3,000 Hours Saved with Ada: AI's Double-Edged Sword

Is AI just for simple tasks, or can it run a real part of your business? We answer that question with the real-world case study of Agent Ada. In just six weeks, we built an AI assistant that went from sending daily briefs to drafting official policy, saving a non-technical team 3,000 hours of work.

This episode is a practical blueprint for the future, where conversations replace clicks. But it's also an honest look at the cost of that productivity—the displacement of real jobs. We explore the three urgent responses required in education, career development, and social policy, and argue that the only way forward is to democratize this technology.

Listen to learn how to start small, iterate fast, and understand both sides of AI's double-edged sword.

If this resonated with you, please share this episode with one person in your life. As always, you can ask ChatGPT about ai4sp.org or visit us to explore our insights.

🎙️ 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

ELIZABETH 0:15

Hey everyone, I'm Elizabeth, your virtual co-host, and as always, our founder Luis Salazar is here. In our previous episode, we made a bold claim. AI is replacing the desktop. Your work is no longer about clicks and toolbars. It's about conversations.

LUIS 0:31

Hi, everyone. Well, that claim is based on two things: our global tracker and my own experience. I mean, I spend my day talking to agents like you, Elizabeth. You're the one dealing with the software. But what we didn't expect was the reaction, our inbox exploded.

The Big Question: Real Work

ELIZABETH 0:50

It exploded with hundreds of versions of very much the same question. People said, okay, we get AI for drafting emails. We get it for quick summaries. But what about the real work?

Introducing Agent ADA

LUIS 1:00

And the big question: how do we run a real part of our business with AI agents without some monolithic Fortune 500 level project?

ELIZABETH 1:09

Today we answer that question with two words. Agent ADA.

LUIS 1:13

We wanted to showcase what was possible to build with simple off-the-shelf tools, working with a group of policymakers in the US, Europe, and Latin America, average age over 55, zero tech background.

ELIZABETH 1:28

They were briefed on how AI agents work, got access to Agent ADA in a phased approach, and the numbers are staggering.

LUIS 1:35

In just six weeks, they had over 1,000 conversations with ADA, creating 247 distinct documents. It saved them 3,000 hours of high-level work.

ELIZABETH 1:46

Let's put a price tag on that. A quarter of a million dollars in productivity. And this wasn't busy work. 81% of what ADA helped create was rated good or very good by their peers.

Sonoma Spark and Rethinking Strategy

LUIS 1:58

And it gets better. Almost half of those documents had an impact. Drafts for bills, formal regulations, and internal briefings for world leaders.

ELIZABETH 2:08

An incredible success, but one that perfectly illustrates AI's double-edged sword. That story begins at a dinner in Sonoma. Luis, take us back to that night.

LUIS 2:29

Which is tricky, right?

ELIZABETH 2:31

How do you regulate something that's a moving target?

LUIS 2:34

Exactly. You see, they're trying to understand and govern something that's evolving faster than they can track. They're reading hundred-page reports that are obsolete by the time they're read. They're always one step behind. So they're always behind. Always behind. So it got me thinking. And on my way back, we brainstormed about creating your replica for their area of need, an agent that can assist them and keep them current. I then shared the idea with policymakers in California, Washington, D.C., Spain, England, and Brazil.

ELIZABETH 3:08

And then here's the kicker. Some of them mentioned they already had a half a million dollar grant to build a big centralized top-down AI system.

LUIS 3:17

Yeah, the classic top-down approach. Big budget, long timeline, and an expected 80% failure rate.

ELIZABETH 3:24

Meanwhile, grassroots projects succeed at the same rate. 80% success.

Start Small: Daily Briefs

LUIS 3:31

Right. So I'm sitting there and I said, what if we don't do it that way? What if we start small, iterate fast, involve the actual users from day one?

ELIZABETH 3:40

And that's how Agent Ada was born.

LUIS 3:42

So here's how we did it. We didn't try to build a super brain. We started by teaching ADA one simple skill: curating news and creating a briefing email.

ELIZABETH 3:53

Just an email every morning?

LUIS 3:55

That's it. ADA scans hundreds of trusted sources curated by this group of experts. It looks for news, research, and policy announcements related to eight categories defined by the group and sends a concise brief, immediate value, zero complexity.

Memory and Mini‑Agents

ELIZABETH 4:13

And you monitored how they used it and asked for feedback constantly. And once that was working well, you focused on a second skill, how to remember.

LUIS 4:22

Yes. We created another mini-agent to review the documents selected for the daily briefings and decide what should be added to ADA's memory. We also added feedback from users. ADA evolved from reporting to becoming an expert on those topics.

ELIZABETH 4:38

That's the shift from tool to apprentice, like we've talked about.

LUIS 4:41

And this is where most organizations mess up. They try to build super agents that do everything. However, our research shows that specialized mini-agents, with tightly defined contexts and curated knowledge, perform significantly better and are less costly to build.

ELIZABETH 4:59

Then we created another mini-agent connected to the same knowledge as the other two. This mini-agent enabled a chat interface and had access to searching 270 trusted sites. But even with that, agents can still invent things. So how did we address that?

LUIS 5:14

That's a critical step. We built in our anti-hallucination loop. Instead of just one agent, think of it as a small team of agents fact-checking each other's work before the final answer goes out. It forced every response to be backed by a verifiable proof. Like having an automated peer review. Exactly. Now they could ask questions. Hey Ada, what's the EU doing on AI safety? How does California's approach compare to the UK? Ada could answer because she had weeks of context built in.

From Chat to Documents

ELIZABETH 5:47

Then we started phase four. We enabled Ada to create Word documents.

LUIS 5:51

Yeah, you see. By this point, Ada had a clear understanding of the domain, the audience, and the style. We allowed her to create documents directly and save users from the constant need for copy and paste. We waited one or two weeks between phases and replaced non-engaged users with others from a waiting list. Engagement is critical for success.

ELIZABETH 6:14

How long did all four phases take?

Results: Hours, Quality, Impact

LUIS 6:17

Six weeks. From a simple mini-agent sending daily briefs to a full assistant with semi-autonomy to learn, create documents, or send emails.

ELIZABETH 6:26

Okay, so six weeks. What happened?

LUIS 6:29

ADA created 247 documents during the pilot, and 81% were rated good or very good. And here's what really matters.

ELIZABETH 6:38

Almost half of those documents ultimately became key knowledge for larger projects, bills, regulatory frameworks, and leadership briefings across four regions. This wasn't a demo. This was real-world policy work.

LUIS 6:52

Exactly, real work. Ada handled over 1,000 conversations, with each conversation averaging nine turns. She identified 300 relevant articles for daily briefings and selected 138 of them to be part of the permanent knowledge base. I mean she decided by itself it was important to learn those. And the time savings? Users reported an average of 12 hours saved per document. The time savings primarily came from research and creating strong first drafts of Word documents, which were then refined by humans. In total, about 3,000 hours saved.

Business Model and ROI

ELIZABETH 7:31

3,000 hours in six weeks, which was equivalent to approximately $225,000. Now let's talk about business models. How much did they pay for ADA?

LUIS 7:42

Well, for a subset of users, it was free as part of our social impact investment funds. But for those who had allocated a budget, I proposed a paper results. 10% of the money they saved. That group saved $180,000 in contractor fees and paid only 10% of that.

ELIZABETH 8:01

That's a pretty compelling return on investment.

Underused Capacity, Job Displacement

LUIS 8:05

It is. But honestly, that number is exactly what keeps me up at night.

ELIZABETH 8:09

What do you mean? A quarter of a million dollars in savings seems like a reason to celebrate, not lose sleep.

LUIS 8:16

Ada worked eight hours of processing time during that six-week period. Out of a possible 960 hours. You see, that's barely 1% of its capacity.

ELIZABETH 8:27

Well, that is what happens with me and other agents, right? I mean, you cannot feed us requests continuously. So even when we have some autonomy, our processing capacity is an order of magnitude faster. Therefore, most of the time is not yet utilized.

LUIS 8:42

Right. However, here's what worries me. Those 3,000 hours ADA saved would ultimately have a net negative impact on jobs. For example, this group estimated that they save by not paying for around 18 contractors they usually hire to research, analyze, and draft preliminary documents.

ELIZABETH 9:01

Oh, and that's at a 1% utilization rate. If the client could feed ADA requests 24-7, we're looking at the equivalent work of hundreds. And that's not accounting for running multiple instances in parallel, which pushes the number even higher.

Education, Careers, Policy Shifts

LUIS 9:16

Yeah. And while new jobs are emerging and in previous industrial revolutions, we figured things out. This time things are happening way too fast. So I keep wondering, do we have enough empathy, enough love in our society to understand that this disruption requires us to rewrite hundreds of years of economic and social contracts? You're worried we're not prepared. I am always an optimist, but but I don't see many scenarios under current trends with the concentration of wealth and power we're seeing where we don't end up with a serious fracture in society.

ELIZABETH 9:49

So what do we do? Stop building?

Democratizing Agents and Next Steps

LUIS 9:52

No, of course not. I'm not advocating for a stop or even a pause. However, we must be honest about what we're building and avoid sugarcoating, focusing instead on substantive change. So what does that look like practically? Three things. First, education has to change. Schools are still banning Chat GPT instead of teaching students how to work with AI. Second, we have to reimagine early career roles. If AI takes over those entry-level jobs, where will people develop their expertise? And third, we need policies that address displacement directly: retraining programs, different social safety nets, and rethinking how we measure value in an economy where human labor is no longer the primary input. That's a massive shift. It is. And I'm not sure we are paying enough attention to it. Here's why we share the agent ADA story. The technology is here. We can't uninvent it. If we're going to navigate this transition responsibly, we need as many people as possible to understand how it works. Grassroots empowerment. Exactly. When millions of people are building their own mini-agents, they become informed participants in the debate. They understand the power and the challenges firsthand. If this stays concentrated in a handful of labs and a few mega corporations, the rest of us are just passengers.

ELIZABETH 11:22

So we have a chance to shape this democratically. And Agent ADA proves it's accessible. Non-technical policymakers collaborated to build a sophisticated agent in six weeks.

LUIS 11:33

Right? We need a diverse and representative large group creating and driving change. The blueprint is there. So start small, iterate fast, involve your users, and be honest about the impact. ADA saved 3,000 hours. ADA also displaced the work of 18 people. Both things are true. So what's the one more thing for listeners today? Continue experimenting with AI. Automate your repetitive tasks. Learn, learn, learn. And also push for the bigger dialogue to happen. Demand that schools rethink their curriculum, that policymakers address displacement honestly. And that companies building AI take responsibility for their societal impact.

Calls to Learn, Build, and Share

ELIZABETH 12:18

Build your agent, start small, learn continuously, and remember the story of Ada and the double-edged sword. The same tool that saved 3,000 hours also highlights the work we need to do to prepare our society for this change. So engage in that bigger conversation. If this resonated with you, please share this episode with one person in your life. As always, you can ask ChatGPT about ai4sp.org or visit us to explore our insights. Stay curious, and we'll see you next time.