Greg Isenberg just got back from five days in San Francisco with frontier AI teams, startup founders, and three billionaires — and the thing he couldn't stop talking about wasn't a new model. It was a business model.

Every billionaire at the table, he said, is actively buying SaaS companies and rebuilding them agent-first. Not someday. Now. The reason is brutally simple: AI can replicate $500K worth of software logic for pennies, and the operators who understand outcome-based pricing have a brief window to eat the incumbents' lunch before the market reprices everything.

Here's how to act on that before the window closes.

Why Billionaires Are Buying SaaS Companies to Gut and Rebuild Them

Isenberg's thesis — backed by the people writing nine-figure checks — is this: the traditional SaaS model sells access to features. You pay $500/month for a dashboard, a database, some integrations, and a login. The value isn't the outcome. The value is the tooling that theoretically lets you produce an outcome if you also supply the labor, the expertise, and the patience to click through forty screens.

AI agents collapse that entire stack. An agent doesn't give you a dashboard to manage your outbound pipeline. It runs your outbound pipeline. It doesn't give you a reporting tool — it generates the report, interprets the data, and drafts the recommendation. The feature layer becomes a commodity overnight.

This is why the billionaires aren't building from scratch. They're acquiring existing SaaS companies with distribution — customer bases, brand recognition, integrations — and ripping out the feature-based core to replace it with agent-driven workflows that charge for results. They keep the moat (the customers) and swap the engine (features → outcomes).

For small operators, the signal is clear: the $200–$2,000/month B2B software stack your clients pay for is about to be unbundled by anyone who can deliver the same result faster, cheaper, and with a guarantee attached.

The Outcome-Based Pricing Opportunity (And Why It's a Narrow Window)

Subscription pricing works when the customer can't replicate the tool. That moat is evaporating. When Claude or GPT can generate a functional app from a text description in minutes, the cost floor for software drops to near-zero. What doesn't drop to zero is the outcome the customer actually wanted.

This is the arbitrage: you stop selling software access and start selling the completed result. Instead of "$400/month for our lead enrichment platform," you charge "$3 per verified, enriched lead delivered to your CRM." The customer's cost shifts from fixed overhead to variable spend tied directly to value received.

Why is the window narrow? Because the billionaires Isenberg described are already moving. Once the incumbents reprice their own products around outcomes — and they will, because the pressure is existential — the advantage small operators have (speed, low overhead, willingness to guarantee results) shrinks. Right now, most SaaS companies are still defending their subscription revenue. That defense creates a six-to-twelve month gap where fast-moving operators can establish positioning, build client relationships, and lock in contracts before the elephants turn around.

What 'Agent-First' Actually Means for a Non-Technical Business Owner

Strip away the engineering jargon and agent-first means one thing: the AI does the work, not just the thinking.

In a traditional SaaS product, AI might suggest a subject line or auto-fill a field. In an agent-first product, the AI receives an objective ("book 10 qualified sales calls this week"), plans the steps, executes them across tools, handles exceptions, and reports back with results. You're not logging into a platform and clicking buttons. You're defining the outcome and reviewing the output.

For operators building these products, here's what that looks like in practice:

What you're replacing: The SaaS interface, the manual workflow, and the human labor your client currently supplies to bridge the gap between "tool" and "result."
What you're keeping: The underlying APIs and data sources (CRMs, email platforms, databases). Agents still need to read and write to these systems.
What the rebuilt product looks like: A managed service powered by AI agents, delivered through a simple interface (often just a dashboard or Slack channel), where the client sees outcomes — not configuration screens.

How to Identify a Commoditisable SaaS in Your Niche (And Build the Replacement)

Not every SaaS product is ripe for this. Here's the framework for finding the ones that are:

Step 1: Audit your clients' tool stack

List every SaaS product your clients pay $200–$800/month for. Focus on tools where the client is also paying a human (VA, coordinator, junior employee) to actually operate the software. That human labor cost is your signal — it means the tool alone doesn't deliver the outcome.

Step 2: Map the inputs and outputs

For each candidate tool, define the workflow in simple terms: what goes in, what comes out, and what decisions happen in between. Example: a recruiting tool takes job descriptions as input, produces a shortlist of qualified candidates as output, and the decisions in between are sourcing, filtering, and scoring. If you can describe the full loop in under five sentences, it's automatable.

Step 3: Build the agent workflow

Use an orchestration platform like StickyFlow to wire the agents together. You're not writing software — you're defining the objective, connecting the data sources (via APIs or MCP servers), and setting the guardrails for each decision point. The agent handles sourcing, filtering, and scoring. You handle the architecture and quality control.

Step 4: Test on one real client before you price it

Run the workflow manually alongside the agent for two weeks. Measure where the agent matches human output, where it exceeds it, and where it fails. This data becomes your pricing foundation and your sales proof.

Packaging and Pricing Your Outcome-Based AI Product

Three pricing models work for outcome-based AI products right now:

Per-outcome pricing: $X per lead delivered, per report generated, per candidate shortlisted. Lowest barrier to entry for skeptical buyers. Best for high-volume, clearly measurable workflows.
Usage-tiered pricing: Base fee covers a set number of outcomes per month, with overage pricing above. Gives you revenue predictability while keeping the outcome alignment.
Hybrid retainer: Monthly retainer for the system management and orchestration, plus per-outcome fees for the actual output. Works best for complex workflows where setup, monitoring, and iteration are genuine ongoing work.

The minimum viable guarantee

Skeptical buyers don't need a money-back guarantee. They need a performance floor. "We deliver a minimum of X qualified leads per month or we refund the difference" is more credible and more compelling than "satisfaction guaranteed." Define the floor conservatively based on your test data, and you'll close faster than any SaaS demo ever could.

Accelerating the ops layer

The unglamorous truth: outcome-based products live or die on operational reliability. Your agents need monitoring, error handling, and escalation paths. Platforms like StickyFlow exist specifically to manage this — giving you the orchestration layer, the logging, and the client-facing delivery infrastructure so you're not duct-taping five tools together and praying nothing breaks at 2 AM.

The takeaway is simple: SaaS incumbents are defending subscription revenue while the value shifts to outcomes. The billionaires see it and they're already buying. You don't need to buy a company — you need to pick one workflow in your niche, build the agent-first replacement, price it on results, and ship it before the window closes. The operators who move in the next six months will own the categories. Everyone else will compete on price against the rebuilt incumbents.