Report · April 2026
State of Applied AI
A structured view of how organizations are deploying AI today, based on 200 real-world implementations.
Founder of Applied. About →
Since the release of ChatGPT in 2022, AI adoption has accelerated. New tools continue to emerge and capabilities are improving quickly, yet their real impact remains difficult to measure. Organizations across industries are experimenting with AI, but how is it actually being used today?
To address this gap, I created Applied, a living map of how organizations deploy AI across industries and business functions. It documents real implementations, the tools behind them, and the areas where measurable change is already occurring.
This first analysis covers approximately 200 AI use cases identified between 2025 and 2026, spanning more than 200 companies, 300 tools, 25 industries, and 15 business functions. We'll keep adding cases, tools, and outcomes over time.
Key Insights
- AI adoption is concentrated in Operations (39%) and Software Engineering (21%) workflows.
- Speed is the most common measurable outcome of AI deployment.
- Financial ROI is primarily framed as avoided cost rather than revenue growth.
- Enterprise AI deployments follow layered stack architectures rather than standalone tool adoption.
- AI embedded in enterprise software deployments (47%) exceeds direct model integration (31%).
- Technology (27%) and Financial Services (14%) companies lead adoption among documented deployments.
Dataset and Scope
The dataset is based on public company case studies, vendor documentation, and recent implementation reports. We analyzed and structured these sources to identify the tools used, reported outcomes (KPIs), and the business stakeholders involved. Only use cases involving AI-native products and implementations from 2025 onward were included.
Top Industries Using AI
Figure 1. Technology and Financial Services lead documented enterprise AI deployments
The sample is primarily US-based companies (60%), followed by the United Kingdom (8%), Germany (3%), and Japan and Australia (2% each). The dataset also skews toward large organizations. Nearly 46% of the companies included have more than 5,000 employees, while 15% have fewer than 50 employees.
Most are already in production. About 54% of use cases are classified as mature and 43% as actively scaling. Only five cases (2.6%) remain in pilot phases.
Because the analysis is derived from publicly reported implementations by companies and vendors, it reflects successful deployments rather than failed cases.
Top Industries Using AI
Technology accounts for 27% of use cases, followed by Financial Services (14%), Healthcare (8%), and Professional Services (7%). Together, these industries represent more than half of all documented deployments.
Several large global industries appear underrepresented. Manufacturing accounts for four use cases (1.7%), and Logistics & Transportation for five (2.1%). Given their economic weight, this likely reflects low disclosure or sample size, not lower adoption.
Where AI Gets Deployed
Figure 2. Operations and software engineering account for the largest share of enterprise AI deployments
Given the sample size, industry distribution should be interpreted as directional rather than exhaustive.
Top Business Functions Transformed by AI
Operations accounts for 39% of use cases, followed by Software Engineering (21%), Marketing (12%), and Customer Service (12%). Together, these four functions represent the majority of deployments.
Finance, Sales, Security, and Product Development appear less frequently in the sample, while Business Intelligence, Supply Chain, and Human Resources remain only marginally represented.
AI in Operations
Support (customer service), physical operations, document processing, and large-scale automation represent the highest-impact cluster.
Four recurring patterns emerge across industries:
Internal IT and support deflection
Organizations are reducing time spent handling internal requests and repetitive tickets. Examples include Palo Alto Networks (351,000 staff hours saved), Albemarle (80% ticket resolution without IT), and Databricks (73% ticket deflection).
Fleet and physical operations
Operational and sensor data are increasingly used to automate routing, maintenance, and safety decisions. C.H. Robinson automates ~5,500 orders daily, Massey Services saves $1.3M annually in fuel, and CoxHealth reduced maintenance time by 94%.
Document and knowledge operations
Review and retrieval workflows are accelerating across contracts, policies, and internal knowledge bases. Thomson Reuters deployed AI across 3,000 subject-matter experts, Fifth Dimension reduced memo preparation from days to 30 minutes, and Deutsche Telekom saved ~5 hours per attorney per week.
Scale and volume automation
Some deployments are already operating at workforce scale. SoftBank logs 4,500 FTE-equivalents per year, Gordon Food Service saves more than 20,000 hours per month, and Shinhan Bank records 1.2 million hours annually.
Reported AI Outcomes
Disclosed metrics when describing AI results
Figure 3. Speed-related outcomes appear far more frequently than revenue effects
AI in Software Engineering
Software Engineering is the second largest area of AI deployment in the dataset, with adoption concentrated in code generation, system modernization, and tools that expand software creation beyond engineering teams.
Three recurring patterns emerge:
Code generation and acceleration
Classmethod delivers a 90% reduction in development time, Headstart generates 90–97% of client code with Claude, and Stripe deployed Claude Code to 1,370 engineers.
System modernization
CSL exited 29 data centers, Experian saved 300 engineering days migrating .NET systems, and ChargeGuru reduced three-month migrations to two days.
Platform-level software creation
Replit notes that 75% of users are non-developers, and Epic Systems that more than half of Claude Code usage comes from non-developers. Similar patterns appear in platforms such as Anything, tl;dv, and Fifth Dimension.
Productivity gains include 550,000 hours saved at Factory, more than 1,500 hours per month at Cisco, 1,150 hours per developer annually at Postman, and seven hours per week per early adopter at Money Forward.
AI in Marketing
Marketing deployments are concentrated in content production workflows, with smaller clusters in audience engagement and automated lead generation.
Three recurring patterns emerge:
Content operations
Large teams are reducing the time required to write, review, and publish marketing materials. Efficiency gains range from 30% to 80% at organizations including KPMG, Vodafone UK, and Salesforce.
Fan and audience engagement
Some organizations are using AI to personalize experiences at scale. Examples include MrBeast, Scuderia Ferrari, and Pacers Sports & Entertainment, supporting engagement across audiences ranging from one million players to hundreds of millions of fans.
Lead generation and pipeline automation
A smaller group of deployments focuses on outbound automation. Lusha posts a 300% increase in outbound leads and a 10× increase in conversion, while Sommo generates 500 to 800 leads per month through automated workflows.
Strategic functions such as campaign planning, media allocation, or brand direction barely appear in the sample.
AI in Financial Services
Financial Services shows broad AI adoption across customer service, fraud detection, reconciliation, compliance, and internal automation.
Three recurring profiles emerge:
Digital challengers moving quickly
Klarna resolves 80% of customer queries autonomously (700 FTE equivalent), N26 automated 70% of targeted processes in its first year, and Chipper Cash hits fraud detection accuracy above 95%.
B2B financial infrastructure providers
Fiserv avoids $10 million in SLA penalties through automation, while Raiffeisen reduced a 30-day query process to 12 minutes. Similar deployments appear across nCino and Vanguard.
Compliance and accuracy appear more frequently as explicit outcomes than in other industries, including Bradesco (95% accuracy, 5% escalation), Campfire (90% reconciliation reduction), and Suncoast (100% automated review coverage).
Hours Saved Across Deployments
Organization-wide
Per person
Figure 4. Several organizations report AI impact directly in hours saved rather than financial metrics.
Types of Impact Observed
Speed is the most frequent outcome. Among 792 impact metrics, time improvements dominate, followed by automation and adoption; cost reduction and revenue appear less often.
Speed and time savings account for 14.8% of outcomes, versus 5.1% for cost reduction and 3.9% for revenue or growth. Organizations use AI mainly to accelerate workflows, not to cut costs or drive new revenue.
This distribution challenges the common assumption that AI adoption is driven primarily by cost cutting.
Hours Saved Across Deployments
Many organizations measure AI impact in hours saved rather than dollars. Examples include Notion (2.3 million hours), administrative workflows (1.6 million hours), Microsoft developers using GitHub Copilot (550,000 hours), Shinhan Bank (1.2 million hours annually), and Gordon Food Service (20,000+ hours per month).
Other improvements include SoftBank (4,500 FTE-equivalents annually), RMIT University (17,000 projected hours), and Cisco (1,500+ engineering hours per month).
Financial Claims: Range and Concentration
Most financial outcomes relate to cost savings rather than revenue generation.
Examples include $150 million in estimated time savings from GitHub Copilot at Microsoft, $10 million in avoided SLA penalties at Fiserv, and $1.3 million in fuel cost savings at Massey Services. Additional deployments show staffing cost savings in the $1–2.2 million range and productivity gains of approximately $18,000 per lawyer annually.
Financial Claims with AI
Figure 5. Reported financial impact from AI varies by several orders of magnitude, from thousands to hundreds of millions
Most Prominent Categories in AI Tooling
Machine learning platforms account for 11%, followed by CRM and sales platforms (10%), data platforms (7%), agentic management systems (7%), and large language models (7%). These categories reflect infrastructure trends, not competing tool types.
Deployment Modes
Deployments follow three approaches:
- Platform-first (47%): AI capabilities embedded within enterprise software such as Harvey, Agentforce, Moveworks, and Writer, where model selection is abstracted from the user.
- API-first (31%): Direct integration with foundation models (Claude, ChatGPT), typically in smaller tool stacks and engineering-led environments.
- Hybrid (22%): Deployments combining direct model access with orchestration and enterprise data platforms.
AI Tooling Landscape
Figure 6. ML Platforms, CRM & Sales, and Data Platform are the most popular categories
Emerging Architecture Patterns
Most deployments follow a three-layer architecture: foundation model (for example Claude, ChatGPT, Bedrock, or Gemini), orchestration platform (such as Agentforce, LangChain, Make, or N8n), and enterprise data platform (including Databricks, Snowflake, or Salesforce Data Cloud).
In many cases, tool categories translate into infrastructure layers together rather than standalone solutions.
Although naming conventions are still evolving given the pace of innovation, many subcategories now fall under the agents category. Chatbots are a clear example of this.
SMB and enterprise stacks overlap little. Smaller environments often combine tools such as ChatGPT, Make, and n8n, while enterprise deployments rely more frequently on platforms such as Claude Enterprise, Agentforce, and Databricks.
Closing Observations
Across 200 documented deployments between 2025 and 2026, AI adoption is concentrated in Operations, Software Engineering, and content production. Technology and Financial Services companies are already reporting strong results in time savings and automation, while financial impact is more often framed as cost savings than revenue growth.
Most implementations follow a layered architecture that combines foundation models, orchestration platforms, and enterprise data infrastructure. Organizations integrate AI first where workflows are structured, measurable, and operationalizable.
Applied maintains a living map of enterprise AI deployments across industries and business functions. The full set of use cases is available here and will continue to expand as adoption evolves.