Most organizations are now spending significantly on AI, but many are not seeing it show up in the P&L.

What the data shows

• Nearly one-third of companies invested at least 1.7% of revenue in AI last year, and by 2026 that figure is expected to be closer to two-thirds.
• Yet about 60% of companies report minimal or no value from AI in terms of cost reductions or revenue gains.
• Nearly two-thirds say AI scaling expenses are becoming uncontrollable, and a similar share face hallucination and explainability issues.
• Roughly three-fourths cite security concerns and challenges with unstructured data.

Why this happens

• Too many fragmented pilots. Companies run lots of proofs of concept without clear priorities, so efforts are diluted and rarely reach scale.
• Weak data and tech foundations. Pilots may work in isolation, but scaling them requires robust data, architecture, and testing. Without that, organizations rely on manual checks that erode value.
• Limited training and adoption. Employees often lack the skills or incentives to use new AI tools, so they revert to old ways of working.
• No real workflow redesign. AI is layered on top of existing processes instead of reimagining how work should be done. In practice, only about 10% of value comes from algorithms, 20% from tech and data, and roughly 70% from process change.
• Poor linkage to financial outcomes. Efficiency gains are not systematically tracked or translated into budget changes, headcount decisions, or margin improvements, so they never become a true cost advantage.

In contrast, AI leaders treat AI and cost transformation as a single agenda. They embed AI into end-to-end process redesign, focus on a few high-impact use cases, and rigorously track financial impact so that efficiency gains become measurable P&L results.

AI leaders are not just deploying tools; they are rethinking how work gets done and how costs are managed.

Measured impact of AI leaders
BCG analysis shows that AI leaders deliver, on average:

• 3x greater cost reduction
• 1.6x higher EBIT margins
• 2.7x the return on invested capital compared with peers

What they do differently

1. Start with proven, high-impact use cases.
   They avoid spreading resources thin and instead pick a small number of mature, high-value workflows to prove impact and fund the broader journey. Procurement is a common starting point because:
   • Spend is large and transactions are relatively straightforward.
   • Commercial AI solutions already exist and can be deployed quickly.
   Typical outcomes in procurement include:
   • Supplier reviews: 5–25% savings in 3–6 months.
   • Specification reviews: 5–10% savings in 3–6 months.
   • Inventory optimization: 5–15% savings in 3–9 months.
2. Reinvent workflows, not just automate tasks.
   Leaders reimagine end-to-end processes across functions, using digital and AI to integrate steps and remove handoffs. This kind of redesign can generate 3–4x the impact of incremental improvements because it:
   • Eliminates redundant work.
   • Shortens cycle times.
   • Improves transparency and decision speed across the value chain.
3. Apply agentic AI where it fits best.
   They use AI agents in complex, lower-risk environments where autonomous action can materially reduce cost, such as HR, finance, customer service, IT, and engineering. Examples include:
   • Marketing and product development: One global consumer goods company rolled out 10 custom agent workflows to 500+ users, cutting time spent on key workflows by 25–40%, getting products and campaigns to market twice as fast, and achieving 90%+ user satisfaction.
   • Engineering design: A shipbuilder used multi-agent AI to cut design lead times from 5 days to 1 day and reduce engineering costs by 45%, while maintaining or improving accuracy.
   • Software modernization: A bank reduced the time engineers needed to understand legacy code by up to 30%, with projected savings of more than 70% at scale.
4. Link efficiency gains to financial outcomes.
   Leaders put in place governance and tracking so that:
   • Productivity gains are quantified.
   • Budgets and resource plans are adjusted accordingly.
   • Savings are reinvested in further AI and transformation initiatives.

By combining these elements, AI leaders build operating models that not only lower costs today but also create transparency, faster decision making, and capital to reinvest in future growth.

A practical AI-first cost transformation starts small but deliberate, with a clear path from pilots to scaled impact.

1. Pick a few proven, high-value workflows
Begin with areas where:

• Spend is significant.
• Processes are relatively standardized.
• Commercial or well-tested AI solutions already exist.

Common starting points include:

• Procurement (supplier reviews, specification optimization, inventory management).
• Marketing analytics and campaign optimization.
• Software engineering productivity.
• Customer service centers and field support.
• Finance processes (e.g., reconciliations, reporting).

Use these early deployments to generate quick, measurable savings that can fund broader transformation.

2. Invest in data, tech, and talent foundations
To scale beyond pilots, you’ll need:

• Reliable data and architecture so AI solutions can be deployed enterprise-wide without excessive manual checks.
• Training and upskilling so employees know how to use AI tools and understand new expectations.
• Clear governance around security, risk, and explainability, especially given that roughly three-fourths of companies report security and unstructured data challenges.

3. Redesign at least one end-to-end process
Choose a critical process and reimagine it from scratch across the value chain, rather than just automating existing steps. Aim to:

• Integrate activities across functions.
• Remove unnecessary handoffs and manual work.
• Use AI to orchestrate decisions and workflows, not just provide insights.

This kind of redesign can deliver 3–4x the impact of incremental tweaks.

4. Experiment with agentic AI in the right pockets
Identify complex but lower-risk areas where AI agents can observe, plan, and act with human oversight. Examples include:

• Marketing content and concept development.
• Engineering and design support.
• Internal knowledge retrieval for customer service or IT.

Keep humans in the loop where risk, compliance, or ethics require it, and use guardrails to manage quality and safety.

5. Tie everything to measurable financial outcomes
From the outset, define how you will:

• Measure productivity and cost improvements.
• Translate those gains into budget and resource decisions.
• Reinvest a portion of savings into further AI and operating model changes.

By following this staged, disciplined approach, you move from isolated AI experiments to an AI-first cost agenda that reshapes how your organization operates and competes.