---
What Is the Unused SaaS License Problem and How Bad Is It?
Roughly 53% of enterprise SaaS applications are unused or underutilized, according to data from Ramp citing industry research. Only 34% of subscriptions are actively used, meaning roughly two thirds of subscription spending may deliver zero measurable value. The estimated annual cost of this waste per organization: approximately $21 million.
That number isn't a rounding error. It's a structural condition.
The figure has been independently corroborated across multiple industry data points a LinkedIn post by Eric Marquez puts unused licenses at 52.7%, and CX Today's survey data notes the number of apps per organization has already shrunk by 18% between 2022 and 2024. This isn't a new crisis being "discovered" it's a chronic dysfunction that enterprises have tolerated for years because procurement moved faster than governance.
More than 10% of entire IT budgets disappear into software that no one logs into, according to Ramp's analysis. In larger organizations, that share climbs higher. The education sector leads all verticals in waste rates at 47%, followed by energy and technology companies the same verticals most aggressively being targeted with AI upsell campaigns.
The practical consequence: before any AI initiative can be connected to existing SaaS infrastructure, someone has to audit what's actually being used, who owns it, and whether the underlying data is trustworthy. That's not an AI problem. It's a portfolio hygiene problem that AI is now forcing into the light.
---
Why Is Enterprise SaaS Consolidation Accelerating Now?
Four forces are compressing the timeline. Integration costs alone can consume 25–35% of the total cost of an AI project often more expensive than the software itself (CX Today). That single data point is reshaping procurement logic faster than any vendor roadmap.
Force 1: AI is making the hidden cost of fragmentation visible. When an enterprise attempts to deploy a GenAI workflow across five different disconnected SaaS tools, the integration bill arrives immediately. Consolidating first isn't idealism it's cost control.
Force 2: PE capital is actively funding roll-ups. In capital markets specifically, growth equity firms including Accel-KKR, Battery Ventures, and Francisco Partners are backing vertical SaaS consolidation plays. The 2025 deal activity from Broadridge, ION, SS&C, and Nasdaq are described in the research as "deliberate plays for long term dominance," not opportunistic acquisitions. When 50%+ penetration is already achieved within a niche, organic growth stalls M&A becomes the only credible path to expanding revenue per customer.
Force 3: Buyer preference has consolidated before the market did. 84% of buyers want integrated solutions over fragmented tools, per the LinkedIn/Fintech M&A research. That number is a vendor death sentence for single-point solutions that can't demonstrate deep platform connectivity.
Force 4: Regulatory complexity is compounding. Germany's SaaS market alone is projected to grow from €6.85 billion to €16.3 billion by 2025, driven substantially by GDPR and data residency requirements (Insentra Group). Across GDPR, HIPAA, SOC 2, and emerging AI governance mandates, fewer integrations and consistent SSO/IAM isn't just operationally cleaner it's a compliance necessity. Fragmentation is a liability on an audit trail.
---
Is the AI Upsell Working or Is It a Trap?
In the short term, AI upsells are generating real incremental revenue for market leaders. ETR Research's data shows early indications that AI features are adding measurable revenue for Microsoft, Salesforce, and ServiceNow. One ETR panelist was direct: "best of breed SaaS will benefit enormously [from AI upsells]… players like Microsoft and Salesforce have enormous leverage." Customers appear willing to pay premiums for AI capabilities from platforms they already trust.
That's the honest answer right now, it's working.
But the structural question underneath is harder. A commenter in Eric Marquez's LinkedIn thread put it sharply: "Why would an enterprise pay a premium for resold AI tokens when it can increasingly access the same frontier models directly?"
That's not a rhetorical question. It's the actual procurement conversation happening in IT departments right now.
The counterforce to the upsell optimism: vendors are responding to AI competitive pressure by layering LLMs on top of legacy tooling. HubSpot is building "dozens of agents. Salesforce has Agentforce. ServiceNow is building ITSM/HR agents. HFS Research, cited in a Medium analysis, argued these AI layer approaches face a structural constraint the underlying datasets are shaped by software design decisions made years ago, not by the actual operational reality of the enterprise using them. As one commenter noted: "SaaS assumed the workflow was mostly known. AI starts by admitting the workflow is messy, undocumented, and different at every company."
That tension isn't going away. And the pricing experimentation makes it worse. ETR Research documents that generative AI pricing remains unsettled per user fees, usage based models, and outcome based pricing are all in play simultaneously. High surcharges deter adoption; usage based plans introduce cost unpredictability. Neither is a clean story to sell to a CFO who is simultaneously being told to cut SaaS spend.
Salesforce's stock experienced what HFS Research described as its "worst trading day in two decades." That's not proof the AI upsell is failing but it signals that the market is asking the same structural questions that buyers are.
---
What Does Consolidation Actually Cost and What Does It Save?
The math on consolidation is more honest than most vendors will show you. Research cited by the Subwise Blog (drawing on data from Binadox and Montro) shows that discovering and eliminating duplicate subscriptions produces double digit spend reductions. The starting point matters: enterprises routinely manage SaaS portfolios numbering in the hundreds of tools.
The savings aren't just line item subscription cuts. Fewer integrations means fewer custom connectors to maintain, fewer failure points in data pipelines, and simplified audit trails for compliance. For any enterprise attempting to build a coherent AI data layer, cleaner data models with fewer conflicting SaaS schemas produce clearer signals and that has compounding value over time (per SAP News, cited in the Subwise analysis).
What consolidation doesn't solve: the ~6 new AI applications entering enterprise environments per month even during active consolidation efforts (Insentra Group). Procurement governance hasn't caught up with the velocity at which individual teams adopt point AI tools. Shadow IT never died it just rebranded as AI experimentation. Enterprises that consolidate their core stack and ignore the edges will find themselves back at 50%+ waste within 24 months.
The operational recommendation from Ramp's research is concrete: renegotiate contracts armed with utilization data, pursue true down clauses and rollover credits, opt for shorter contract terms where possible, and pressure test whether usage based pricing reduces total cost relative to seat licenses. None of this is novel it's just rarely enforced because someone has to do the work of pulling the utilization data first.
---
What This Means for Vendors Betting on AI Upsells
Vertical SaaS vendors who have already hit 50%+ penetration in their niche face a specific math problem. Growth equity firms don't fund flat ARR. When you've saturated your addressable market on seats, AI becomes the primary lever to expand revenue per existing customer not because it's necessarily the right product decision, but because it's the available one.
That creates a misalignment between vendor incentive and buyer need that's worth naming directly. Vendors need upsell revenue. Buyers need their existing tools to work better. Sometimes those align. But when a vendor deprecates existing features to force migration to a pricier AI tier a tactic Insentra Group's research specifically flags as a market behavior the trust damage is real and often irreversible at renewal.
Gartner's projection that 80% of enterprises will have deployed GenAI enabled applications by 2026 sounds like a tailwind for every SaaS vendor. But read the detail: many of those GenAI deployments will be net new SaaS tools, not capabilities added to existing platforms. That's a market signal, not a guarantee for incumbents.
1 in 3 capital markets firms already use AI in operations. Another 23% plan to implement within 12 months. That's real demand. But 84% of those buyers want integrated solutions. The vendor who wins isn't necessarily the one with the most impressive AI demo it's the one whose data architecture is clean enough to make the AI actually work on a buyer's actual, messy enterprise data.
---
How Should Enterprise Buyers Respond?
Enterprise buyers have more leverage in this cycle than vendor marketing would suggest. The consolidation pressure is real, but so is the negotiating position created by having utilization data. A procurement team that walks into a renewal conversation with documented seat utilization rates showing that 50%+ of their licenses are idle has a structurally stronger position to demand pricing concessions, shorter terms, or usage based alternatives.
The strategic sequence that emerges from the research is logical, if not glamorous: audit first, consolidate second, then evaluate AI capabilities on top of a rationalized stack. Attempting to bolt AI onto a fragmented, half idle SaaS portfolio doesn't produce intelligence it produces expensive integration overhead that consumes 25–35% of total AI project cost before a single use case goes live.
For buyers evaluating AI upsell pitches from existing vendors, the right question isn't "is this AI impressive?" The right question is: does this vendor's AI have access to our actual data, or is it operating on a constrained dataset shaped by their software design rather than our operational reality?
That distinction between AI built on clean, relevant enterprise data versus AI layered on top of legacy schema is where the actual value gap lives.
---
The Bottom Line
The unused license problem isn't a budgeting failure. It's a signal that enterprise software purchasing outran enterprise software governance and AI is now the forcing function that makes the bill come due. Organizations that treat consolidation as a cost cutting exercise will capture the easy savings. Those that treat it as a data architecture decision will be the ones whose AI projects actually deliver something.
For vendors, the AI upsell is a viable short term revenue strategy ETR Research confirms it's working for market leaders right now. The medium term risk is the pricing credibility question: when a buyer can access the same frontier model directly, the value of the SaaS wrapper has to be the data layer, the workflow integration, and the operational trust. If those aren't genuinely stronger than the alternative, the premium won't hold.
The consolidation wave isn't finished. With ~6 new AI tools entering enterprise environments per month even during active cleanup efforts, the portfolio sprawl is self-replicating. The enterprises that build procurement governance around utilization data not just initial license counts will be the ones that compound savings rather than cycling through cleanup exercises every two years.
---
The question worth sitting with: If 84% of enterprise buyers say they want integrated solutions, but ~6 new point AI tools enter the average enterprise environment every month anyway who's actually making those purchasing decisions, and do they have any idea what it's costing the organization to integrate them later?
Loading comments...