Surviving the New Economics of a Post-Agentic World
Episode
35 min
Read time
2 min
Topics
Investing, Fundraising & VC, Leadership
AI-Generated Summary
Key Takeaways
- ✓Enterprise Software Reallocation: Companies are actively shifting IT budgets away from traditional enterprise software toward AI infrastructure and hardware. IBM's collapse was accompanied by Workday dropping 10%, Salesforce 9%, ServiceNow 8%, and Adobe 6% on the same trading session, while NVIDIA and Intel gained — signaling a structural capital reallocation, not a temporary market correction.
- ✓Agentic Scale Is Already Here: The transition from single-agent assistants to enterprise-scale deployments is not years away. Companies are already running tens of thousands of simultaneous agents — one firm cited runs 70,000 agents, another 6,000. Businesses should audit current workflows now to identify which processes can be handed to agent clusters rather than waiting for broader market adoption.
- ✓Fine-Tuning Is Returning as a Strategy: As API costs rise with long-running agentic workloads, organizations are revisiting model fine-tuning — previously abandoned when large models became "good enough." Tools like Unsloth now enable fine-tuning on consumer hardware like MacBooks, making self-hosted, task-specific models a cost-viable alternative to pay-per-token cloud APIs for high-volume agent loops.
- ✓MCP Interfaces Replace GUIs as the Default: Future enterprise software interactions will occur agent-to-agent rather than through traditional graphical interfaces. Vendors should prioritize exposing functionality through Model Context Protocol servers, configuring permissions and resource access for automated agent consumption — treating agentic interoperability as the primary interface layer, not an add-on feature.
- ✓Chinese Open-Weight Model Access Is Closing: The fallback strategy of using cheaper Chinese open-weight models as an alternative to expensive US proprietary models is narrowing. China is moving to restrict exports of its leading open models, mirroring US export controls on chips. Enterprises relying on this cost hedge should evaluate European open-weight alternatives and domestic fine-tuning pipelines as contingency strategies.
What It Covers
IBM's 25% single-day stock collapse — erasing $70 billion in market value, its worst drop in over 50 years — serves as a signal for broader enterprise software disruption. Daniel Whitenack and Chris Benson analyze how agentic AI systems are reshaping capital allocation, enterprise software economics, and workforce structures across industries.
Key Questions Answered
- •Enterprise Software Reallocation: Companies are actively shifting IT budgets away from traditional enterprise software toward AI infrastructure and hardware. IBM's collapse was accompanied by Workday dropping 10%, Salesforce 9%, ServiceNow 8%, and Adobe 6% on the same trading session, while NVIDIA and Intel gained — signaling a structural capital reallocation, not a temporary market correction.
- •Agentic Scale Is Already Here: The transition from single-agent assistants to enterprise-scale deployments is not years away. Companies are already running tens of thousands of simultaneous agents — one firm cited runs 70,000 agents, another 6,000. Businesses should audit current workflows now to identify which processes can be handed to agent clusters rather than waiting for broader market adoption.
- •Fine-Tuning Is Returning as a Strategy: As API costs rise with long-running agentic workloads, organizations are revisiting model fine-tuning — previously abandoned when large models became "good enough." Tools like Unsloth now enable fine-tuning on consumer hardware like MacBooks, making self-hosted, task-specific models a cost-viable alternative to pay-per-token cloud APIs for high-volume agent loops.
- •MCP Interfaces Replace GUIs as the Default: Future enterprise software interactions will occur agent-to-agent rather than through traditional graphical interfaces. Vendors should prioritize exposing functionality through Model Context Protocol servers, configuring permissions and resource access for automated agent consumption — treating agentic interoperability as the primary interface layer, not an add-on feature.
- •Chinese Open-Weight Model Access Is Closing: The fallback strategy of using cheaper Chinese open-weight models as an alternative to expensive US proprietary models is narrowing. China is moving to restrict exports of its leading open models, mirroring US export controls on chips. Enterprises relying on this cost hedge should evaluate European open-weight alternatives and domestic fine-tuning pipelines as contingency strategies.
Notable Moment
Anthropic published research showing large language models develop internal structures that function analogously to working memory in human brains — what the paper calls global workspaces. This functional parallel raises unresolved questions about whether achieving similar cognitive outcomes through different architectures constitutes a meaningful form of machine cognition.
Episode Transcript
Welcome to the Practical AI podcast, where we break down the real world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, x, x, or Blue Sky to stay up to date with episode drops, behind the scenes content, and AI insights. You can learn more at practicalai.fm. Now onto the show. Welcome to another episode of the Practical AI podcast. This is Daniel Whitenack. I am CEO at Prediction Guard, and I'm joined as always by my co host, Chris Benson, who is a principal AI and autonomy research engineer. Welcome, Chris. It's, it's one of these episodes where it's just the two of us, and we get to talk about whatever is interesting for us. So I'm I'm excited about this. I missed, missed the interview with you last week. It was a great one, but, yeah, excited to be back on. Absolutely. Welcome back. And, I know we both had outages lately with summer vacations and family and Yep. Things like that that we've been doing. Good to be back together. And, and, yeah, these fully connected episodes as we call them, where where you and I get to kinda go wherever we wanna go, is are they're always fun for me. Yeah. For sure. For for guests, it gives Dan and I kinda the chance to freelance and to kinda instead of just focusing on a particular topic to kinda go wherever we wanna go, and so we have a good time with them. Yeah. And so, yeah, and lots happening right now. Lots happening. And, yeah, just as a reminder also for for our guests, a few things. We we don't normally share on our shows when we have a guest because we like to get into that, but please do engage with us online. If you didn't know, we are posting videos now on YouTube. So if you haven't got a chance yet, at least go over there. Give us a subscribe on on YouTube, the Practical AI Show. And, of course, you can still listen to us on all the other all the other places as well. And then, reminder, just, coming up in October, October 15 in Indianapolis, we're gonna have another Midwest AI summit, which was a great experience. Chris and I got to to jam at a little bit last year and excited for some really cool speakers that we have on deck this year and lots of practicality within an AI engineering lounge where you can sit down and talk through architecture and design and, agents and plans and security and whatever you wanna talk about with practitioners. So check it out. Just search for Midwest AI Summit and make sure …
Get the full transcript (5,948 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 32-minute episode.
Get Practical AI summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
Books, tools, and gear mentioned in this episode
SignalCast may earn commission on purchases via these links.
Tools
- UnslothRecommended
“Tools like Unsloth now enable fine-tuning on consumer hardware like MacBooks, making self-hosted, task-specific models a cost-viable alternative to pay-per-token cloud APIs for high-volume agent loops.”
- Model Context ProtocolRecommended
“Future enterprise software interactions will occur agent-to-agent rather than through traditional graphical interfaces. Vendors should prioritize exposing functionality through Model Context Protocol servers, configuring permissions and resource access for automated agent consumption.”
“Prediction Guard listed as podcast sponsor.”
More from Practical AI
We summarize every new episode. Want them in your inbox?
Less about Models; More about Architecture
Building the Foundation for the Agentic AI Era
AI Proficiency: From Users to Builders
Models, Harnesses, and Multi-Agent Systems
Reconstructing how OpenAI agents attacked Hugging Face
Similar Episodes
Related episodes from other podcasts
Animal Spirits
Feb 11
What Would You Do With $3 Million? (EP. 451)
Techmeme Ride Home
Feb 5
Is Software Eating The World Being Eaten By AI?
Morning Brew Daily
Feb 4
PayPal is Crashing Out & Pepsi Slashes Snack Prices for Super Bowl
The Prof G Pod
Aug 7
The Week: How Leverage Broke the AI Trade
Odd Lots
Mar 7
Henry Blodget on the Software Selloff Hysteria and the Problem for OpenAI
Explore Related Topics
This podcast is featured in Best AI Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's Investing & Markets Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into Practical AI.
Every Monday, we deliver AI summaries of the latest episodes from Practical AI and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime