AI awareness and access have skyrocketed, yet real enterprise value and ROI remain rare. How will you hardwire new behaviours and redesign workflows to make AI‑enabled work the norm?

Key takeaways

• AI investment is high, but value lags because 93% of budgets go to tech, leaving only 7% for the behaviour and capabilities needed to drive real value.
• Embedding AI into redesigned workflows drives real, scalable ROI. Fund the system of work so your AI spend doesn't become shelfware.
• People‑focused adoption works: our controlled behaviour pilot with CIBC achieved 90% developer adoption and a 10–14% productivity lift by embedding AI into workflows and enabling leader‑led behaviour change.
• Supporting new AI habits can boost productivity lift to over 20%, increase experimentation by up to 28%, and enhance openness to embrace future AI developments by 18%.  

AI tools are now universal across organizations. Developer copilots, knowledge assistants, and autonomous agents are being approved and rolled out at record speed. Yet for most organizations, measurable productivity gains and ROI remain elusive.

In our survey of 1,854 executives, 85% of organizations increased their AI investment in the past 12 months, and 91% plan to increase it again this year. However, only 6% of respondents said their increased AI investment saw payback within a year.¹ Another survey of over 3,000 leaders found that, globally, fewer than 60% of employees with access to AI regularly.²

What’s holding back AI value realization within organizations?

They continue to pour most of their AI spend into technology instead of redesigning how work gets done. It’s a bit like investing in gym equipment without designing the proper fitness programs. The result: deployments stall, experimentation fragments, and ROI remains low or difficult to quantify.

To deliver enterprise-scale value, organizations cannot simply bolt AI onto legacy processes. Instead, they need to hardwire a new system of work through AI-fluent behaviours, reimagined workflows, and leader-led AI fluency.

That’s how AI moves from “faster tasks” to better decisions, improved customer experiences, and scalable growth through joint accountability across the organization.

With the right approach, you can shift AI from an optional productivity tool to measurable enterprise value.

High access, low uptake: The gap that’s limiting ROI

Just 10% of surveyed organizations say they are realizing significant ROI from agentic AI.³ It’s a predictable outcome when 93% of AI budgets go to technology and only 7% go toward the people and workflows expected to drive value.⁴

Notable examples of this gap between technology enthusiasm and on-the-ground usage:

• 76% of executives believe employees are enthusiastic, but only 31% of individual contributors agree.⁵
• While there is a growing excitement around autonomous AI agents’ potential to unlock productivity gains and allow human employees to focus on strategy and innovation, only 11% of organizations have successfully deployed agents in production.⁶

Why AI access (or investment) isn’t translating to ROI

1. Adoption is treated as training, not behaviour change

Organizations are directing most of their limited people investment into AI tool training, while underinvesting in what actually drives durable adoption: behaviour change, reinforced through clear expectations, social norms, and incentives. As a result, employees fall into the “knowing versus doing” gap: understanding what AI can do, but failing to use it consistently.

When systems, processes, and leaders don’t reset defaults or reinforce new ways of working, the burden of change rests on the individual. That makes day-to-day behaviour change feel effortful and optional. To create real impact, organizations must design an environment where AI use is visible and rewarded, leaders model and expect AI-enabled execution, and new norms for how work gets done are explicit and consistently reinforced.

2. AI is deployed, but not embedded in how work gets done

Most organizations are lagging in one of two ways:

1. they bolt AI onto inefficient workflows; or
2. they provide broad access to AI tools without clear guidance on when, where, and how AI is expected to be used.

In both cases, employees must constantly decide whether AI fits into their work. When AI is optional rather than embedded in workflows, employees must constantly decide when to engage it. That added discretion creates friction, reduces adoption, and limits value to isolated individual use instead of turning it into a scalable, enterprise-wide capability.

3. Leaders aren’t equipped to lead AI‑enabled work

Middle managers are where strategy turns into day-to-day work. In many organizations, they aren’t clearly set up or held accountable to make AI part of how work actually gets done. When that happens, AI adoption stalls not because employees push back, but because expectations are unclear.

Our internal research shows a simple pattern: people use AI more when leaders are clear and consistent about why it matters. In practice, employees take their cues from what leaders do and reward in real work—not from messages or presentations.

That makes leadership behaviour operational, not symbolic. When leaders use AI in everyday decisions, build it into work requests, and talk openly about how time saved should be used, they make it clear what “good” looks like. This is how AI moves from something people try occasionally to a standard way of working—and how small productivity gains add up to real, enterprise-level impact.

Why people-focused AI adoption is the unlock

AI creates value when it reshapes the workday—liberating human effort from routine tasks, so employees experience a real shift in how they use their time and talents, unlocking opportunities to contribute in ways that were never possible before. That requires deliberate shifts in workflows, roles, and operating models, not just new tools. Yet 84% of organizations haven’t redesigned roles for a human + AI future.⁷

Real-world example: CIBC boosts developer productivity

We collaborated with CIBC to pilot and scale Microsoft’s GitHub Copilot across 1,800+ developers. The result: a 10–14% productivity lift and 90% adoption.⁸

Productivity was measured year-over-year using quantitative data including the quantity and speed of pull requests and pull-request lead time.

This result was not driven by the tool alone. It was driven by how GitHub Copilot was embedded into the system of work:

• Repetitive coding work was AI-augmented by default.
• Leaders actively modeled AI usage.
• Enablement was targeted, not generic.
• Peer learning reinforced adoption.
• Governance and responsible AI guardrails were built in from the start.

By integrating GitHub Copilot directly into developer workflows, CIBC avoided the common trap of isolated experimentation and unlocked repeatable, scalable value.

Key insights from CIBC project

• Behaviour change beats training. Long-term ROI comes from resetting expectations and defaults, not from one-time enablement.
• Leadership modelling is the missing multiplier. When leaders use AI visibly and consistently, adoption becomes normal rather than optional.
• Targeted nudges, personalized adoption journeys, and social proof turn trial into habit.
• Instrumentation matters. Productivity gains must be measured in operational terms (cycle time, throughput, quality) not self-reported sentiment or time savings.

How Deloitte helps turn lagging AI adoption to leading AI advantage

The tipping point to AI ROI is your organization’s ambition to change how work gets done.

We help organizations:

• move beyond pilots and fragmented experimentation;
• start in the highest potential business domains, measure impact, and scale with confidence; and
• align technology, people, and process to deliver measurable, repeatable ROI.

Ready to turn AI investment into sustained business impact?

Connect with our AI and transformation leaders to redesign work, accelerate adoption, and unlock AI advantage.