agents-for-growth-turning-ai-promise-into-impact

Growth, Marketing & Sales Practice Agents for growth: Turning AI promise into impact As CEOs and CMOs ask where AI is moving from hype to real results, frontrunners demonstrate that tighter human–AI collaboration and sharper governance is required. by Greg Kelly, Lisa Harkness, and Steve Reis November 2025 -- 1 of 7 -- Key takeaways — Value comes from end-to-end change. Broad productivity wins are table stakes. Impact comes from prioritizing the biggest growth problems and then solving them end to end in a domain. — Reimagine workflows, not tools. Growth comes from mapping decisions and handoffs, and embedding agents where they change outcomes, not bolting them onto legacy steps. — Scale with a new operating model. End-to-end transformation requires cross-functional human–AI teams, shared data products, and governance that treats agents like managed talent. A global retailer sees demand for a top product surge in one region while inventory piles up in another. Within seconds, a team of AI agents reallocates ad spend, adjusts pricing, reroutes stock, and refreshes creative assets to offers that match shopper intent. In this scenario, what comes next is coordinated action triggered by customer signals, orchestrating business growth in real time. This is not a fantasy scenario; it’s the new frontier of AI in growth functions. Agentic AI embeds automated reasoning directly into marketing, sales, and customer service workflows. We estimate that agentic AI will power more than 60 percent of the increased value that AI is expected to generate from deployments in marketing and sales.1 It’s no exaggeration to say that marketing and sales represent the tip of the spear when it comes to translating agentic AI’s potential into meaningful value.2 Early leaders are already seeing measurable impact. For example, according to McKinsey analysis, some Fortune 250 companies have estimated that they are seeing campaign creation and execution speed up 15-fold, driven by faster innovation cycles and process optimization. The value from agentic AI comes from the tasks it is able to do. Unlike gen AI and chatbots that largely assist in the completion of marketing and sales tasks, AI agents can act, decide, and collaborate. They are able, for example, to optimize prices, advance leads, tailor offers and manage customer interactions end-to-end. As organizations deepen their adoption of agentic AI, gains can scale. Our analysis shows that effective and scaled agent deployments could deliver productivity improvements of three to five percent annually and potentially lift growth by 10 percent or more. Most organizations, however, have yet to realize meaningful value from AI generally. Nearly eight in ten report no significant bottom-line gains from AI generally, mostly due to constraints stemming from fragmented pilot programs, weak data, and insufficient governance foundations. The leaders breaking through and realizing value from AI are redesigning how growth happens by integrating AI agents into their workflows. From our experience across industries, organizations that are finding breakthroughs and turning agentic AI from promise into performance in marketing and sales are following four lessons: 1 Early applications show gen AI could unlock $2.6 to $4.4 trillion in annual value, with as much as 20 percent of the expected productivity lift concentrated in marketing and sales. 2 Agentic AI is a system based on gen AI foundation models that can act in the real world and execute multistep processes. AI agents can automate and perform complex tasks, often using natural language processing, which would normally require human effort. Agents for growth: Turning AI promise into impact 2 -- 2 of 7 -- 1. Go where the value is Impact begins with identifying where agents can move the needle—whether in conversion, pricing precision, or customer engagement—and deploying them to accelerate those outcomes. Consider personalization, where the opportunity is both proven and profound. McKinsey research shows that 71 percent of consumers expect personalized interactions, and 76 percent become frustrated when they don’t happen. AI-driven personalization can enhance customer satisfaction by 15 to 20 percent, increase revenue by 5 to 8 percent, and reduce the cost to serve by up to 30 percent. Agentic AI makes this possible at scale, using contextual reasoning and real-time decisioning to refine offers, content, and experiences with each interaction. According to McKinsey analysis, a European insurer, for instance, reimagined its sales operation with AI agents that personalized campaigns across hundreds of microsegments, adapted scripts to buyer cues, and coached sales teams with real-time feedback. The result: conversion rates two to three times higher, 25 percent shorter customer service call times, and continuous learning loops that manual reviews could never match. Other organizations are using similar practices with AI to elevate customer experience by anticipating what each customer needs next and delivering it in the right moment. A US airline used predictive insights to tailor compensation for flight disruptions, differentiating between frequent fliers and occasional travelers. The resulting impact was a 210 percent improvement in targeting at-risk customers, an 800 percent rise in customer satisfaction, and a 59 percent reduction in churn among high-value travelers. That same type of intelligence is also proving useful to enhancing pricing. Agentic AI can sense market shifts, model outcomes, and act instantly by adjusting prices or reallocating inventory in real time based on competitor moves, customer behavior, or demand forecasts. Airlines, for instance, are already using agentic AI to create personalized bundles that combine fares, seating, and add-on offers, updating prices dynamically based on live signals such as search trends, weather, and booking patterns. 2. Think in terms of workflows, not agents Organizations realizing meaningful impact from agentic AI are going beyond simply deploying new agents to improve existing tasks; they are redesigning workflows. Agents enhance value creation when used to improve end-to-end processes and journeys through automation and coordination—their power is limited, however, when used to improve isolated steps. Enhancing product discovery, for example, delivers limited impact if purchasing and fulfillment remain slow or disjointed. In traditional processes, work moves sequentially, often across departments: Marketing hands off to sales, service escalates to support, and pricing follows. Each of these functions has made Agents for growth: Turning AI promise into impact 3 -- 3 of 7 -- tremendous progress in recent years by incorporating digital and analytics capabilities, and agentic AI builds on those improvements by automating and orchestrating tasks across teams and functions. Overcoming the persistent challenges of coordination across complex operational silos and workflows can allow organizations to achieve faster cycle times, as well as greater consistency and responsiveness at a scale no level of human coordination could match. Crucially, success calls for designing processes around agents—not bolting agents onto legacy processes. For example, rather than using agents to help customer service teams respond to complaints faster, leading organizations use agents to predict potential issues, trigger outreach before a customer calls, and resolve cases pre-emptively with personalized offers. The European insurer offers a clear view of what this looks like in practice. According to McKinsey analysis, in just 16 weeks, the company re-architected its commercial model around a connected network of agents working across the full customer journey. The improvements generated included the following: — Knowledge agents centralized over 1,000 policy and product documents, enabling frontline staff to retrieve accurate answers instantly. — Coaching agents introduced AI-driven call transcription and grading, automatically reviewing 95 percent of sales calls versus 3 percent previously. — Integration agents connected these capabilities into the existing CRM and agent portal—adhering to single-sign-on security policies and providing real-time performance dashboards. Together, these agentic systems shortened average call times by 25 percent, reduced manual cross-functional handoffs, and created a continuous feedback loop. As agents learned from each engagement, they continually refined next-best actions, message sequencing, and product pairing to stay aligned with evolving customer needs. Value creation with AI agents for end-to-end change depends, however, on matching the right agent to the right task: domain-specific agents that handle complex, contextual actions; generalist agents for tasks such as data synthesis or content generation; agents that check for errors; and orchestration agents that direct and synchronize the system as a whole. Humans have a crucial role in this effort. They can work closely with agents to supervise and verify, as well as manage issues that AI agents escalate to them. The most advanced organizations combine these human–agent collaborations into adaptive workflows that evolve with each iteration and customer signal. Agents for growth: Turning AI promise into impact 4 -- 4 of 7 -- 3. Build collaborative agents, not just add-on tools To scale agentic AI, organizations need to stop thinking of agents as add-on tools and start treating them as collaborative, digital partners. That means defining the agents’ roles, onboarding them properly, and managing them with clear performance expectations—not unlike human team members. The right metrics for measuring AI agents’ performance differ from traditional productivity KPIs, however. Rather than focusing on call counts or campaign volume, for example, leading organizations track a mix of indicators such as conversation quality, task-completion accuracy, escalation precision, and learning velocity, reflecting how effectively agents incorporate feedback and adapt to changing buyer cues. Because every agent action is logged and traceable, these metrics can be monitored continuously. Real-time dashboards surface performance drift, benchmark outcomes against human baselines, and flag when retraining or recalibration is needed. A leading US homebuilder demonstrates how this discipline translates into impact. Seeking to improve digital engagement and appointment conversion, the company trained AI sales agents to emulate its top-performing human sellers. McKinsey analysis of more than 500,000 sales transcripts revealed dozens of conversation states—greeting, objection handling, follow-up, close—and the patterns most associated with success. Using these insights, the team developed agent personas with unique styles, tempos, and conversational approaches. Every AI-led conversation was then benchmarked against human baselines using a scoring agent that evaluated accuracy, personalization, and flow. Dashboards highlighted drop-off points and tone mismatches, enabling rapid tuning. Conversion-to-appointment rates tripled, weekly appointments doubled, and the best-performing agents reached human-level parity in empathy and flow. 4. Build the agentic growth organization As agents take on workflows that cut across marketing, sales, and customer service, companies need to rethink how growth is organized. The traditional model where each function operates in its own silo is giving way to an integrated system where agents coordinate activities, share data, and connect the entire customer journey from awareness to loyalty. Campaign design, lead conversion, and customer engagement are no longer sequential steps but parts of a single, learning loop. This shift requires a new, hybrid human–AI operating model. In this system, agents handle orchestration and execution, while humans provide strategy, creativity, and oversight. Growth teams become cross-functional by design, with marketers, sellers, customer service reps, and data scientists collaborating around shared workflows and common KPIs. Agents are reused across functions rather than duplicated: One agent that fetches customer data can support campaign planning, sales calls, or post-purchase service interactions. Agents for growth: Turning AI promise into impact 5 -- 5 of 7 -- Without effective governance and an agentic architecture, however, this scale can lead to “agent chaos” through redundant builds, inconsistent quality, and unmanaged risk. To scale effectively, leading companies are standing up agent factories: dedicated hubs that industrialize how agents are built, deployed, and governed. These hubs standardize reusable blueprints, shared data products, and guardrails for security and compliance. And the standardized agents they build are assigned clear, role-based responsibilities, so that lead agents orchestrate work, practitioner agents execute tasks, and QA and compliance agents monitor performance. Several global banks exemplify this approach, standing up agent factories to transform their due-diligence processes. Each factory deploys agent squads to handle discrete steps, from data extraction to validation and quality assurance, reducing manual work while improving accuracy and control. A leading North American manufacturer of outdoor lifestyle products applied similar principles to customer service. According to McKinsey analysis, after analyzing more than 30,000 service tickets and call transcripts, the company redesigned the function so agents handle diagnosis, data retrieval, and summarization, while humans focus on empathy and resolution. Adoption succeeded through a tailored change-management program that included leaders being trained on KPI dashboards, frontline staff receiving job aids for AI-assisted workflows, and technical teams learning model maintenance and tuning. Continuous feedback loops and shared dashboards keep both human and digital agents aligned to drive faster resolution times, higher satisfaction, and measurable revenue uplift. As these systems mature, the differentiator becomes human capability. The role of people shifts from completing tasks to supervising, refining, and improving how the work gets done. Managers and specialists must learn to delegate to agents, review outputs, identify exceptions, and guide learning loops. Emerging skills—such as prompt design, outcome tracking, and escalation management—are fast becoming core to modern growth roles. Many organizations already target 25 to 50 percent of employees to work regularly with agentic AI—a clear signal that fluency in collaborating with AI is becoming a defining business capability. One year into the agentic AI era, the lesson is clear: growth won’t come from tools alone but from how leaders choose to build and deploy them. Competitive advantage will depend not on how many agents a company launches but on how effectively they are designed, managed, and scaled. The companies pulling ahead are already putting new mindsets into practice. This is only the beginning of the change that agentic AI will bring—larger questions will soon loom, including: — When your sales agent negotiates with your customer’s buying agent, how will your Agents for growth: Turning AI promise into impact 6 company differentiate itself? -- 6 of 7 -- — If execution becomes commoditized, what will your brand stand for? — As workflows span silos, how will you ensure accountability and control? These are the leadership tests of the agentic era. Very soon the central question for leaders will move on from “what can this agent do for us” to “what outcomes am I prepared to deliver with it, and how can I best use the space it creates to allow humans to do what only they can do even better?” The sooner that organizations can embed agentic AI into their marketing, sales, and customer support operations, the sooner they will be able to answer those larger questions. Greg Kelly and Steve Reis are senior partners in McKinsey’s Atlanta office, and Lisa Harkness is a partner in the Connecticut office. The authors wish to thank Barr Seitz and Cindy Van Horne for their contributions to this article. Copyright © 2025 McKinsey & Company. All rights reserved. Agents for growth: Turning AI promise into impact 7 -- 7 of 7 --