real‑time guidance that encourages deeper thinking and better decisions. Learning becomes continuous, adaptive and role specific — not something people step away from work to do, but something that happens as the work itself evolves.

At this stage, organizations will want to ask:

• What does the “day in the life with agents” look like?
• What roles move from doing to supervising to designing?
• What new skills are needed (prompting, exception handling, validation, oversight)?

5. Responsible AI and risk

We help AI initiatives adhere to both ethical and regulatory requirements by applying our proprietary EY.ai Confidence Index and Responsible AI Framework. Using these guidelines when designing AI systems, deployment and monitoring helps safeguard ethical and risk considerations by design.

At this stage, organizations will want to ask:

• What compliance and risk rules govern each workflow?
• Where must humans remain in the loop?
• How do we achieve transparency in agent decisions?

EY: A client-zero story to bolster long-term results

True ROI emerges when workflows are redesigned around agents and success is measured differently. As agents are embedded into end‑to‑end processes, value shifts from tracking tool usage to managing outcomes — cycle time, quality, throughput and risk reduction. Leaders move from overseeing individual productivity to managing portfolios of agents against clear business metrics. This is the difference between adopting AI and operating as an agentic enterprise.