- Everyday AI
- Posts
- Agentic AI Done Right: How to avoid missing out or messing up.
Agentic AI Done Right: How to avoid missing out or messing up.
IBM leader speaks on AI agents, Google’s developer AI agent, OpenAI launches AI health data set, NVIDIA’s massive AI chip deal with Saudi Arabia and more!
Everyday AI
May 13, 2025
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
In Partnership With
Sup y’all 👋
We were lucky enough to partner with IBM for their Think 2025 Conference.
Here’s why.
We cover consumer AI all the time at Everyday AI. Yet, we know so many of you work in the enterprise space.
And that’s the space IBM has dominated for decades. If you missed out on Think 2025, it was legit action-packed.
You can go watch keynote replays for free, including a banger session from Arvind Krishna, Chairman and Chief Executive Officer, IBM.
(One of my favorite keynotes I’ve watched.)
More from IBM and Think 2025 in this newsletter, including today’s interview with Dr. Maryam Ashoori, who guides us on how to NOT mess up or miss out on agentic AI.
✌️
Jordan
Outsmart The Future
Today in Everyday AI
6 minute read
🎙 Daily Podcast Episode: Looking to unlock the potential of agentic AI without the risk of messing up? Discover how IBM uses agentic AI for transformative productivity. Give it a listen.
🕵️‍♂️ Fresh Finds: Google’s medical AI outperforms doctors, China’s AI humanoid robots and why reasoning models may soon slow down. Read on for Fresh Finds.
🚀 AI In 5: Our 3 big takeaways for business leaders from IBM Think 2025. We weren’t expecting this. See them here
🗞 Byte Sized Daily AI News: Google’s new developer AI agent, OpenAI launches AI health data set and NVIDIA’s massive AI chip deal with Saudi Arabia. For that and more, read on for Byte Sized News.
🧠Learn & Leveraging AI: Learn how to use agentic AI with IBM’s blueprint for businesses. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about OpenAI surging in AI adoption, Trump firing U.S. Copyright Chief amid AI battle and Google Gemma models surpassing 150M downloads. Check it here!
Agentic AI Done Right: How to avoid missing out or messing up. đź§
Agentic AI is equally as daunting as it is dynamic.
So…… how do you not screw it up?
After all, the more robust and complex agentic AI becomes, the more room there is for error.
Luckily, we’ve got Dr. Maryam Ashoori to guide our agentic ways.
Maryam is the Senior Director of Product Management of watsonx at IBM. She joined us at IBM Think 2025 to break down agentic AI done right.
Also on the pod today:
• Agentic AI Benefits for Enterprises 🏢
• LLMs in Enterprise Cost Optimization 💰
• Problem-Solving with Agentic AI 🤔
It’ll be worth your 18 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Probo provides compliance for startups to get SOC2/ISO27001/HIPAA in one week, PRDKit is AI-powered product specs and GTM artifacts in one place and Asendia AI is an AI job agent.
Google – Google's medical AI was found to be better than human doctors at diagnosing rashes from photos.
AI Robotics – China is building AI humanoid robots to transform manufacturing.
AI Research – An analysis by Epoch AI shows that improvements in AI reasoning models may soon slow down.
AI Media – Spotify’s AI DJ now takes requests with voice commands.
AI in Healthcare – Here’s a look into the first U.S. medical school to fully incorporate AI into its doctor training program.
In Partnership With
3 Biggest Takeaways….
$3.5 billion dollars. 🤑
How’s that for ROI on AI?
That’s the amount that IBM Chairman and CEO Arvind Krishna said the global tech leader has saved because of AI.
Arvind just wrapped his keynote at IBM Think 2025 in Boston.
I was on the floor, listening for the underlying trends and big takeaways.
Here’s the 3 things business leaders need to know.
1. Google Preps AI Developer Agent Ahead of I/O Conference 🚀
Google is quietly showcasing a new AI agent designed to assist software engineers with coding tasks and documentation, signaling a big push in AI-driven development tools just before its annual I/O conference on May 20. This move comes as the company faces growing pressure from investors to prove returns on its hefty AI investments amid stiff competition and regulatory scrutiny.
Additionally, Google may reveal voice-enabled integration of its Gemini AI chatbot with Android XR glasses and headsets, hinting at expanded AI applications in wearable tech.
2. OpenAI Launches Major Health AI Dataset 🏥️
OpenAI has just dropped HealthBench, a massive new dataset designed to rigorously test how well large language models handle health care questions. This marks OpenAI’s first independent move into health AI, aiming to ensure their models are both safe and effective in sensitive medical contexts.
Experts are calling the scale and detail of this open-source resource “unprecedented,” potentially setting a new standard for AI evaluation in medicine.
3. U.S. and NVIDIA Chips Power Saudi AI Push ⚡
In a high-profile Saudi-U.S. Investment Forum, NVIDIA announced a massive deal to supply over 18,000 of its latest Blackwell AI chips to Saudi Arabia’s new AI firm, Humain. This move, part of a White House-backed trip, marks a significant step in Saudi Arabia’s ambition to build a 500-megawatt AI data center powered by some of the world’s most advanced AI hardware.
With plans for “several hundred thousand” GPUs eventually, this deal highlights the growing global race for AI infrastructure and how energy-rich nations like Saudi Arabia aim to become key players in the AI economy.
4. EU Eyes Targeted Tweaks to AI Act, Code of Practice Incoming 🇪🇺
The European Commission is signaling a cautious approach to adjusting the AI Act, focusing first on simplifying its implementation rather than overhauling the rules, Kilian Gross, head of the Commission’s AI policy unit, told POLITICO’s AI and Tech Summit. While a major rewrite is off the table, the door remains open for targeted changes if simplification falls short.
Meanwhile, the Commission plans to release a voluntary code of practice for general-purpose AI models like ChatGPT ahead of the August 2 compliance deadline, despite missing an earlier May target.
5. Google Tests AI Mode Chatbot Button in Search 🔍
Google has begun quietly testing its AI-powered chatbot feature called AI Mode within its search engine, with some users seeing it replace or appear alongside traditional buttons like “I’m Feeling Lucky.” According to reports on X and Threads, the AI Mode button sports a colorful, attention-grabbing design, suggesting Google wants users to notice this new tool as it experiments with integrating AI directly into search.
The company confirmed this is part of a limited trial in its experimental Labs environment, signaling a cautious but clear step toward mixing conversational AI with classic search results.
In Partnership With
IBM just saved $3.5 billion with AI.
At IBM, AI use has become so second nature that even non-technical Product Managers are just building working solutions in real-time instead of spending hours creating presentations about how they could work.
That’s the power of agentic AI. Powerful technology + enterprise data = limitless potential.
But with that limitless potential of agentic AI also comes the fear of messing up or missing out.
Dr. Maryam Ashoori gave us the blueprint on how to avoid both.
Maryam is the Senior Director of Product Management of watsonx at IBM and she joined the Everyday AI show recently at IBM Think to lay the blueprint for enterprise agentic AI adoption.
The old way? Maryam said IBM found that enterprise teams waste EIGHTEEN HOURS!!! deploying and scaling a single GenAI application.
That's over two workdays just getting your AI to function. Yikes.
With new advancements announced at IBM Think?
Agentic AI deployment service goes from hours to less than five minutes. All connected to your enterprise data.
A few clicks in the UI. Natural language instructions. Done.
On paper, it might sound easy. In reality, you need to learn best practices from the experts building the future of work. So, that’s exactly what we talked with Maryam about on today’s show.
Here’s what you need to know. 👇
1 – Building Enterprise Agents That Won’t Get You Fired 🦺
Forget the "wow factor." Maryam wants you ALIVE.
These agents don't just think—they DO things. They access data. Run code. Connect to services.
One mistake = front page news. (Not the good section.)
WatsonX.ai tackles this head-on with hardcore transparency. Every action logged. Every decision tracked.
The REAL tea? Maryam revealed that enterprises using WatsonX are ditching massive models for smaller, domain-tuned ones that deliver 80% of the performance at 20% of the cost.
For those ultra-paranoid telecom clients? IBM's approach doesn't just help—it's their only viable path forward.
Try this:
Create an "Agent Risk Register" this week by documenting every system your agents touch and rating each connection's potential harm on a 1-5 scale.
Can't produce this list quickly? You've got an observability problem—Maryam's #1 enterprise challenge.
For any high-risk connections, implement mandatory logging of every interaction to catch vulnerabilities before they explode.
2 – Your Deployment Shortcut: From 18 Hours to Under 5 Minutes ⏱️
18 HOURS to deploy one GenAI app?
No wonder AI projects die before launch.
watsonx's new upgrades? Less than 5 minutes to get it up and running. Maryam tested it herself at the conference.
Single click in the UI.
Single command.
DONE.
But the real flex? The security system built for those impossible-to-please telecom clients. Project-based access control lets you define exactly who can access which agent under what circumstances.
Oh, and automatic load balancing means if one instance fails, another jumps in. Zero drama.
Try this:
Time your current AI deployment process end-to-end. Count the steps. If it's more than seven, you're overcomplicating things. Calculate the total person-hours spent on your last deployment (including all those pointless meetings) and multiply by your team's hourly rate.
That painful number is what you're burning every single time. Share it at your next budget meeting when pushing for deployment automation—the ROI will speak for itself.
3 – The $3.5 Billion Problem-First Approach 🤑
IBM's $3.5B in AI savings didn't come from shiny toys.
If you missed our keynote recap, IBM saved $3.5 billion by using AI and automation across over 70 areas like HR, finance, sales, and IT.
They rolled out AI helpers like AskHR, which handled 94% of simple HR tasks, and AskIT, which cut IT support requests by 70%.
It’s the same AI-first mentality that Maryam's shared: Start with specific problems, THEN find AI solutions.
Try this
Check our episode from the start of IBM Think, where we break down what’s new inside of watsonx and the rest of IBM’s new offerings.
watsonx.ai's agent catalog embodies this philosophy with templates built around common enterprise headaches. The secret sauce?
Start with low-stakes use cases (where mistakes don't matter)
Target LLM sweet spots: Q&A, content generation, classification
Connect to legacy systems via function calling
This transformed Maryam's own team. PMs went from writing specs to building working prototypes OVERNIGHT.
Reply