Why CIOs need to connect AI spending to business outcomes

Measurable AI gains raise a larger investment question

A chief information officer (CIO) can show productivity gains from artificial intelligence (AI) while still struggling to sustain returns across the enterprise. Improving one activity shows that the technology can help, but extending that benefit requires changes to how the organization funds, assigns and manages work. Agentic AI, systems that can act independently within a workflow, makes these organizational decisions central to the investment.

The responsibility carries personal pressure too. In a survey published in April by Writer, an AI software company that benefits from enterprise adoption, 61% of tech leaders said they feared losing their jobs if they failed to guide their organizations through the AI transition. Enterprises are increasing AI investment to improve productivity even as returns remain unclear and the technology changes. So CIOs must make spending decisions while showing they can lead the changes those decisions require.

Yet leaders already have benefits to build on. Nearly half of tech leaders report measurable value from AI, while a report released in July by Accenture recognized agentic AI’s productivity potential alongside C-suite difficulty achieving lasting business value. Accenture sells technology and transformation services, giving it a commercial interest in companies pursuing that work. These findings concern different investment stages: a measurable gain can appear before an enterprise has established how to maintain and extend it.

Agent autonomy and business value require separate decisions

The prospect of greater value is driving plans for more independent systems. BCG forecasts that agentic systems will account for two-fifths of all AI value by 2030, and its report found that 42% of companies expect to give agents autonomy by then. BCG provides consulting services for business and technology change, so it benefits commercially when companies invest in AI transformation. Its forecast concerns agents’ expected economic contribution; the companies’ plans concern how much freedom agents will have to act.

But what will those permitted actions achieve? Lukic argues that autonomy alone does not create value and links stronger results to companies that have prepared their operations, data, responsibilities and controls for agents. More freedom may expand an agent’s technical capabilities, but the business result depends on how those capabilities change the work. Investment decisions therefore need to connect each proposed expansion of autonomy to the outcome it should improve.

Workflow redesign determines how agents enter daily work

Connecting agent actions to outcomes requires leaders to examine the whole workflow. Lukic distinguishes agentic AI from earlier prediction models by its ability to change work from beginning to end, beyond helping with an individual task. Companies are reporting value at that broader level, Lukic says. A faster task can improve one activity while leaving delays or coordination problems elsewhere in the process unchanged.

To address those problems, leaders need to examine the operating model: how an organization assigns work, decisions and responsibility. Lukic links stronger agent results to companies that have adapted this model alongside their data and controls. The assessment should establish which decisions agents will make, where responsibility sits and how the revised process improves the overall result. Data supports the work agents perform, ownership establishes who is accountable, and controls set the conditions under which agents may act, giving them a defined place in everyday operations.

Employees must then use these arrangements for the investment to affect daily work. Lukic warns that strong technical infrastructure cannot deliver the intended return when workflows remain unchanged or employees do not adopt the technology. Adoption therefore requires practical preparation for different tasks and responsibilities. Leaders need to include that preparation when assessing what an agentic deployment will require.

Lukic describes the work involved: “The bulk of the work is in redesigning processes, changing roles and ways of working, building new skills, and managing the organizational change required to make AI part of day-to-day operations,” Lukic said. Process redesign determines how work proceeds, while role changes establish who carries it out and takes responsibility. Skill development helps employees perform the revised work, making organizational change part of implementation.

BCG’s recommended allocation of AI-overhaul effort gives organizational preparation the greatest weight: 10% for algorithms, 20% for technology and data, and 70% for people, organization and processes. Algorithms are the computational methods that determine how a system performs its tasks; technology and data provide the environment and information those methods use. Each technical component retains a role, but most effort goes toward changing how the organization works. These percentages describe effort, so treating them as a mandatory budget split would change the recommendation.

CIOs must connect deployment, costs and outcomes

Once agents change work across activities and teams, the CIO needs a view across those boundaries too. Putting a system into operation establishes that it can run, but sustaining its contribution requires oversight of employee use, responsibility and business results. The workflow being redesigned should determine the extent of organizational change. Leaders can then judge whether the organization is prepared to use the system in the work it intends to change.

That judgment starts with visibility into where AI operates, which outcomes it supports and how much it costs overall. An inventory identifies deployments, but investment decisions require a connection between each deployment’s purpose, spending and result. Without it, leaders can see how widely AI is used while still struggling to decide where further spending will help.

Lukic places this responsibility within an expanded CIO role: “More often, CIOs need to play the role of a chief integration officer by setting the architecture and controls,” Lukic said. “Helping the business scale what works and making sure spend can be connected back to value, rather than managed as a collection of disconnected projects.” This role connects technical decisions with responsibility for extending successful uses across the business. It also puts the relationship between spending and results within the CIO’s remit.

Architecture and controls help the CIO carry out that responsibility across deployments. Architecture defines how systems fit together, while controls govern how they operate and what actions they may take. Both matter when an agent participates in a workflow involving several activities and owners. A decision that works for one deployment can affect other systems or responsibilities, so the CIO needs to assess it within the wider operating environment.

This enterprise view also helps leaders assess the total financial commitment. Managing projects separately can leave spending and benefits disconnected, making it harder to judge which uses deserve expansion. For an expansion decision, leaders need to establish which workflow changes, adoption measures and assigned responsibilities must carry over, and what additional organizational work the next deployment requires. Including that work in the investment decision gives the CIO a basis for committing the resources needed to extend the deployment.

Key executive takeaways

Connect early gains to lasting returns: CIOs assessing AI investments need to identify the funding, ownership and workflow changes required to sustain measurable productivity gains across the enterprise.
Tie agent autonomy to outcomes: Each proposed expansion of an agent’s authority needs a defined business outcome, accountable owner and controls. These requirements give CIOs a basis for deciding where greater autonomy warrants investment.
Fund workflow and workforce changes: Deployment plans need to cover process redesign, revised roles, employee skills and adoption. BCG recommends allocating 70% of AI-overhaul effort to people, organization and processes; that recommendation concerns effort rather than a mandatory budget split.
Make deployment costs and results visible: CIOs need an enterprise-wide view connecting each AI deployment’s purpose, total spending and business results. Expansion decisions also need to account for the organizational work required to carry successful uses into other workflows.