The missing operating discipline behind enterprise AI

Insights

• Organizations have moved from exploring AI to deploying AI across sales, service, and marketing. Early pilots have shown that technology can improve productivity, automate routine work, and enhance customer experiences.
• Yet many organizations are discovering that deploying AI and realizing value from AI are two different challenges. Successful pilots often fail to translate into repeatable business outcomes across the enterprise.
• The reason is rarely the technology itself. Data fragmentation, disconnected workflows, governance challenges, workforce adoption, and unclear measures of value frequently become barriers to scale.
• Organizations that achieve the greatest impact treat AI as an operating capability rather than a collection of individual use cases. They build the disciplines needed to prioritize opportunities, integrate AI into workflows, coordinate execution, and measure outcomes.
• This is what "Less AI Work. More AI That Works." means in practice: shifting the focus from launching more AI initiatives to creating measurable business value from the AI that has already been deployed.

AI is already helping customer-facing teams make better decisions through recommendations, insights, and automation. The next stage is the rollout of AI agents, tools that can operate autonomously in pursuit of defined goals, which are taking a more active role in customer-facing operations.

Across sales, service, and marketing, AI agents are already demonstrating their ability to improve productivity, accelerate response times, and automate routine work. In many cases, the technology is proving its value through successful pilots and early deployments.

Yet a persistent challenge remains. While some organizations are extending these successes across the enterprise, others struggle to move beyond isolated use cases. AI initiatives continue to attract investment and attention, but sustained business impact often proves harder to achieve than expected.

Understanding why this gap exists has become increasingly important as organizations move from using AI for recommendations to using AI for execution.

CRM enters the age of autonomous execution

Customer relationship management (CRM) platforms have traditionally served as systems of record, helping organizations capture customer information, track interactions, and manage sales, service, and marketing activities. AI added another layer of value through recommendations, lead scoring, summarization, and predictive insights, while employees remained responsible for making decisions and taking action.

CRM is expanding beyond its traditional role as a system of record to support customer-facing actions and workflow execution. AI agents can qualify leads, resolve service requests, recommend next-best actions, orchestrate customer interactions, and execute follow-up activities within predefined guardrails. Gartner predicts that by 2029, agentic AI will autonomously resolve 80% of common customer-service issues, reducing operational costs by 30%.

Salesforce's vision for Agentforce, Data Cloud, and its broader AI-first CRM strategy reflects this shift, bringing together AI agents, trusted customer data, governance controls, and observability capabilities to support more autonomous customer-facing operations.

Agents are easier to deploy than they are to operationalize

Although agentic AI offers the potential to respond faster, personalize engagement, and improve customer experience, many organizations remain stuck in pilot purgatory. Business value often remains confined to individual initiatives rather than extending across the enterprise. Infosys Knowledge Institute research found that only about 20% of AI use cases are delivering on all business objectives, while another 30% are close to doing so, highlighting the challenge of converting AI experimentation into sustained value at scale.

This is where many organizations discover that deploying AI and operationalizing AI are fundamentally different challenges. The first requires technology. The second requires operating discipline. That is the difference between AI work and AI that works.

Many organizations continue to carry significant data and process debt, including fragmented data, governance challenges, legacy integrations, and workflows that were never designed for human-agent collaboration. The same research found a strong relationship between AI success and changes to operating models and data architecture, suggesting that technology deployment alone is rarely sufficient to deliver value at scale. As AI moves from experimentation into day-to-day operations, these shortcomings become increasingly difficult to ignore.

Infosys' client engagements illustrate these challenges. At a large US health-plan organization, customer-service representatives were required to search across Salesforce knowledge articles, internal applications, intranet content, and external documents while responding to member queries. Infosys’ generative AI-powered platform helped the client to bring all the required information together, contributing to lower average handling time, higher first-call resolution, and improved member and representative experiences. This example shows how information spread across Salesforce and other enterprise systems can constrain service performance.

Additionally, organizations face growing pressure to demonstrate business value while managing the economics of adoption. Consumption-based pricing models, infrastructure choices, model-selection decisions, and expanding workloads can introduce uncertainty around costs and return on investment. A solution that appears viable during a pilot can become far more difficult to justify as adoption expands across business units and geographies.

Infosys encountered a similar challenge at a UK media organization. The company faced a projected 70% increase in customer-service cases while its 85-person contact center was already operating at peak capacity. Infosys helped the company by analyzing requests by complexity and channel before introducing self-service, spam filtering, and AI-assisted service capabilities. The pilot achieved approximately 30% call deflection, identified around 20% of emails as spam, and created projected capacity savings of up to 50%.

This demonstrates that AI initiatives must begin with a clear understanding of the operational constraints and suitable use cases. The organization faced a clear capacity constraint, and the solution was designed around service complexity and channel requirements rather than technology alone.

While these situations appear different, they reveal that the obstacles to scaling AI frequently extend beyond technology into data, workflows, governance, workforce readiness, and measurement. Organizations struggle when data, workflows, governance, workforce readiness, and measurement evolve more slowly than the technology itself. Employees need to understand how AI changes day-to-day work, where human judgment remains essential, and how success will be measured if organizations hope to achieve adoption at scale.

Organizations must do the work to identify opportunities, redesign work, coordinate execution, manage costs, and continuously measure outcomes before they move from pilots into production. In other words, success depends on creating less AI work and more AI that works.

Build the operating discipline for enterprise AI

Organizations that successfully scale agentic AI recognize that deploying agents is the starting point. Sustained value depends on an operating discipline that determines how AI opportunities are prioritized, embedded into business processes, coordinated across the enterprise, and measured over time.

Infosys' approach defines four disciplines to help organizations move from selecting AI opportunities to defining use cases, deploying solutions, and measuring business outcomes.

1. Opportunity discipline

Many organizations begin their AI journey by pursuing multiple use cases simultaneously. However, BCG research found that organizations generating the greatest value from AI focus on a smaller number of high-impact initiatives rather than spreading investments across a larger portfolio. Leading organizations prioritize opportunities based on business impact, implementation complexity, scalability, and strategic relevance.

Infosys' Agentic Discovery approach helps organizations identify and prioritize AI opportunities with the greatest potential business impact. For example, a healthcare organization developing an enterprise AI roadmap worked with Infosys to evaluate multiple AI opportunities before defining its enterprise roadmap. Rather than pursuing the most advanced technology available, the organization assessed business value, governance requirements, scalability, and long-term operating implications before prioritizing the opportunities most likely to generate sustainable results.

The key takeaway here is that successful organizations are selective. They focus investment on opportunities that solve meaningful business problems, can scale across the enterprise, and have a clear path to measurable value.

2. Workflow discipline

AI creates the greatest value when it becomes part of how work is performed. Rather than treating AI as a standalone capability, organizations need to embed it into the day-to-day activities of employees and customer-facing teams.

This requires a clear understanding of where AI can reduce manual effort, accelerate decision-making, improve access to information, or automate repetitive tasks within existing business processes. Infosys’ use-case framework supports this effort by helping organizations define how AI fits into workflows before deployment.

The benefits of this approach are already visible. For example, a large US financial institution embedded Salesforce Einstein capabilities, including knowledge recommendations, case summarization, and AI-assisted response generation, into customer-service workflows. Agents were able to access relevant information more quickly, create and update knowledge content more efficiently, and reduce the time required to understand and resolve customer issues. The initiative improved agent productivity, accelerated knowledge creation, improved case-resolution times, and reduced operational effort.

The lesson here is that AI adoption becomes easier when employees can use it within the flow of their work and see a clear benefit from doing so. Organizations that successfully scale AI focus on integrating AI into business processes rather than introducing it as a separate technology layer.

However, technology alone does not guarantee adoption. As AI becomes embedded in day to-day operations, employees need clarity on how work will change, where human judgment remains essential, and how success will be measured. Infosys Knowledge Institute research found that effective employee engagement and change-management practices can improve AI success rates by up to 18 percentage points, highlighting the importance of workforce readiness alongside workflow redesign.

3. Orchestration discipline

Organizations need a way to connect data, workflows, decisions, and AI capabilities across teams and systems. This affects how customers experience their services. Effective orchestration ensures that agents operate with consistent context, governance, and accountability.

Enterprise orchestration is also built on trust. AI agents must operate on reliable customer data, remain within defined governance boundaries, and provide sufficient transparency for business users, customers, and regulators to rely on autonomous decision-making.

This is where Salesforce's investments in Agentforce, Data Cloud, governance controls, and observability capabilities become increasingly important. Infosys complements these capabilities through enterprise orchestration frameworks and broader AI ecosystem integration, helping organizations maintain flexibility while supporting governance, interoperability, and consistent customer experiences.

4. Value realization and cost discipline

Many organizations still evaluate AI success through activity metrics such as the number of agents deployed or use cases launched. Leading organizations focus on outcomes.

They evaluate whether AI improves customer satisfaction, reduces service effort, increases productivity, accelerates revenue generation, improves retention, and delivers measurable business value.

They also monitor the costs of adoption. As AI usage grows, costs can be influenced by factors such as pricing models, infrastructure requirements, technology choices, and increasing levels of usage. A deployment that appears successful during a pilot can become difficult to scale if costs rise faster than the value it creates.

This challenge was evident at another healthcare organization as it compared AI approaches ahead of its own roadmap decisions. While some options offered more advanced capabilities, the organization prioritized scalability, governance, and cost efficiency to maintain viability as adoption expanded.

Sustained value requires more than periodic reporting. Organizations need a structured way to connect AI initiatives to business and operational outcomes, define success measures before deployment, and continuously monitor performance after launch. Infosys Business Value framework is one approach that helps establish this connection by linking AI use cases to business and operational key performance indicators throughout the life cycle.

Post-deployment AI care is equally important, requiring organizations to continuously monitor performance, optimize costs, address emerging risks, and refine AI capabilities as business conditions evolve.

Institutionalize AI as an operating capability

As customer-facing operations become more reliant on agentic AI, leaders must manage AI as an enterprise capability rather than as a series of standalone projects.

Establish executive ownership
AI initiatives often span technology, operations, customer experience, risk, and business functions, making accountability difficult. Leaders should define clear ownership for outcomes, adoption, governance, and investment decisions across the AI life cycle.

Create common operating principles
As AI agents take on greater responsibility for customer-facing activities, organizations need clear guidelines for how humans and AI work together. This includes decision rights, escalation paths, governance controls, workforce expectations, and standards for coordinating AI across platforms, business functions, and data environments.

Measure outcomes and costs continuously
Leaders should establish mechanisms to constantly track customer experience, productivity, revenue impact, operational efficiency, adoption rates, and workforce engagement. They also need visibility into model costs, infrastructure choices, consumption-based pricing, and long-term operating expenses to ensure value creation remains aligned with investment.

Organizations that derive the greatest value from AI treat it as a managed business capability rather than a collection of isolated deployments. They invest not only in technology, data, and governance, but also in workforce readiness and change management to ensure AI becomes part of how work gets done across the enterprise.

The next competitive advantage will be operational

As agentic AI becomes embedded across sales, service, and marketing, CRM platforms are evolving to capture, shape, and execute customer interactions. The implications extend beyond automation. Organizations will need to rethink roles, redesign operating models, establish new governance mechanisms, and create new forms of collaboration between humans and AI agents.

Competitive advantage will come from how effectively organizations operationalize AI across the enterprise. Organizations that prepare for embedding AI into their business will be best positioned to create differentiated customer experiences, adapt to changing market conditions, and unlock new sources of growth.