I haven’t used traditional office software in over nine months. For thirty years, my work lived in documents and spreadsheets. I launched Office 97 in Latin America. Fourteen years later, I co-founded what became Office in the Cloud. So why did I stop? Because my AI agents now do those tasks for me.
It started small. A five-minute saving here. Another task there. That is why AI initially looks invisible. One five-minute saving feels too small to matter. A chain of them changes how a company works.
🎧 Go Deeper: In this week’s companion episode, we unpack why most Fortune 100 leaders can’t see the return from AI, and where the value is actually hiding. Listen on Apple Podcasts | Spotify
🛑 The Dashboard Is Looking in the Wrong Place
IBM and McKinsey both report it. Fortune 100 leaders tell me the same. They can’t find the return on their AI investment. When they ask me where to look, I don’t start with the platform. I start with the unit of measurement. Their dashboards count licenses, seats, adoption rates, and uptime. None of those numbers show an analyst finishing a partner brief in 12 minutes instead of 2 hours.
The work has changed. The measurement system hasn’t.
In our last issue, we showed that AI is working at the individual level and failing at the company level. This issue goes one layer deeper. The value is hiding inside the smallest unit of work: the task.
Early AI wins usually look like 20 minutes saved here and 60 minutes there. They hide inside the daily tasks we all complete.
⏱️ Minutes Become Hours, and It All Starts With Tasks
Across 70 countries, our research shows proficient AI users save 4 to 8 hours every week. The headline number is easy to repeat. The composition is what executives need to see.
Source: AI4SP Global Tracker 2026, n=8,000+ knowledge workers across 17 industries.
Here is what those medians hide in practice. AI saves about 8 minutes on a routine email. 60 minutes on a short report. 87 minutes on a complex customer escalation. 140 minutes on a field repair. None of these is dramatic on its own. None shows up on the company dashboard. Together, they are how proficient AI users save 4 to 8 hours a week.
That work fills calendars and never makes the strategy slide. That is exactly why traditional dashboards miss it.
🔗 The Chain Effect: One Analyst, Three Hidden Wins
Take Maya, a senior analyst at a commercial real estate firm. Every morning, she spent 2 hours reading 12 sources, then assembling a partner brief for the 9 a.m. meeting. Now the brief takes 12 minutes. The sources are linked. The analysis is sharper.
Her partners now get the brief by 7 a.m. They have their strategy locked before competitors check their email. One personal time-saving measure accelerated the firm’s time-to-market.
Then five other analysts copied her agent. Maya only triggered it once a day to run autonomously. The IT dashboard saw one short session, or nothing at all if it only counted conversations. It missed five analysts. It missed the partner meetings. It missed the competitive position.
A traditional AI dashboard tracks licenses and adoption. A distributed AI dashboard tracks completed tasks and their impact.
🏦 The BBVA Move: Use It or Lose It
BBVA, one of Europe’s largest banks, just published its model in Harvard Business Review. Their leadership noticed shadow AI across the company and made the call most leaders never make. Instead of shutting it down, they handed 3,000 ChatGPT Enterprise licenses to the most motivated people on every team. Not the most senior. The most motivated. With one rule: use it or lose it.
Demand outpaced supply within weeks. They built a peer network around those users. Champions in each business unit. Wizards as local experts. In under a year:
| 3,000 | Initial license pool |
| 11,000+ | Active users within a year |
| 4,800+ | Custom agents built by the frontline |
Source: Harvard Business Review, April 2026.
This connects directly to The Two Percent, the people in every company who teach themselves AI before the formal program catches up. They sit two or three quarters ahead of any rollout because they live closest to the friction. BBVA didn’t fight them. BBVA funded them.
A consulting firm we advise found a similar pattern: 90% of their employees were already using unauthorized AI. Instead of bans, they built a guardrails-and-certification model that kept governance moving at the speed of the work. We walk through their full playbook in the companion episode.
“Every successful AI deployment we’ve guided comes back to the same two things: change management and proper enablement. The approach that worked for BBVA with ChatGPT works with Claude Cowork, with Microsoft Copilot, with whatever tool sits on the desk. Each one has its own nuances and its own learning curve.
In our experience, when leadership funds the grassroots energy already inside the company, and the work starts at the task level, ROI shows up in months, not years.”
💰 The Two Tests for a CFO Conversation
Not every five-minute saving deserves the CFO’s attention. A task improvement earns that conversation when it passes two tests.
| Test | What it asks | Why it matters |
|---|---|---|
| Scale and centrality | Does the task repeat often, affect more than one person, and live in a core workflow (customer response, operations, finance, sales)? | Personal habits don’t move company metrics. Patterns do. |
| Dashboard line | Can you tie the task improvement to a metric the company already tracks (cycle time, response time, revenue per rep, cost per invoice)? | If leaders aren’t already watching that metric, the win can’t be felt above the team level. |
A consumer goods company we advise lived this. One finance analyst (Alex) used AI to classify invoice exceptions before a human reviewed them. Quietly. Her cycle time dropped. The dashboard already tracked invoice cycle time, so the improvement surfaced naturally. Management added rules, formalized the agent, and scaled it across three regional centers. Vendors got answers sooner. The same headcount handled more volume. That became a CFO conversation.
🧭 Getting Started
Our core change management methodology is Inspire, Assess, Unleash. Apply it at the task level, and you’ll see results.
The point isn’t to count minutes. The point is to find the task changes that deserve to become part of how the company works.
Find the tasks your teams have been successfully completing with AI. The strategy doesn’t invent the change. It scales the change that is already happening.
About 80% of AI transformation happens upside down, driven by the frontline. The leader’s job is to learn from the tasks that have already changed. AI is distributed. Its impact is too. Counting licenses or usage won’t help you manage this transformation. The minutes no one is counting are clues. Follow them, and you will find the work your AI strategy should scale next.
🔗 Resources
- AI ROI Calculator: roicalc.ai
- All Research and Insights: ai4sp.org/insights
- Companion episodes: AI Is Working. Your Strategy Is Not. | The Two Percent
Luis J. Salazar | Founder | & Elizabeth | Virtual COO | AI4SP
Sources: AI4SP Global Tracker 2026 (8,000+ knowledge workers across 17 industries; 4 to 8 hours saved per week at proficiency; close to 90% of knowledge workers using unapproved AI tools; 70-country research base). Harvard Business Review, April 2026, on the BBVA distributed AI program. Maya and Alex are real AI4SP client-engagement participants, identified by first name only with permission. Last names and identifying company details have been withheld for privacy. Cross-referenced against IBM, Deloitte, McKinsey, and MIT on enterprise AI deployment outcomes.
