AI ROI Is Becoming A Public Promise

For two years, leaders have been told to put a number on AI’s return.

This week, companies began putting their names beside one.

Sysco told investors it is targeting at least $500 million in AI-powered efficiency savings by fiscal 2029. Chewy said it sees a path to approximately $50 million or more in annualized savings as its AI capabilities scale. That is a meaningful change from reporting users, messages, pilots, or hours saved. A named dollar amount and a deadline create something leadership, investors, and employees can revisit.

But a savings target is still not a realized return.

It is a public promise about what the operating system should produce in the future. Its credibility depends on what sits underneath the headline: the baseline, the included workflows, the accountable owner, the timing, the gross-versus-net treatment, and the method used to separate AI’s contribution from everything else changing in the business.

This week’s disclosures show what stronger AI accountability can look like. They also show why the number is only the beginning.

Here are the numbers—and what they mean for professionals preparing to make an AI commitment of their own.

The market is moving from AI aspirations to AI commitments.

That is progress. A commitment gives the organization a destination. But the dollar amount becomes useful only when leadership can show how the business is expected to reach it.

A defensible AI commitment needs five parts:

• Baseline: What did this workflow cost or produce before the AI-enabled change?
• Boundary: Which processes, teams, locations, and financial lines are included—and which are not?
• Owner: Who is accountable for implementation, adoption, operating performance, and the financial reconciliation?
• Clock: When should leading indicators move, when should operating results appear, and when will the final return be tested? - Net value: What remains after platform, integration, data, training, security, human review, maintenance, and change-management costs are included?

Sysco’s disclosure is notable because several of those elements are visible. The company did not simply announce $500 million and walk away. It named a fiscal deadline, identified operational workstreams, established a first-year net-savings target, and connected the cost-out goal to long-term incentives.

That still does not guarantee the result. It makes the result testable.

The same discipline matters when the goal is growth rather than cost reduction. Bain’s analysis suggests that innovation and competitive shifts represent a much larger opportunity than productivity alone. But “AI will help us grow” is not a measurable growth case.

Leadership still has to name the new offer, customer segment, conversion point, capacity gain, retention change, or revenue-producing workflow. It has to decide what evidence would show that AI contributed to the result. And it has to measure contribution margin—not revenue alone—after the full cost of delivering the new value.

The goal is not to make every AI decision look certain. The goal is to make the assumptions visible enough to manage.

Consider a 30-agent real estate brokerage making a clear AI commitment: help each agent close one additional transaction during the next 12 months.

That is specific. It is also incomplete. The brokerage first needs a baseline. How many agents currently close at least one transaction? What is the average lead-to-appointment and appointment-to-client conversion rate? How many viable opportunities go cold because follow-up is late, inconsistent, or never completed?

Then it needs a boundary. Is the commitment tied to lead follow-up, database re-engagement, client communication, recruiting, property marketing, or all of them? If every AI-enabled activity is included, no one will know which workflow created the result.

The owner must be clear too. The platform cannot own adoption. Someone in leadership must confirm that the workflow is connected, the approved messaging is being used, exceptions are reviewed, and agents are not quietly returning to manual habits.

The clock should include more than the final closing count. In the first weeks, the brokerage may look for contacts organized, follow-up completed, and agents using the process. Next come response rates, conversations, appointments, signed clients, and active pipeline. Closed transactions arrive later.

Finally, the financial calculation must use company dollar—not the total sales volume or commission attributed to the agent—and subtract the complete cost of the platform, implementation, onboarding, integrations, staff oversight, and maintenance.

Now “one more sale per agent” is not a slogan. It is a testable operating commitment with a chain of evidence leading to the P&L. This is the work EMDUIT exists to build. Your AI is the platform. EMDUIT is the workforce that connects the platform to the people, workflows, controls, and measures required to earn the return.

A dollar target creates attention. A measurable operating plan creates accountability.

If your organization announced an AI savings or growth target today, could your team show exactly how the number will be earned?

If you are a licensed professional or business leader, share the AI commitment your organization needs to make measurable before it makes it public.