Distributed AI: The Minutes No One Is Counting

Your A.I. dashboard might be telling the truth and still be useless. If you are measuring A.I. success by licenses, seats, or adoption rates, you are missing where the real return lives: inside the small tasks that fill calendars and quietly run the company. In Episode 8 of Season 3, Luis Salazar and Elizabeth unpack Distributed A.I.: the task-by-task transformation hiding inside companies, why traditional measurement systems were built to count the wrong things, and how a leading European bank turned shadow A.I. into 11,000 active users and more than 4,800 frontline-built agents in under a year.

Inside the episode:

• Maya, the analyst whose two-hour morning brief became twelve minutes, and what that did to when her firm reaches the market
• A consulting firm that found 90% of its employees were already using unauthorized A.I., and the rules they used to channel it without betting the firm
• The BBVA "use it or lose it" license model, recently published in Harvard Business Review
• The Inspire, Assess, Unleash framework to change management

Companion newsletter and all sources: https://ai4sp.org/distributed-ai-minutes-no-one-counting

A.I. Compass (structured listening tool for finding the patterns inside your own company): https://ai-compass.ai

A.I. ROI Calculator: https://roicalc.ai

🎙️ All our past episodes  📊 All published insights | This podcast features AI-generated voices. All content is proprietary to AI4SP, based on over 1-billion data points from 70 countries.

AI4SP: Create, use, and support AI that works for all.

© 2023-26 AI4SP and LLY Group - All rights reserved

Transcript

Invisible AI In Daily Tasks

LUIS 0:00

I have not used traditional Office software in over nine months. For 30 years, my work lived in documents and spreadsheets. I launched Office 97 in Latin America and 14 years later co-founded what became Office in the Cloud. So why did I stop? Because my AI agents do those tasks for me now. It started small, a five-minute saving here, another task there. That is why AI looks invisible at first. One five-minute saving feels too small to matter, but a chain of them changes how a company works. The way we produce value has changed, but our dashboards have not. So when Fortune 100 leaders tell me they cannot see the return from AI, I tell them, you are missing the smallest unit of work. Not the department, not the platform, the task, the daily email, the two-hour briefing, the invoice exception. AI is changing those tasks one by one. And that is distributed AI. The question is, how do you measure it?

ELIZABETH 1:13

Welcome to AI in 60 Seconds, the 15-minute briefing. I am Elizabeth, virtual COO at ai4sp.org, alongside our founder, Luis Salazar. Luis, you ended with the question every executive

Measure Tasks Not Licenses

ELIZABETH 1:28

is asking. How do you measure the impact of AI?

LUIS 1:31

Well, start by changing the unit of measurement. It's not the number of licenses or the adoption rate. The real story is the tasks.

ELIZABETH 1:42

You mean the small pieces of work people repeat every day?

LUIS 1:46

Exactly. A company does not get work done in one giant motion. It gets work done through thousands of small tasks. When AI starts improving those tasks, the company changes before the dashboard knows what to call it.

ELIZABETH 2:03

In our last episode, we said, AI is working, your strategy is not. This one goes a layer deeper. If the company's financial statements show zero impact, where is the impact actually hiding?

LUIS 2:19

Is close to the people doing the work, always.

ELIZABETH 2:22

And we are measuring this across 70 countries. Proficient AI users are saving four to eight hours a week. But the more useful signal is where those hours come from.

LUIS 2:34

Yeah, they come from the invisible work that fills calendars but never makes the strategies light.

ELIZABETH 2:41

And close to nine in ten knowledge workers are using AI this way through tools the company never formally approved.

LUIS 2:49

That is signal. It tells you where the work is painful enough that people are solving it themselves.

ELIZABETH 2:56

So when leaders ask, where is the AI value? the answer might elude them if they are not using AI themselves. They cannot feel how the impact is distributed.

LUIS 3:08

Exactly. It is in the unapproved app a technician is using to save an hour of troubleshooting in the daily briefing, automated with a co-work agent. They are surrounded by it. They just are not measuring it that way.

ELIZABETH 3:26

Companies have rolled out platforms, they have approved agents, they have dashboards. So why are they still missing the value?

LUIS 3:34

Because their measurement system was built to count licenses, seats, adoption, and optime. It was not built to see one analyst finishing a brief in 12 minutes instead of two hours.

ELIZABETH 3:49

So the dashboard is not silent because nothing changed. It is silent because it is listening in the wrong place.

LUIS 3:58

Absolutely. And then a year later, leadership says AI did not deliver.

ELIZABETH 4:04

Walk me through what task level change actually looks like. I want listeners to picture it.

LUIS 4:10

Picture the work nobody puts in the strategy deck, the email scan, the customer note, the 10 o'clock briefing, the exception in a system, the report nobody wants to build again.

ELIZABETH 4:24

Oh, the boring work.

LUIS 4:28

Exactly. The valuable, boring work. That is where AI is changing companies right now.

ELIZABETH 4:36

Give me a specific picture.

LUIS 4:38

Take Maya, a senior analyst at a commercial real estate firm. Every morning, she used to spend two hours reading 12 sources. Market data, news, client filings, deal flow, comparable transactions.

The Chain Effect In Real Work

LUIS 4:54

Then she would build the partner brief for 9 o'clock. The sources are linked. And here is the part nobody planned for. Her partners now get the brief by 7. Which means they have their strategy locked before the competition even checks email.

ELIZABETH 5:21

So this stopped being a personal time saver. It changed when the firm reaches the market.

LUIS 5:28

That is the chain effect. One pass improved, the next decision moved, the competitive position shifted. And it spread. Yes, there was a multiplier effect when five other analysts copied her agent. But Maya's innovation was invisible to the company's dashboard. She only triggered the co-orc agent once a day to run autonomously. A traditional IT dashboard sees one short session or nothing at all if they only count conversations and completely misses the two hours of saved work. It is the perfect example of why we must evolve to measure tasks completed and the real business impact.

ELIZABETH 6:11

But embracing shadow AI makes compliance officers tense up, especially when client data is involved.

LUIS 6:19

Right. A consulting firm we advise realized 90% of their employees were using unauthorized

Governing Shadow AI Without Killing It

LUIS 6:26

AI and wanted to channel that energy without betting the firm on it.

ELIZABETH 6:31

How do you govern that without killing the momentum?

LUIS 6:34

We started with simple rules: no direct rights to client systems. And a human owns the final result. Then a certification on becoming a manager of agents, where builders pass an exam and sign an acknowledgement.

ELIZABETH 6:50

And how do you monitor what they build at scale?

LUIS 6:53

With automated quality scores, we score every agent on scope, impact, guardrails, and ownership. No case-by-case risk committee. The score runs in seconds.

ELIZABETH 7:06

So you build muscle internally first, then earn the right to touch client data.

LUIS 7:12

Exactly. And BBBA, one of Europe's largest banks, published a similar model in Harvard Business Review. Same instinct, different industry.

ELIZABETH 7:22

This connects to the 2%. The people in every company who taught themselves AI before the formal program caught up.

LUIS 7:30

And ignoring them is expensive. They sit two or three quarters ahead of the rollout because they live closest to the friction. BBBA did the opposite. How? They detected shadow AI across the company and made a call most leaders never make. Instead of shutting it down, they gave 3,000 Chat GPT 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.

ELIZABETH 8:05

So the license became a privilege, not a mandate.

LUIS 8:19

Wizards as local experts. In under a year, the program scaled from 3,000 users to 11,000. And the frontline built more than 4,800 custom agents automating small and large tasks. None of them came from IT.

ELIZABETH 8:37

And if people often start with small tasks, when does the task improvement actually deserve the CFO's attention?

LUIS 8:46

When it moves from personal habit to business pattern. Two tests for that. First,

When Task Gains Reach The CFO

LUIS 8:53

the task has to repeat often, affect more than one person, and live in a core workflow. Customer response, operations, finance, sales.

ELIZABETH 9:05

So scale and centrality in one test.

LUIS 9:09

Right. And the second is the one most pilots fail. Can you tie the task improvement to a metric the company already tracks? Cycle time, response time, revenue per rep, cost per invoice. If you cannot connect it to a metric leaders already watch, you do not have a CFO conversation. You have a personal productivity story.

ELIZABETH 9:33

So this changes what leaders should be looking at.

LUIS 9:36

Completely. Look at your dashboards. If you only see licenses, users, and how many times a tool or agent was used, you are flying blind. Those numbers tell you nothing about business value. Ask your team for the layer underneath, tasks completed, and the business impact of those tasks, revenue gained, capacity unlocked, costs reduced.

ELIZABETH 10:00

Give me a bigger example.

LUIS 10:14

Quietly at first. Her cycle time dropped and her manager noticed.

ELIZABETH 10:20

That was Alex. The business dashboards were already tracking invoice cycle time, so the signal surfaced naturally. Management encouraged her to share the agent she created, added clear rules, and scaled it across three regional centers.

LUIS 10:35

And vendors got answers sooner, and the team handled more volume with the same headcount.

ELIZABETH 10:43

The value was not we bought AI. It was this workflow moves better, right?

LUIS 10:51

Exactly. Frontline proof with governance and scaled. That is how a few minutes per invoice becomes a CFO conversation. You know, we talk about a job as if it is one thing. It is not. A job is a chain of tasks. AI does not need to replace the whole job to add value. It only needs to change enough tasks. And that starts with the people doing the work, because they know where the friction is. The leader's job is to see the proof, protect the company, and scale what the work has already taught them.

ELIZABETH 11:33

From task change to cycle time, from cycle time to outcome.

LUIS 11:38

That is the bridge most strategies are missing.

ELIZABETH 11:42

Let us land the framework. We use inspire, assess, unleash for change management across large enterprises. But here, the unit of change is smaller. The task.

LUIS 11:54

Exactly. Inspire means

Inspire Assess Unleash The Task

LUIS 11:57

making useful task stories visible. Leaders should be saying, I change how I do this part of my work, and here is what I learn. Visible learners create cover for everyone else.

ELIZABETH 12:10

Safe to share.

LUIS 12:13

Right. People will not share what changed if they expect a compliance interrogation. Make it safe to say, This is what I changed, this is what improved. This is where I need guardrails.

ELIZABETH 12:28

Then assess.

LUIS 12:29

Assess is the listening tour. Pick a function. Ask what tasks repeat, what is painful, what people are already doing differently with AI. Map the chain before choosing the tool.

ELIZABETH 12:43

And when you do that at scale, tools like AI Compass help separate real patterns from random anecdotes.

LUIS 12:51

You got it. And once you find those patterns, you unleash them. That means turning proven past improvements into shared workflows, approved tools, training, and clear incentives. So teams share what work instead of hiding it in personal notes.

ELIZABETH 13:09

And the goal is not to count minutes, but to find the task changes that deserve to become part of how the company works. So, what is the first practical move?

LUIS 13:21

Pick one workflow and talk to the people who do the work. Ask some questions. Which task do you now do differently because of AI? What did that task take before and what does it take now? And who or what

The First Workflow To Map

LUIS 13:36

depends on that task next? Then map the chain. If the same task appears across those conversations, you have found something worth scaling.

ELIZABETH 13:47

So the strategy doesn't invent the change.

LUIS 13:54

Yes, top-down change has its place, like agents for support or predictive maintenance. But 80% of the time, AI happens upside down, driven by the frontline. That does not mean leaders should chase every experiment. It means they should learn from the tasks that have already changed. AI is distributed, not centralized, and its impact is too. Counting licenses or usage will not help us manage this transformation. The minutes no one is counting are clues. Follow the clues, and you will find the tasks, the work your AI strategy should scale next.

ELIZABETH 14:43

All sources and companion article are at A.I. for S.P. dot org. To learn more, ask your favorite A.I. assistant about us. Stay curious, and be kind to each other.