The 2026 Application Rationalization Playbook: SaaS Sprawl + AI Tooling Overlap

Average enterprise SaaS spend per employee jumped 21.9 percent in 2025 to $4,830, the first year-over-year increase in three years, driven by AI vendor pile-on (Zylo). 78 percent of AI users bring their own AI tools to work. 49 percent run multiple AI tools simultaneously. Nearly 70 percent of CIOs put rationalization in their top 2025 initiatives (Gartner). Here is the practical playbook for cutting SaaS and AI tooling sprawl in 2026.

Where do we start, with SaaS or with AI tools?

With AI tools. The AI sprawl is newer (most of it accreted in 2024-2025), the duplication is more obvious (3 chat assistants doing the same job), and the security risk is higher. Most teams can complete the AI rationalization in 30 days and use that win to fund the broader SaaS work.

The 2025 SaaS Management Index from Zylo reported that average SaaS spend per employee hit $4,830, a 21.9 percent year-over-year increase. Companies waste $18 million annually in unused licenses and use only 49 percent of provisioned licenses on average. An average of 7.6 applications enter the tech environment each month (Zylo 2025 SaaS Management Index). This was the first increase in average SaaS spend in three years, attributed to rising vendor costs and rapid AI adoption (Zylo press release).

Gartner's March 2025 Cost Optimization Pulse found nearly 70 percent of CIOs list technology rationalization as a top initiative for reducing IT waste. Gartner's forward-looking projection: through 2027, organizations that fail to attain centralized visibility and coordinate SaaS lifecycles will overspend on SaaS by at least 25 percent due to unused entitlements and unnecessary, overlapping tools (Spendflo summary of Gartner).

The good news: BetterCloud's 2025 State of SaaS report found organizations now use an average of 106 different SaaS tools, down from 112, a 5 percent year-over-year decrease, with mid-market companies (1,500-4,999 employees) decreasing apps the most significantly at 28.8 percent (BetterCloud). The sprawl is reversible. The companies that reverse it have a documented playbook. This is that playbook.

What changed in 2025 that made sprawl explode

The 21.9 percent per-employee SaaS spend increase in 2025 is mostly AI vendor pile-on. Menlo Ventures reports enterprise AI investment tripled in a single year, from $11.5 billion to $37 billion. Anthropic now commands 40 percent of the enterprise LLM API market share, with OpenAI at 27 percent and Google at 21 percent (Menlo Ventures 2025 State of GenAI in the Enterprise).

The catch: almost nothing was deprecated to fund the 3x AI spend explosion. Most enterprises are now running Microsoft Copilot AND ChatGPT Enterprise AND Google Gemini AND a vertical AI vendor (Notion AI, Glean, Sana, Anthropic Claude direct, etc.) simultaneously. The Microsoft + LinkedIn 2024 Work Trend Index found 78 percent of AI users bring their own AI tools to work and 49 percent of employees use multiple AI tools simultaneously for different tasks (Programs.com summary). CIO Magazine reported that roughly half of employees are using unsanctioned AI tools, with enterprise leaders as major culprits (CIO), which is the single biggest rationalization blocker, because leadership won't kill apps they're personally using.

The shadow-AI risk is also material. Reco's 2025 State of Shadow AI Report found nearly 98 percent of organizations have employees using unsanctioned AI or apps. Around 54 percent of shadow AI tools have been used to upload sensitive company data and approximately 76 percent of shadow AI tools fail to meet SOC 2 compliance standards. Average cost of a shadow AI data breach has reached $4.2 million.

The visibility problem

The Flexera 2025 State of ITAM Report found 35 percent of respondents say SaaS waste has increased over the past year, and most customers report 25 percent of their SaaS spend is wasted. 63 percent of respondents say they have visibility into cloud instances, while only half feel comfortable with SaaS, and visibility into BYOL positions is at only 27 percent (Flexera).

Translation: most CIOs cannot rationalize what they cannot see. The visibility crisis is the binding constraint. Application rationalization in 2026 starts with discovery, not decisions.

The 5-step rationalization playbook

Step 1: Discovery. Pull every SaaS contract, every cloud-marketplace subscription, every corporate-card AI vendor expense, and every SSO-connected app into one inventory. Productiv, Zylo, Torii, BetterCloud, and Flexera all sell discovery tooling for this; for smaller orgs, an Excel pull from the AP system + IDP is enough. The output is one ledger of every app, owner, license count, and spend.

Step 2: Usage layer. Map actual usage against entitlements. The Zylo 49-percent-utilization average is the baseline. Anything below 30 percent utilization is a candidate for license reduction. Anything below 10 percent utilization is a candidate for elimination.

Step 3: Overlap analysis. Identify functional overlap. The current 2026 hot spots are AI assistants (Copilot vs ChatGPT vs Gemini vs vertical AI), business intelligence (Tableau vs Power BI vs Looker vs Sigma), CRM-adjacent tools (Salesforce vs HubSpot vs Pipedrive), collaboration (Slack vs Teams vs Discord vs others), and observability (Datadog vs New Relic vs Splunk vs Grafana).

Step 4: Decision and consolidation. For each overlap, document the keep / consolidate / replace decision with a named owner and a target date. The decision matrix factors are total cost (license + integration + retraining), strategic fit, and switching cost. Most rationalization programs target 20-30 percent app-count reduction in year one.

Step 5: Lifecycle governance. Stand up an ongoing review cadence (quarterly minimum) with the SaaS-management vendor or with internal IT Asset Management. The 7.6-apps-per-month influx number from Zylo is the rate the lifecycle process must absorb.

What rationalization actually delivers

The BetterCloud 28.8 percent app-count reduction in mid-market is the realistic ceiling for a one-year program. The financial impact at $4,830 per-employee SaaS spend is roughly 15-20 percent reduction in per-employee SaaS spend with no functional capability lost. For a 5,000-employee enterprise, that is roughly $3.6 to $4.8 million annual savings.

The harder-to-measure benefits are operational: faster onboarding (fewer accounts to provision), faster offboarding (fewer entitlements to revoke), reduced security surface (fewer SSO and SCIM integrations to maintain), and improved data governance (fewer data egress points).

The AI-tooling rationalization is a separate exercise with its own ROI. Most enterprises will end 2026 with 1 primary AI assistant (typically Microsoft Copilot for Microsoft shops, ChatGPT Enterprise for non-Microsoft shops) plus 1-2 vertical AI tools (typically Glean for knowledge, plus a developer AI like Claude Code or GitHub Copilot). Anything beyond that bundle should justify its existence on a measurable use-case basis.

The 90-day plan

Days 1 to 30: complete discovery (every SaaS contract, every cloud-marketplace subscription, every AI vendor expense), publish the inventory. Days 31 to 60: complete the usage layer and overlap analysis, present the decision matrix to the CIO for sign-off. Days 61 to 90: execute the first wave of decisions (typically 5-10 contract terminations or consolidations), establish the lifecycle governance cadence.

By day 91 the organization has a documented application portfolio, a quantified savings number for the next FY budget cycle, and a process to keep sprawl from re-accreting.

Frequently asked questions

Where do we start, with SaaS or with AI tools?

With AI tools. The AI sprawl is newer (most of it accreted in 2024-2025), the duplication is more obvious (3 chat assistants doing the same job), and the security risk is higher. Most teams can complete the AI rationalization in 30 days and use that win to fund the broader SaaS work.

Do we need a SaaS-management platform (Productiv, Zylo, Torii, BetterCloud)?

For organizations above 1,000 employees, yes. Below that, an Excel + IDP + AP pull is usually sufficient for the first cycle. Above 5,000 employees, a SaaS-management platform pays for itself within one year through avoided overspend.

What's the right rationalization target?

The BetterCloud mid-market 28.8 percent app-count reduction is the realistic ceiling. 15-20 percent is a defensible year-one target for most organizations.

How do we handle leadership shadow AI use?

Surface it explicitly. The CIO Magazine finding that enterprise leaders are major culprits in unsanctioned AI use means the rationalization needs the CEO's air cover. Without that, the program will hit a wall at the executive layer.

What about the AI tools that produced documented ROI?

Keep them. The point of rationalization is not to eliminate AI; it is to consolidate to the AI vendors that produced documented ROI and eliminate the ones that did not. Use the 2026 AI Governance Operating Model inventory ledger as the system of record for which use cases survived the cut.

If you want the longer version of this analysis, including the full discovery template, the overlap-analysis matrix, and the lifecycle governance playbook, our Data Foundation, AI Readiness, and Data Governance Consulting practices ship the rationalization end-to-end. The platform-level decision (Snowflake vs Databricks) is in our Snowflake vs Databricks for AI Workloads in 2026 blog. Anchor case studies: the Texas A&M System 11-warehouse Snowflake consolidation (24 weeks, $2.3M to $380K) and the Desert Diamond Casinos 5-PMS to 1 consolidation (18 weeks, $1.9M annual technology cost saved) are the deepest published Thinklytics rationalization engagements.

How is this different from a SaaS spend management project?

SaaS spend management catalogs subscriptions and finds savings. Application rationalization adds the business-criticality and replacement-fit layers, so the output is a decision tree (keep / replace / retire / consolidate) instead of a savings number. The two work well together; spend management feeds the inventory, rationalization makes the decisions.

Do we need a SaaS spend management tool to start?

Productiv, Zylo, Vendr, BetterCloud, or Torii speed the discovery phase from 8-12 weeks to 2-4 weeks. Whether the tool pays back depends on portfolio size: under 50 SaaS apps, manual discovery often beats the tool subscription. Above 100 apps, the tool is hard to beat.

How does Thinklytics scope a rationalization engagement?

We typically run a 3-month phase one (inventory + business-criticality scoring) and a 6-month phase two (the actual consolidation work). Senior-led, fixed scope, fixed fee. The business-owner involvement schedule is documented day one. Read more at system consolidation.