The AI Tool Explosion: What the Data Shows
We asked three departments how many AI tools they use. They said five. We found seventeen.
That gap between perception and reality is the defining characteristic of AI tool sprawl in 2026. While leadership teams debate AI strategy in quarterly reviews, employees across the business are signing up for AI tools at a pace that makes the early SaaS explosion look tame. And most finance and IT leaders have no idea how deep it goes.
Based on Cledara platform data covering thousands of SaaS subscriptions, 85% of companies now have dedicated AI tools in their software environment. Not AI features embedded in existing products. Standalone AI subscriptions, paid for with company money, used by teams across the organisation. That number rises further when you include SaaS products that have added AI capabilities (and AI pricing) in the last 18 months.
The average company now has 4 to 5 AI subscriptions
Among companies on the Cledara platform, the average business pays for 4.5 AI-related tools. Some have as many as 16. Half of all companies are paying for OpenAI, roughly 39% pay for ChatGPT or Claude, and 30% subscribe to Cursor. That is before counting the dozens of niche AI tools, from ElevenLabs to Midjourney to Lovable, that individual teams adopt without central oversight.
The growth trajectory is even more striking. AI subscriptions grew 84% year over year between 2024 and 2025. In 2022, AI tools represented just 1% of all new software subscriptions. By early 2026, that figure has hit 29% of all new subscriptions. Nearly one in three new tools a company buys today is an AI tool.
Microsoft's 2025 Global AI Adoption report confirms this is not limited to Cledara's customer base: 78% of companies worldwide have adopted AI technologies, with the average organisation now using AI across three or more business functions. Among tech-forward companies in the 30 to 500 employee range, adoption rates are likely even higher.
Most AI investments are not delivering value
Speed of adoption does not equal speed of value creation. According to BCG's 2025 research, 60% of companies generate no material value from their AI investments, and only 5% create substantial value at scale. McKinsey's State of AI report paints a similar picture: while 88% of organisations use AI in at least one function, only 39% see any measurable impact on EBIT, and most of those report improvements below 5%.
The problem is not AI itself. The problem is ungoverned proliferation. When every team buys its own AI tools independently, the result is fragmented spend, duplicated capabilities, and zero visibility into what is actually working. BCG's own finance-focused research found that only 45% of finance executives can even quantify the ROI from their AI initiatives.
Where AI Tools Are Hiding in Your Stack
AI tool sprawl is harder to spot than traditional SaaS sprawl because AI spend hides in three distinct layers, and most companies only track one of them.
Layer 1: Standalone AI tools
These are the obvious ones: ChatGPT, Claude, Cursor, Midjourney, Jasper, Perplexity, ElevenLabs. They show up as their own line items in your card statements, and they are the tools most companies think of when they hear "AI spend." On the Cledara platform, the most widely adopted standalone AI tools are OpenAI (used by roughly half of all companies), ChatGPT and Claude (each at around 39%), and Cursor (30%). Engineering teams are the heaviest users, followed by marketing and design.
But standalone tools are only the visible tip. The Cledara platform's directory of 6,000+ tools now includes over 100 AI-specific vendors, and new ones appear every week. Many of these are purchased by individual contributors on personal cards and expensed later, making them nearly invisible to centralised tracking.
Layer 2: AI add-ons to existing SaaS
This is the layer most companies miss entirely. Over the past two years, nearly every major SaaS vendor has added AI features, and they have not been shy about raising prices to pay for them. Notion eliminated its standalone AI add-on in 2025, folding AI into its Business tier and effectively doubling the cost for many teams. Canva raised its Teams pricing by up to 300%, citing AI feature expansion. Salesforce introduced Agentforce at $125 per user per month on top of existing licences.
When your existing tools add AI features with corresponding price increases, your AI spend grows without a single new subscription appearing in your records. It simply looks like "Notion got more expensive," not "we are now paying for another AI tool." A Gartner report found that 61% of businesses plan to upgrade recently purchased software specifically to gain AI features, which means this hidden AI spend layer is about to get significantly larger.
Layer 3: Developer API usage
The most invisible layer of all. Engineering teams increasingly consume AI through APIs: OpenAI, Anthropic, Cohere, and others billed on a usage basis. On the Cledara platform, Anthropic's API appears in 17% of companies as a separate subscription from the Claude consumer product. These costs are variable, unpredictable, and often buried in cloud infrastructure budgets rather than tracked as software spend.
According to CloudZero's 2026 State of AI Costs report, the average monthly AI spend jumped from $63,000 to $85,500 in just one year, a 36% increase. Nearly half of companies now plan to spend over $100,000 per month on AI, up from just 20% in 2024. The report also found that 72% of leaders say cloud spending, driven substantially by AI, is becoming increasingly unmanageable.
The Cost: More Than Just Money
AI tool sprawl creates three compounding problems that extend well beyond the line items on your P&L.
Redundancy: three teams paying for three different AI writing tools
When marketing uses Jasper, the product team uses Claude, and customer support uses ChatGPT, the company is paying for three tools that do substantially the same thing. This is not hypothetical. On the Cledara platform, the average company with AI tools pays for nearly five AI subscriptions, many with overlapping capabilities. Multiply redundant subscriptions by the number of teams making independent purchasing decisions, and you get tail spend that adds up fast.
AI spend now represents roughly 5.6% of total SaaS expenditure across the Cledara platform. That percentage is climbing rapidly as the share of new AI subscriptions approaches 29%. Without consolidation, companies will find themselves spending more on duplicative AI tools than on entire categories of their traditional software stack within the next 12 to 18 months.
Data risk: company data flowing to 10+ LLM providers
Every AI tool your employees use is another endpoint for sensitive company data. CyberHaven's research found that 71% of AI tools put enterprise data at risk. Nearly half of employees admit to sharing sensitive work data with unauthorised AI tools, including financial information, employee records, and proprietary research.
IBM's 2025 breach report found that one in five organisations experienced data breaches through shadow AI, adding an average of $670,000 to breach costs. Meanwhile, 58% of shadow AI users rely on free tool versions that lack enterprise-grade security features. When company data flows to ten or more LLM providers, each with different data handling policies, the attack surface expands in ways that traditional security controls were never designed to manage.
Integration chaos: every team building different AI workflows
Without centralised governance, individual teams build their own AI workflows, integrations, and automations. Engineering embeds Cursor into their development pipeline. Marketing connects Jasper to their content calendar. Sales hooks up an AI notetaker to their CRM. Finance experiments with AI-powered forecasting tools. Each integration is a standalone decision with its own data flows, permissions, and failure modes, creating a web of dependencies that no single person understands and no one can maintain when something breaks.
This fragmentation also kills knowledge sharing. When three teams use three different AI tools for similar tasks, the organisation never develops shared best practices, prompt libraries, or evaluation criteria. Every team reinvents the wheel, and the company pays for it three times over. The 2025 AI adoption data makes clear that this fragmentation is accelerating, not slowing down.
How to Get AI Sprawl Under Control
Audit first, govern second, optimise third
The instinct when confronting AI sprawl is to jump straight to governance: set policies, restrict purchasing, require approvals. That approach fails because you cannot govern what you cannot see. Start with a complete audit of every AI tool in your environment, including the three layers outlined above.
A thorough AI audit should answer four questions. What standalone AI tools are employees using? Which existing SaaS products have added AI features (and AI-driven price increases)? What AI API consumption exists in engineering and data teams? And which of these tools overlap in functionality?
Once you have visibility, build governance that matches your company's relationship with AI. Set spend thresholds and approval workflows. Establish a preferred vendor list for common AI use cases: writing, coding, image generation, transcription, and data analysis. Map every AI tool to a business owner who is accountable for its value. Review AI spend monthly, not quarterly, because the category moves too fast for quarterly cycles.
Pay special attention to the AI add-on layer. When vendors like Notion, Canva, and Salesforce raise prices citing AI features, evaluate whether your team actually uses those AI capabilities. If not, you may be paying an AI premium for features nobody touches. Understanding which AI features deliver value versus which are simply bundled into higher-tier plans is one of the fastest paths to reducing bloated SaaS costs.
Only after you can see and govern your AI stack should you optimise: consolidating redundant tools, negotiating renewals with volume leverage, and reallocating budget from underperforming AI investments to the ones that actually deliver results.
Do not ban AI: channel it
The worst response to AI sprawl is to restrict AI adoption entirely. Companies that ban AI tools do not eliminate AI usage; they push it underground. CIO Magazine reports that 68% of employees use unauthorised AI tools at work, up from 41% in 2023. Banning AI simply means you lose all visibility into how it is being used and what data it is consuming.
The smarter approach is to channel AI adoption through managed pathways. Make it easy for employees to request and get approved AI tools quickly. Provide company-wide licences for the most commonly needed capabilities. Create fast-track approval for low-risk, low-cost AI tools so teams are not incentivised to go around the process. The goal is to make the governed path the path of least resistance, not a bureaucratic barrier that drives people to their personal credit cards.
How Cledara Discovers and Manages AI Sprawl
Cledara's platform is purpose-built to give finance and IT teams complete visibility into software spend, and AI tools are where that visibility matters most right now.
Engage, Cledara's browser extension, discovers AI tools across all browsers without employees needing to self-report. It detects shadow AI usage in real time, identifying tools like ChatGPT, Claude, Midjourney, and dozens of niche AI applications that never appear on a purchase order. The typical Cledara customer discovers 20+ previously unknown applications, and a growing share of those are AI tools.
Cledara's AI Dashboard centralises all AI-related spend into a single view, connecting to providers like OpenAI and Anthropic via API keys to track usage-based costs with daily spend visualisation. Instead of hunting through cloud bills and credit card statements, finance teams see exactly what the company spends on AI, broken down by tool, team, and trend.
With a directory of over 6,000 software tools, Cledara automatically identifies and categorises new AI tools as they appear, including the long tail of emerging tools that traditional spend management platforms do not yet recognise. Combined with virtual cards per subscription, approval workflows, and one-click cancellation, Cledara gives you the controls to manage AI adoption without slowing it down.
AI is not going back in the box. The companies that thrive will not be the ones that spend the most on AI, or the least. They will be the ones that know exactly where every AI dollar goes, and can prove it is working. Discover every AI tool your company is paying for with a free AI sprawl audit from Cledara.
