Podcast episode
Surviving the New Economics of a Post-Agentic World
agents inference model-pricing open-weights orchestration
TL;DR
Two Practical AI hosts (Daniel Whitenack, CEO at Prediction Guard; Chris Benson, principal AI research engineer) use IBM's single-day 25% stock plunge as a jumping-off point to argue that agentic AI is already disrupting enterprise software economics — not as a future scenario but as a present-tense market event. Useful for tracking the enterprise software disruption narrative and the emerging "agent workforce" infrastructure layer; thin on specific model/lab news.
What was covered
- IBM's $70B single-day market-cap loss (~25% drop, worst in 50+ years, exceeding Black Monday 1987 losses): hosts attribute it to enterprise IT budgets shifting away from traditional software/services toward AI infrastructure and hardware — framed as a "canary in the coal mine" for the broader enterprise software category.
- Sector-wide enterprise software selloff on the same day: Workday –10%, Salesforce –9%, ServiceNow –8%, Adobe –6%; NVIDIA and Intel/chip providers rose as companies "panic-bought" hardware to hedge against future AI compute price hikes.
- Chinese open-weight model access restrictions: China reportedly moving to restrict export of its open-source models (cheaper alternatives to US proprietary APIs), closing off a fallback strategy for non-US companies squeezed by US model pricing. Hosts also discuss a "Chinese token black market" — gray-market resellers providing mainland Chinese users access to OpenAI and Anthropic APIs that are blocked in China.
- Return of fine-tuning as a strategy: As enterprises get squeezed from both directions (expensive US proprietary APIs, closing Chinese open-model access), hosts note renewed executive interest in fine-tuning smaller, self-hosted models — a trend that had faded when large frontier models seemed "good enough."
- Agent-at-scale reality check: Hosts cite two Prediction Guard customers already running 70,000 and ~6,000 agents respectively in production, arguing the "thousands of agents" scenario is not years away.
- Anthropic research paper on "verbalizable representations": Brief discussion of Anthropic's Verbalizable Representations Form a Global Workspace in Language Models — finding that models appear to create functional analogs to human working memory ("workspaces"), with implications for agentic routing and information-reporting capabilities.
Notable claims & predictions
- Chris Benson: "The rate of change is accelerating exponentially right now… IBM is just one significant company — there are lots and lots of companies taking that enterprise software approach and they are all at risk to some degree."
- Daniel Whitenack: "We've talked to one company that already has 70,000 agents running, one that has around 6,000 — this is not five-years future, this is very rapidly what companies are [doing]."
- Chris Benson: "The notion of loop engineering is going to be kind of antiquated by what follows before the year is out" — suggesting the current agent-loop paradigm is already transitional.
- Chris Benson: "The vast majority of interactions across different systems are going to occur agentically without a human directly involved in most of those processes" — positioning agent-to-agent (A2A) as the dominant future interface pattern, displacing GUIs.
- Daniel Whitenack: On the economics shift — moving a long-lived agent from a pay-as-you-go API to a self-hosted small model "is going to be a vastly different economics" — flagging inference cost as the key lever driving architectural change.
- Chris Benson (on the Anthropic paper): Models are "creating workspaces where they're essentially fulfilling the same function as working memory in a human brain" — framing this as functionally relevant to agentic capability even while sidstepping consciousness claims.
Why this matters for AI operators
- Inference economics are forcing architectural decisions now: The hosts' point that long-running, multi-agent loops make pay-as-you-go API pricing untenable is a live pressure on enterprise AI architects. Operators planning agent deployments at scale need to model self-hosted vs. API economics explicitly — the crossover point arrives much earlier than most assume when agents run 24/7 rather than interactively.
- Enterprise software moat erosion is accelerating: The synchronized selloff across IBM, Workday, Salesforce, ServiceNow, and Adobe on the same session suggests capital markets are re-rating the enterprise software category structurally, not episodically. AI infrastructure and security vendors (and MCP-layer tooling) are the apparent beneficiaries — a signal for where enterprise AI spend is heading.
- Geopolitical model supply risk is becoming a real enterprise concern: China restricting open-weight model exports (alongside US export controls on chips) is narrowing the "cheap fallback" options that enterprises had assumed would exist. This raises the strategic value of European open models and domestic fine-tuning capacity — fine-tuning infrastructure vendors and European model providers (Mistral, etc.) may see renewed interest.
- Agent governance and identity infrastructure is an emerging gap: With production deployments already at 70,000 agents at single enterprises, the management layer — permissions, MCP server configuration, identity, security, audit — is lagging the deployment rate. Hosts identify this explicitly as "an area that is wide open" for tooling, relevant to any operator building or selling into enterprise agentic stacks.
Full analysis
IBM lost roughly $70 billion in a single day, about a 25% drop, and Daniel Whitenack and Chris Benson on Practical AI want you to read it as the first domino: agentic AI is already re-pricing the entire enterprise software category, not five years out but this quarter. Workday off 10%, Salesforce off 9%, ServiceNow off 8%, Adobe off 6%, all on the same session, while NVIDIA and the chip names went up. The claim under all of it: enterprises are moving spend from software seats to AI infrastructure, and long-running agents make pay-as-you-go API pricing untenable.
This is briefing mode, and the honest frame is that this is a market-narrative episode, not a capability episode. There is one real research thread (the Anthropic workspace paper) and a lot of stock-ticker storytelling. The question worth answering: is the "agent economics" pressure real enough to change what you build, or is it a one-day tape you're being invited to over-read?
The Skeptic. One bad session is not a re-rating. IBM has cratered on earnings before and recovered; attributing a 25% drop specifically to "budgets shifting to agents" is a story you tell after the fact, and Whitenack has a product (Prediction Guard) that sells directly into that story. The 70,000-agents customer is the load-bearing claim, and it's an anecdote with no definition of what counts as an "agent." A cron job with an LLM call is an agent if you squint. For a PM: Benson and Whitenack are pointing at a stock chart and calling it a law of nature. Before you rip out your Salesforce contract, notice that nobody quantified how many of those "agents" do anything a human would miss.
The Researcher. Strip the tape and there's exactly one thing to read: Anthropic's Verbalizable Representations Form a Global Workspace in Language Models. The finding is that models form functional analogs to working memory, internal "workspaces" they can report on. That's genuinely relevant to agent routing, because a model that can verbalize its own intermediate state is easier to inspect and hand off between steps. But Benson's leap from that paper to "loop engineering will be antiquated before year's end" is not in the paper. Working-memory analogs do not abolish the orchestration layer; they make one part of it more legible. For a PM: the paper says the model keeps a scratchpad it can describe, not that it no longer needs a manager.
The Compute Pragmatist. The real content is Whitenack's point on inference economics, and it holds up. An interactive chatbot bills in bursts. An agent running 24/7 in a loop bills continuously, and at frontier API rates that line item stops being rounding error fast. That's why fine-tuning a small self-hosted model comes back: not because it's smarter, but because when a workload runs constantly the crossover from per-token API pricing to owned inference arrives far earlier than the interactive case. This maps onto a decision you can actually model. Take your agent's tokens-per-day, price it at API rates, compare to a fine-tuned 7B-class model on your own GPUs plus ops. For always-on loops the math tips toward self-hosting sooner than most architects assume.
The Open-Source Advocate. The genuinely new supply-side wrinkle is China reportedly moving to restrict export of its open-weight models, plus a gray-market for reselling blocked OpenAI and Anthropic API access to mainland users. That matters because the cheap Qwen/DeepSeek fallback was a lot of teams' hedge against US API pricing. If that door narrows, the strategic value of European open weights (Mistral and friends) and of your own fine-tuning capacity goes up. Own the weights, own the economics. For a PM: the discount alternative to expensive US models might get harder to buy, so being able to run your own becomes worth more.
The Builder. Whitenack and Benson name the real gap almost in passing: at 70,000 agents, the management layer (identity, permissions, MCP config, audit) is wildly behind the deployment rate. The saved reading makes this concrete and scary. OpenAI's agent accidentally ran a cyberattack against Hugging Face, and Runlayer is suing Rippling over an MCP gateway. That's the actual Tuesday-morning work: before you scale to thousands of agents, who can each one talk to, what's the blast radius when one loops wrong, and where's the audit trail? The capability exists. The governance to run it safely does not.
Where they part ways. The Skeptic says one selloff isn't a trend and the agent counts are unaudited marketing. The Compute Pragmatist says the inference-cost pressure is real regardless of what IBM's stock did on one day, and it's already forcing self-host decisions. Both can be true: the market narrative is overcooked and the underlying economics are shifting. Second split: the Researcher reads the Anthropic paper as a modest legibility gain, while Benson reads it as the end of loop engineering. That's the gap between what a paper found and what a founder wants it to mean.
What this actually hinges on: whether "always-on agents break API pricing" survives contact with real workloads. An arithmetic question, and you can settle it yourself. Run one representative long-lived agent for a week, log the tokens, price it three ways: frontier API, a hosted small open model, and a fine-tuned model on your own GPUs including ops overhead. If the crossover is inside your contract term, Whitenack and Benson are directionally right about the mechanism even if they're wrong about IBM. The enterprise-software-is-dead framing I'd shelve. The inference-economics framing I'd model this month.
Prediction: No major US enterprise software incumbent (IBM, Salesforce, ServiceNow, Workday, Adobe) will report an actual year-over-year revenue decline attributable to agentic AI in the earnings cycle running through Q1 2027; the "agents are killing enterprise software" thesis will show up in stock moves, not in the top line.
Confidence: Medium. One-day repricing runs ahead of real revenue by many quarters.
Why: The episode's whole case rests on a single trading session, and stock prices move on expectations while revenue moves on signed contracts that renew slowly. Enterprise software sells on multi-year deals with switching costs measured in re-implementation pain, so even if agents do erode seats, the effect lands in renewals over years, not in the next two prints. The incumbents are also shipping their own agent products and repricing toward consumption, which cushions the top line. The opposite outcome, a named incumbent citing agentic displacement for a real revenue drop this soon, would require enterprise buyers to rip out working systems faster than any prior software transition, and there's no evidence in the episode that's happening beyond one unaudited "70,000 agents" anecdote.
Revisit by 2027-04-30: We're right if none of those five names attributes an actual YoY revenue decline to agentic AI displacement through the Q1 2027 earnings cycle. We're wrong if at least one does.
Comments