The Short Answer

AI is changing businesses by moving from isolated productivity experiments into the operating core of the enterprise. The biggest value is not a faster email draft or a prettier presentation. It is the ability to redesign judgment-heavy, non-deterministic processes that previously depended almost entirely on human attention.

That shift has a financial, operational, and managerial consequence: companies will need to treat AI as a capability, not as software procurement.

At the TIME100 AI Leadership Forum in New York, senior executives from Publicis Sapient, New York Life Insurance, and American Express spoke about AI in refreshingly practical terms. The message was not that AI will magically transform every company overnight. It was that AI is already exposing which organizations understand process, technology, governance, and talent - and which are simply chasing demos.

The serious enterprise question is no longer “Should we use AI?” It is “Which business processes are mature enough, governed enough, and valuable enough to be redesigned with AI?”

Productivity Is the First Chapter, Not the Book

Nigel Vaz, CEO of Publicis Sapient, described AI as a force for operational efficiency and better problem-solving. That is correct, but it is also only the opening move.

Most companies are still using AI where adoption is easy to explain: summarization, document drafting, internal search, coding assistance, customer-service support, and analytics acceleration. These use cases matter. They reduce friction. They free capacity. They create measurable gains.

But productivity gains alone rarely create a durable strategic moat.

The more important opportunity is process redesign. AI allows companies to execute workflows that are not fully deterministic. In plain English: processes where the next step depends on context, judgment, language, ambiguity, risk, or exceptions.

Examples include:

  • Reviewing insurance claims before escalation
  • Reading and classifying complex vendor contracts
  • Supporting credit-risk analysis with contextual evidence
  • Monitoring customer complaints for regulatory exposure
  • Preparing management reports from messy internal data
  • Assisting developers with legacy-code modernization
  • Coordinating multi-step operational workflows across systems

These are not “chatbot” use cases. They are operating-model use cases.

Technical Debt Has Become Strategic Debt

Vaz also raised one of the most important barriers facing established enterprises: technical debt.

This point deserves more attention than it usually receives. Many legacy organizations want AI outcomes while running on fragmented data, outdated systems, inconsistent permissions, weak process documentation, and years of workaround culture. AI does not magically solve that. In many cases, it exposes it.

If a company cannot answer basic questions about where its data lives, who owns it, which systems are authoritative, and which process exceptions are acceptable, AI implementation becomes fragile.

Technical debt is no longer just an IT concern. It directly affects:

  • Time to deploy AI use cases
  • Accuracy and consistency of model outputs
  • Cybersecurity and data leakage risk
  • Regulatory defensibility
  • Integration costs
  • Employee trust in AI recommendations
  • Return on investment

For CFOs, this means AI business cases must include infrastructure readiness. For COOs, it means process standardization becomes an AI accelerator. For CIOs, it means architecture decisions now shape competitive speed.

An enterprise with cleaner data, clearer workflows, and better API discipline will move faster than a competitor with a larger AI budget but weaker foundations.

The Human Amplifier Argument Is Right, But Incomplete

Deepa Soni, senior executive vice president and CIO at New York Life Insurance, framed AI as a human amplifier rather than a human replacement. This is the right message for conservative industries, especially insurance, banking, healthcare, and regulated financial services.

AI adoption fails when employees feel it is being imposed on them as a surveillance or replacement mechanism. It succeeds when people experience it as leverage: fewer repetitive tasks, faster analysis, better preparation, and more capacity for professional judgment.

Still, “human in the loop” must be designed carefully.

If every AI-supported process requires a human to inspect every step manually, the organization has not transformed much. It has simply added another screen to the workflow.

The better model is this:

  • Humans define the objective, policy, and risk boundaries
  • AI executes repeatable judgment-support tasks within those boundaries
  • Exceptions, uncertainty, and high-risk cases are escalated
  • Humans supervise many processes instead of manually executing one process
  • Performance is monitored continuously with feedback loops

This is the management shift many organizations miss. The goal is not to keep one employee approving one AI output at a time forever. The goal is to help an employee who previously handled a single workflow supervise dozens or hundreds of AI-assisted workflows safely.

That is where productivity becomes operating leverage.

Learning From Failed Assumptions Is Not Failure

Ravi Radhakrishnan, CIO of American Express, made another valuable point: early assumptions about where AI would create value did not always survive contact with reality.

This is normal. In fact, it is healthy.

Enterprise AI should not be managed like a one-time ERP rollout where every requirement is frozen in advance. It should be managed as a disciplined portfolio of experiments, with clear gates for value, risk, adoption, and scalability.

A strong AI experimentation model includes:

  • A business owner who is accountable for the workflow
  • A measurable baseline before AI is introduced
  • A clear definition of success and failure
  • Security and privacy review before deployment
  • Human escalation rules
  • Testing on real operational edge cases
  • A plan for maintenance, monitoring, and retraining where relevant
  • A decision point: scale, revise, or stop

The problem is not failed pilots. The problem is pilots that teach the organization nothing.

AI Is Not Only Technical

One of the most damaging misconceptions in the market is that AI implementation is mainly a technical exercise. It is not.

AI sits at the intersection of technology, management, domain expertise, process design, behavioral change, risk, data governance, and finance. That is why shallow advice can be so expensive, especially for small and mid-sized businesses that may not have the internal filters large enterprises use when evaluating consultants.

There are many self-appointed AI experts. Some know how to produce impressive demonstrations. Far fewer understand how to redesign a claims process, procurement workflow, sales operation, finance close, or compliance review in a way that survives real organizational pressure.

Relevant education matters. Academic depth matters. Business experience matters. Applied implementation experience matters. AI is a multidisciplinary field, and the strongest practitioners often combine technical understanding with deep knowledge of professional workflows.

That combination is what turns AI from a novelty into infrastructure.

The Two Tracks Every Enterprise Needs

Organizations should not choose between AI literacy and AI agents. They need both.

AI literacy builds the workforce’s ability to communicate effectively with models, evaluate outputs, protect sensitive information, and use AI tools responsibly. This is now a basic professional skill. Employees who can brief, challenge, and refine model outputs will outperform those who treat AI like a search box.

Agent development is different. It requires organizational infrastructure for building, deploying, monitoring, securing, and improving AI agents that perform defined tasks across systems.

The surprising part is that agents can sometimes be easier to adopt than general AI tools. A well-designed agent can fit into an existing workflow with minimal behavior change for employees. By contrast, broad AI tools often require people to change how they write, research, analyze, document, and collaborate.

That does not mean agents are technically simple. It means the adoption burden can be lower when the process design is strong.

Enterprises should therefore invest in both tracks:

  • Train employees to work effectively with AI tools
  • Build internal capability to create and manage AI agents
  • Establish governance for agent permissions and auditability
  • Create reusable integration patterns
  • Measure operational impact, not just usage
  • Treat agent management as an emerging organizational function

In time, information systems departments may become something close to human resources departments for AI agents: onboarding them, assigning permissions, monitoring performance, retiring poor performers, and ensuring they operate within policy.

Tool Choices Matter, But Operating Model Matters More

The platform conversation is important, but it should not become a substitute for strategy.

Claude remains one of the strongest options for broad enterprise AI work, particularly where language quality, reasoning, and practical usability matter. Claude Code and Claude’s collaborative work patterns are among the more effective applied AI tools currently available for teams that know how to use them. At the same time, enterprise security, data handling, and compliance requirements must be taken seriously before broad deployment.

Microsoft Copilot is a credible infrastructure layer for many organizations, especially those already committed to the Microsoft ecosystem. It has sometimes moved more slowly than faster AI-native companies, but recent improvements show Microsoft is accelerating. Copilot Studio is also a reasonable path for Microsoft-based agent implementation.

At the same time, tools such as n8n are entering enterprise environments in ways that would have seemed unlikely a few years ago. Large organizations are increasingly open to workflow automation platforms that can connect systems, orchestrate agents, and help teams move faster.

The winning question is not “Which vendor is best?”

The better question is: “Do we have the internal architecture, governance, and professional capability to build useful AI systems repeatedly?”

A company that lacks this capability will struggle even with excellent tools. A company that builds it will be able to switch, combine, and govern tools as the market changes.

Trust, Service, and Security Will Decide the Winners

Radhakrishnan’s final point may be the most important: when the noise settles, customers will still choose companies based on trust, service, and security.

AI does not cancel the fundamentals of business. It intensifies them.

Customers will not forgive a bank because its failed support experience was “AI-powered.” Regulators will not excuse weak controls because the model was advanced. Employees will not trust outputs that cannot be explained. Boards will not accept AI investments that lack financial discipline.

The organizations that win with AI will be those that connect innovation with operational seriousness.

That means:

  • Use AI to improve service, not hide poor service
  • Automate carefully where risk is low and value is clear
  • Keep humans in the loop where judgment, ethics, or regulation demand it
  • Avoid turning human review into a bottleneck
  • Build internal AI capability rather than depending entirely on external vendors
  • Measure AI by business outcomes, not presentation quality
  • Treat cybersecurity and data governance as design requirements from day one

The Boardroom Agenda Has Changed

AI is no longer a side topic for innovation teams. It belongs in board discussions about cost structure, resilience, customer experience, technology debt, workforce capability, and competitive strategy.

The companies making real progress are not necessarily the loudest. They are the ones doing the harder work: mapping processes, cleaning data, training employees, building agent infrastructure, testing assumptions, and creating governance that allows speed without recklessness.

That is the real message from the executive conversation. AI is not just another tool entering the enterprise. It is forcing companies to reveal how well they understand their own operations.

And for many organizations, that may be the most valuable disruption of all.