In a recent BCG analysis, this group of AI leaders delivers 3 times greater cost reduction, 1.6 times higher EBIT margins, and 2.7 times greater return on invested capital than their peers. They’re also creating other advantages, such as increasing transparency, enabling faster decisions, and reallocating capital more effectively to fuel growth and innovation.

In that way, these companies are compounding their cost advantages from AI and improving overall performance. They show what’s possible and offer insights into how others can catch up.

Challenges to overcome

We see some common challenges across cost programs built around AI:

• Too many fragmented initiatives, not enough scale. Many companies run AI experiments everywhere and lack clear priorities. They dilute their efforts and apply AI to areas where it might not have the biggest impact.
• Foundational issues with data and technology. Successful AI pilots can be tough to scale. Organizations often lack the right IT or data infrastructure, and the testing and resiliency requirements for a company-wide implementation are far more complex than those for an isolated initiative.
• Insufficient focus on training and upskilling talent. Employees sometimes ignore a new AI initiative, often because they lack the skills and capabilities required to use the new tools.
• Failure to redesign workflows and processes. In a typical AI implementation, only 10% of the value comes from the algorithms, and 20% comes from the technology and data. The remaining 70% comes from managing process change—mainly from redesigning workstreams and processes end-to-end.
• Inability to turn efficiency gains into financial value. Even when organizations improve efficiency with AI, those gains often evaporate before they impact the P&L.

The four-part plan for success

To overcome these challenges, leading companies focus on integrating AI into a deliberate sequence of traditional cost levers. Their goal is to deliver results immediately and systematically, through four key priorities.

Start with proven applications to fund the journey. Rather than racing to embed AI across every business unit and function, companies should start with a small number of projects using relatively mature solutions that deliver rapid results.

Procurement is a good option. It often represents a big share of company spending, the
transactions are relatively straightforward, the range of potential issues is small, and proven AI solutions are already available to improve performance. For example, when companies use AI to optimize their supply base, standardize pricing, and negotiate for discounts, they can often save 5% to 25% in three to six months.

Other areas where AI applications can generate fast results include marketing analytics, software engineering, customer service centers, product development, finance,  and field support for sales teams.

Reinvent workflows and processes for greater impact. AI can be applied to existing
processes, but the real value comes from optimizing and redesigning workstreams. The goal is to integrate data flows across departments and functions, leveraging digital and AI technologies to dramatically increase efficiency. This is a bigger endeavor, and one where companies sometimes underestimate the difficulty, especially in redesigning processes that cross functional boundaries.

Because of that complexity, a smart approach is to start with one process and design it from scratch, end-to-end, across the entire value chain. That puts companies on the path to generating breakthrough gains in productivity, efficiency, and value creation.

Apply agentic AI in the right situations. AI agents are systems that can observe, plan, and act autonomously, rather than providing insights. That can enable major cost reductions, especially in functions like HR, finance, customer service, and IT. But it’s important to use agents in the right ways.

For very straightforward processes, baseline automation solutions are good enough. In areas with stiff regulatory requirements, human oversight is a must. The sweet spot for agentic AI applications is in the middle: complex processes and environments where risk exposure and ethical or governance sensitivity are comparatively low.

Rigorously track value. Perhaps the most important step is to link AI-related efficiencies to bottom-line impact in the P&L. That entails building a clear business plan with specific metrics, timelines, and projected ROI. Moreover, teams need to make strategic decisions for how freed-up staff time can be reallocated.

For example, if AI improves the efficiency of a specific activity by 15%, teams supporting that work could either operate with leaner staffing levels or redirect additional capacity toward other value-creating activities.  Managers may even decide to give employees that time back to improve morale. Regardless of how it’s handled, senior leadership teams need to think through these implications.

By applying these four measures, companies can integrate AI with their cost efforts and build a lasting competitive advantage.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.