Private equity firms have poured hundreds of billions of dollars into enterprise software firms over the last few decades, on the assumption that terminal values and growth rates were sizable.
But artificial intelligence has raised existential questions about enterprise software valuations because it raises the issue of positive feedback loops, which are inherently stability disrupting.
A positive feedback loop is a process where the output of a system amplifies or intensifies the initial stimulus, driving the system further away from its original state. A negative feedback loop, on the other hand, allows the system to adjust.
source: Citrini Research
Unlike negative feedback that maintains stability, positive loops often cause exponential change in a single direction, without the corrective feedback that allows the system to stabilize itself.
In the context of what AI might do, it could, in principle, reduce employment. Reduced employment might lead to less consumer spending. Less consumer spending should put pressure on gross revenues and profit margins.
That might increase reliance on AI to reduce operating costs, which in turn further reduces employment and spending, and therefore firm revenues and profits.
If AI agents can write code and execute complex workflows autonomously, what happens to the software companies built on charging per human "seat" or user?
What happens to cash flows, and the valuation models built on those cash flows?
How much enterprise software value creation will happen in the future, compared to other segments of the market? And what does that imply for asset holding periods and exits?
Area of PE Operation | The Historical SaaS Era | The Current AI-Disrupted Reality |
Valuations & Multiples | Software enjoyed near-automatic, premium revenue multiples (often 20x+ ARR) due to reliable recurring revenue and low customer churn. | A fundamental re-rating is underway. Investors are questioning long-term defensibility, dragging exit expectations down (e.g., closer to 15x EBITDA). PE firms are facing potential markdowns on legacy SaaS portfolios. |
Due Diligence | Focus was heavily weighted on Net Revenue Retention (NRR), customer acquisition costs (CAC), and the "stickiness" of workflows. | Firms are now commissioning aggressive "AI vulnerability audits." The core question is whether a target's product is an easily replaceable interface, or if it possesses proprietary, structured data that AI agents need to function safely. |
Value Creation (Operations) | Growth relied on expanding sales/marketing teams and steadily increasing the number of licensed human users per client account. | Operating teams are forcing portfolio companies to pivot away from seat-based pricing toward outcome-based or usage-based models. Genuine AI integration is now required for survival, not just for marketing. |
Leverage & Private Credit | Software LBOs were built on highly secure, predictable cash flows, allowing PE firms to utilize significant leverage. | Private credit markets are showing stress (evidenced by the recent gating of tech-heavy credit funds in early 2026). Lenders fear that if AI shrinks a software company's user base, cash flows will contract, increasing default risks. |
Exit Environments | Shorter hold periods (3–5 years) with reliable exits via public markets (IPOs) or sales to strategic buyers/larger PE funds. | Hold periods are extending. Exits are incredibly difficult to execute without hard proof that a portfolio company has successfully evolved into an AI-augmented platform rather than a legacy SaaS tool facing obsolescence. |
The sort of interesting issue is that lots of observers have wanted to shift pricing from a more-commodity-like “seats” model to a “success-based” or “outcomes” pattern, the thinking being that this approach provides higher margins and perceived product value.
Such an approach, in principle, makes more revenue sources available to suppliers: not just a share of the information technology budget but a share of saved labor costs; revenue shrinkage; customer acquisition; churn; marketing or other operations costs.
Company, Product | Function | Outcome Metric | Pricing Mechanism | Strategic Benefit |
Intercom Fin | AI customer support | Tickets resolved | ~$0.99 per successful resolution | Aligns price with support deflection value; drives product improvement (Pragmatic Institute - Corporate) |
Zendesk AI agents | Customer service | Successful resolutions | Charge per resolved ticket | Hybrid model linking AI value to outcomes (L.E.K. Consulting) |
Chargeflow | Payments / fintech | Chargebacks recovered | % of recovered revenue | Captures share of direct financial benefit (rezoomex) |
Riskified | E-commerce fraud | Approved fraud-free transactions | Charge per approved transaction | Direct link between trust and revenue protection (L.E.K. Consulting) |
Vendr | SaaS procurement | Savings negotiated | % of savings | Monetizes cost reduction achieved (tanayj.com) |
ServiceNow pilots | Workflow automation | Efficiency improvements | Full payment only if targets met | Demonstrates ROI in enterprise automation (Monetizely) |
InsideSales / XANT | Sales acceleration | Conversion & velocity improvements | Performance-linked pricing | Aligns cost with revenue impact (Monetizely) |
HubSpot performance tiers | Marketing automation | Campaign performance metrics | Discounts tied to results | Incentivizes effective usage (Monetizely) |
Hitachi Rail “trains-as-a-service” | Industrial IoT | On-time performance | Payment linked to punctuality | Transfers reliability risk to vendor (Pragmatic Institute - Corporate) |
Identity verification (iDenfy) | Fraud / KYC | Verified users | Charge per approved identity | Clear, measurable success metric (L.E.K. Consulting) |
To use a simple example, instead of selling shovels, one sells “holes.” In some ways, it is similar to the shift from “product” pricing to “services” pricing. Instead of selling a car, one sells transportation services, with recurring fees rather than one upfront purchase.
Business Function | The "Outcome" Being Sold | Real-World/Current Examples | Pricing Mechanism |
Customer Support | A successfully resolved customer inquiry. | Intercom (Fin AI), Zendesk | No charge for "chats"; you only pay (e.g., $0.99) when the AI solves the issue without human intervention. |
Legal & Compliance | A completed, filing-ready legal document. | EvenUp, Harvey | Instead of a monthly license for paralegals, firms pay per AI-generated demand letter or contract audit completed. |
Sales & Marketing | A "Sales Qualified Lead" (SQL) or booked meeting. | 11x.ai, Salesforce (Agentforce) | Companies pay for the meeting booked by the AI agent, rather than paying for a seat for a human Sales Development Rep (SDR). |
Fintech / Payments | A recovered chargeback or fraud-free transaction. | Chargeflow, Riskified | The vendor takes a percentage of the money recovered or a fee only for transactions that are proven to be legitimate. |
Cloud Operations | Hard-dollar savings on infrastructure bills. | Viewnear, ProsperOps | The software is free or low-cost; the vendor takes a 20-30% "success fee" of the actual savings generated on the client’s AWS/Snowflake bill. |
Supply Chain | A successfully renegotiated vendor contract. | Pactum, Luminous | The AI agent autonomously negotiates with thousands of "long-tail" vendors; the fee is a share of the cost-reduction achieved. |
Up to this point, such outcomes-based pricing has been quite difficult to implement. AI might make it imperative.
No comments:
Post a Comment