Agentic AI is about to alter how corporations create worth. Morgan Stanley estimates AI might add $13 to $16 trillion in worth to the S&P 500, with practically $500 billion estimated coming from agentic AI alone. But, most enterprises aren’t prepared. In response to new analysis from Asana’s Work Innovation Lab, solely 20% of organizations have efficiently scaled AI brokers, whereas practically a 3rd don’t also have a plan. Staff, in the meantime, already anticipate to delegate 43% of their workload to AI brokers inside three years.

The issue isn’t the know-how — it’s the planning and execution. Too many pilots stall out as a result of CIOs haven’t constructed the methods, guardrails and tradition to maneuver past experiments. The expectation has shifted: We’re now not requested to “attempt AI.” We’re answerable for scaling it safely and displaying clear enterprise worth.

Right here’s my sensible playbook for making that shift.

From particular person instruments to collaborative brokers

We’ve entered the period of human-agent collaboration, the place a number of AI brokers can coordinate throughout total groups and capabilities. It is a basic shift from the “one individual, one AI instrument” mannequin. But most organizations stay caught in pilot mode, the place brokers are restricted to slender duties like summarizing notes or drafting emails.

The actual alternative lies in designing methods the place brokers don’t simply help people however work collectively — sharing info throughout groups, coordinating dependencies and producing outcomes that leaders can belief.

Think about one agent serving the wants of a number of stakeholders without delay, moderately than dozens of remoted assistants duplicating effort, and studying on the run from human suggestions.

So how will we get there? Our analysis exhibits the hole comes all the way down to construction. In organizations that present formal coaching, practically 80% of staff really feel assured utilizing AI, in comparison with simply 40% the place coaching is absent. And governance issues simply as a lot: Solely 19% of staff say their firm has clear guidelines for when to depend on AI versus individuals. The takeaway is to deal with AI like infrastructure, not one-off experiments.

What CIOs ought to do:

• Begin with outcomes, not governance. The primary query to ask isn’t about insurance policies or buildings — it’s “The place can AI assist us clear up our hardest enterprise issues?” CIOs who anchor AI in enterprise outcomes create a transparent mandate for adoption and guarantee governance serves technique, not the opposite approach round.
• Construct the construction round it. As soon as the priorities are clear, create the guardrails to scale responsibly. That usually means a cross-functional AI council that brings IT, authorized, safety, HR and enterprise leaders into one room. Assign homeowners for every agent — like product homeowners — who’re accountable for security, updates and outcomes. This stops shadow initiatives earlier than they unfold and provides leaders visibility into experiments whereas they’re nonetheless manageable.
• Empower individuals with each guidelines and instruments. Governance alone gained’t change how work will get accomplished. Outline what brokers can and might’t do, but in addition give staff the coaching and freedom to make use of AI on their most repetitive work and their hardest issues. Encourage shared studying throughout groups so profitable use instances compound. This grassroots innovation, paired with a top-down technique, is what takes AI past scattered pilots and makes it a part of how the enterprise runs.

You may’t automate chaos. Redesign workflows first

Too many corporations bolt AI onto messy processes and anticipate transformation. The actual alternative is to reimagine workflows with AI as a collaborator from the beginning. Which means asking: “What ought to this course of appear like if brokers are serving to us run it?”

Begin with high-volume, low-risk processes that have already got construction — like IT ticket triage, expense approvals or advertising marketing campaign consumption. Redesign them in order that they work easily with people and brokers collectively. In any other case, AI simply scales the chaos sooner than individuals ever might.

Steps to take:

• Redefine workflows for AI-human collaboration. Map possession, inputs and outputs and escalation paths — not simply to tidy up, however to resolve the place brokers add probably the most worth and the place people want to remain in management.
• Construct an approval and deployment course of. Require consumption types, sandboxing, safety opinions, privateness checks and audit logs. Maintain a registry of all fashions and brokers with clear service ranges and troubleshooting guides.
• Create a data base. Publish a central library of accessible brokers with utilization directions and real-world examples.
• Discover champions. Establish early adopters to co-design new workflows and train others finest practices.

Executed proper, even easy rule-based brokers change into the inspiration for higher-autonomy methods — as a result of the workflows they’re embedded in had been designed for them, not simply retrofitted after the actual fact.

Choice-making is shifting (quick) — govern it

Brokers are transferring from simply executing duties to supporting choices — and more and more influencing them. CIOs must resolve: The place ought to brokers have autonomy and what oversight is required?

A tiered method works finest:

• Simulate: Brokers recommend actions, however people resolve (e.g., advisable contract modifications).
• Shadow: Brokers act in parallel and people confirm (e.g., price range allocation recommendations).
• Partial automation: Low-risk autonomy with human escalation (e.g., approving bills below $500).
• Supervised autonomy: Larger autonomy with common human checkpoints (e.g., routing and resolving tier 1 assist tickets).

This framework clarifies your threat tolerance whereas assembly compliance and audit necessities. As CIO, you personal the matrix of what’s delegated, what’s supervised and what will get logged.

Make ‘agent supervisor’ a core workforce ability

AI modifications jobs greater than it eliminates them. The important thing ability now’s managing brokers: directing their work, checking their output and enhancing their directions.

Our analysis exhibits staff are cautiously optimistic — 52% suppose AI will assist their work, however 29% fear it’d substitute them. CIOs have a important function to play in serving to staff see AI as a teammate, not a menace — by steerage, coaching and clear examples of the place it provides worth.

A sensible method:

• Position-based coaching tracks: AI data isn’t one-size-fits-all. Builders (usually IT or operations) want the technical depth to create and keep brokers. Reviewers (threat, governance and high quality groups) must know easy methods to consider and approve outputs. Operators (each day enterprise customers) want sensible steerage on easy methods to work with brokers and supply suggestions. Structuring coaching this fashion avoids generic “AI 101” classes and ensures individuals be taught what’s instantly related to their function.
• Make the “agent supervisor” function specific: Don’t deal with agent supervision as casual or further work. Replace job descriptions, efficiency metrics and incentives so staff know that supervising, correcting and enhancing agent workflows is a part of their core job.
• Accomplice with HR on sentiment: Coaching solely works if staff really feel supported. Work with HR to commonly verify belief ranges, readiness and adoption emotions. Catching issues about job safety or workload early permits you to modify coaching and communication earlier than skepticism turns into resistance.

This method doesn’t simply drive adoption — it builds resilience in a blended human-AI workforce the place staff see themselves as companions with brokers, not opponents.

The brand new ROI: Return on intelligence

Conventional ROI metrics — effectivity and price — don’t seize AI’s full impression. CIOs ought to measure return on intelligence: the mixed enchancment in determination high quality, throughput and innovation when people and brokers work as one system.

Key metrics to trace:

• Effectivity: cycle time, cost-to-serve, throughput per worker
• Choice high quality: error charges, rework prevented, approval lead instances
• Innovation: time-to-insight, experiments launched, function velocity
• Belief & adoption: weekly lively customers, human-in-the-loop acceptance fee, audit cross charges
• Resilience: incident response instances, rollback velocity, change failure charges

Construct these metrics into your deployment from day one. The objective isn’t “extra AI” — it’s higher outcomes and extra engaged groups.

The CIO playbook: 4 strikes to start out this quarter

1. Anchor AI in enterprise outcomes. Start by asking: “The place can AI assist us clear up our hardest enterprise issues?” This frames adoption round worth, not know-how.
2. Kind a cross-functional AI council. As soon as priorities are clear, carry IT, authorized, HR, threat and enterprise leaders collectively to set threat ranges, determination rights and consumption processes. Assign homeowners accountable for every agent’s efficiency.
3. Pilot rule-based brokers in key workflows. Select three high-volume processes and transfer intentionally by the trail from simulate to shadow, then partial automation and supervised autonomy. This builds belief and a path to increased autonomy with out surprises.
4. Observe ROI on two ranges. Measure effectivity positive aspects, but in addition human outcomes like engagement and burnout. Pair this with role-based coaching so staff have the abilities and confidence to make AI a part of their each day work.

Scale methods, not simply instruments

The businesses that scale responsibly now gained’t simply get extra environment friendly — they’ll set the tone for a way staff belief and undertake AI over the following decade. For CIOs, that is greater than deploying know-how: it’s shaping the way forward for work itself.

Do that proper and also you gained’t simply modernize your organization. You’ll present what clever, resilient, human-centered organizations appear like within the age of AI brokers.

This text is printed as a part of the Foundry Knowledgeable Contributor Community.
Wish to be part of?