Close Menu
The Financial News 247The Financial News 247
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
What's On

Preventing Data Loss, At Scale

November 18, 2026

Companies are spending Trump’s tariff refunds on bonuses – not reimbursing shoppers: report

November 18, 2026

Hiding In The Software Budget

November 18, 2026

AI Delivered The Savings, But Can Your Business Sustain Them?

November 18, 2026

Millionaire wages war on Saks after his mom gets punched, kicked by customer — then fired

November 18, 2026
Facebook X (Twitter) Instagram
The Financial News 247The Financial News 247
Demo
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
The Financial News 247The Financial News 247
Home » Hiding In The Software Budget

Hiding In The Software Budget

By News RoomNovember 18, 2026No Comments6 Mins Read
Facebook Twitter Pinterest LinkedIn WhatsApp Telegram Reddit Email Tumblr
Share
Facebook Twitter LinkedIn Pinterest Email

Rotem Alaluf | Founder and CEO of Wand AI, an enterprise AI software provider delivering AI solutions and accelerating AI transformation.

​Companies can measure what AI costs. Most are less prepared to measure the work their agents perform.

Ask a CFO what the company spent on AI last quarter, and there is usually a report: model fees, cloud consumption and software licenses. Ask how many exceptions its agents resolved, or which business outcomes they improved, and the answer is harder to find.

That isn’t a reporting problem. It’s a management problem.

In a previous Forbes article, “AI Stops Being A Tool The Moment It Owns An Outcome,” I argued that an AI system changes organizational status when it receives an ongoing responsibility rather than an isolated instruction. An agent tasked with keeping unreconciled transactions below 0.5% needs defined authority, escalation rules and a manager. It also needs to be measured against the objective it owns.

Two Workforces, Two Sets Of Numbers

Companies have systems for managing people: organizational structures, compensation, performance objectives and functional budgets. Their AI agents, meanwhile, appear in technical dashboards showing token consumption, API calls and error logs. Both are useful but don’t reveal how humans and agents perform as one workforce.

A department can reduce labor costs while increasing the human effort required to review faulty AI output. Neither outcome is visible in a cloud invoice.

The operating question isn’t whether an agent costs less than an employee. It’s which combination of human and agent effort delivers the required result, at what cost, quality, speed and risk. To answer that, both forms of work need to appear in the same management picture.

Why The Agent-Hour Is A Meter, Not A Result

Human labor has long been budgeted in hours. It’s an imperfect measure. A timesheet tells you how much time someone worked, not what the work accomplished. The hour gives organizations a way to plan capacity and compare the cost of delivering work.

In my book, The Agentic Age, I describe the agent-hour as a basic unit of AI labor: a measured hour of an identifiable agent operating under a defined mandate and governance controls. Unlike a bare compute charge, a governed agent-hour can be connected to the agent’s identity, assigned objective, authority, actions, escalations and audit trail.

That distinction matters. An hour is still an input, not an outcome. Two agents can consume the same number of hours and deliver different results. A company should measure agent-hours alongside verified output, quality, human supervision and total cost. It requires treating AI labor as something that can be managed, not merely purchased.

Consider a reconciliation team. Knowing that its agents consumed 2,000 governed hours gives the CFO an indication of capacity and spending. Knowing how many exceptions were resolved, how many were escalated and how many reopened tells the CFO whether those hours were productive.​

Not Letting Agents Certify Their Own Success

A dangerous shortcut is using an agent’s own completion message as evidence that it succeeded. “Task complete” may mean the agent produced an answer, not that a payment cleared, a claim was settled or a discrepancy disappeared from the system of record.

Verification has to come from evidence beyond the agent’s assertion: the reconciled ledger, the completed transaction, the policy-compliant decision or an independently checked sample. A second model can assist with evaluation, yet its judgment should not substitute for the underlying business evidence where that evidence is available.

The more useful scorecard operates at the level of the business function. It combines cost per verified result, turnaround time, error and rework rates, escalation frequency and the human effort still required. Those measures show whether adding agents improves the operation and where authority should or shouldn’t expand.

Work That Was Never Counted

AI ROI asks how many employee hours the technology saved. That misses an important category: valuable work that organizations left undone because the cost of performing it was too high.

Agents can expand the coverage and frequency of this work, provided the results are verified and exceptions receive oversight.

There may be no historical labor cost to eliminate because the work was never performed. The benefit could appear as fewer unresolved exceptions, earlier detection of risk or better service, not a reduction in headcount.

If AI is measured only against jobs it replaces, companies will miss the value it can create.

From Corporate Ledgers To National Accounts

Official GDP includes economic output regardless of whether it was produced with help from people, software or machines. What national accounts don’t separately show is the contribution made by autonomous AI agents within that output.

I use the term agentic GDP for an analytical view of the economic value attributable to agentic labor. It’s not additional GDP to add on top of existing output, and counting agent-hours alone cannot establish it. Measuring it would require connecting governed work to economic results while avoiding double counting in operations where people, agents and conventional technology collaborate.

As agents take on more work, the relationship between output, employment, wages and tax receipts may change. Governments will need better evidence to understand where productivity gains accrue, how income is distributed and whether fiscal rules need to adapt. The data cannot begin with an estimate invented at the national level. It must begin with what organizations can observe.

Counting It Before You Scale It

Start with one business function, not a companywide transformation program. Give every agent a named business owner and a defined objective. Put its governed hours and fully loaded costs on the same page as verified results, quality, escalations and human review effort. Use the same business outcome metrics to evaluate the combined human-agent operation.

​The first scorecard will be imperfect. That’s acceptable. What’s not acceptable is declaring an agent productive because a dashboard shows it was active or because it says the job is done.

Companies need to stop confusing deployment with performance. Before giving an agent more responsibility, leaders need to answer a question: When it says the work is done, what independent evidence proves it?

A workforce you cannot measure is a workforce you cannot manage.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Rotem Alaluf
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email

Related News

Preventing Data Loss, At Scale

November 18, 2026

AI Delivered The Savings, But Can Your Business Sustain Them?

November 18, 2026

When To Build, Buy And Automate

November 18, 2026

A Personal AI Assistant For Every One Of Us? Good Idea Or Not?

November 18, 2026

Your AI’s Confidence Score Won’t Catch Products You Don’t Carry

November 18, 2026

The Real Value Of Enterprise AI Lives In The Boring Work

November 18, 2026
Add A Comment
Leave A Reply Cancel Reply

Don't Miss

Companies are spending Trump’s tariff refunds on bonuses – not reimbursing shoppers: report

Business November 18, 2026

Major US companies are using the government’s refunds of President Trump’s tariffs to reward their…

Hiding In The Software Budget

November 18, 2026

AI Delivered The Savings, But Can Your Business Sustain Them?

November 18, 2026

Millionaire wages war on Saks after his mom gets punched, kicked by customer — then fired

November 18, 2026
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo
Our Picks

When To Build, Buy And Automate

November 18, 2026

A Personal AI Assistant For Every One Of Us? Good Idea Or Not?

November 18, 2026

Nvidia’s Jensen Huang agrees to testify before Senate, Chairman Tim Scott says

November 18, 2026

Your AI’s Confidence Score Won’t Catch Products You Don’t Carry

November 18, 2026
The Financial News 247
Facebook X (Twitter) Instagram Pinterest
  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact us
© 2026 The Financial 247. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.