Research

Agentic AI:
from copilot to autopilot

AI is evolving from individual assistance tools into interconnected systems that can plan, coordinate and execute increasingly complex tasks. The latest Lünendonk study shows how organisations are approaching this shift and which capabilities are becoming critical for scaling AI agents across the enterprise.

As AI adoption matures, focus shifts to secure integration and scalable architecture: 98% are investing in data quality and integration, while 94% invest in orchestration platforms and IT infrastructure.

The study

Status quo, potential, and next steps from the perspective of executives

Since the breakthrough of generative AI, companies have experimented with new applications, introduced copilots and embedded AI into individual workflows. AI agents take this development further: they can break down complex goals into individual steps, combine information from different sources, interact with software applications and complete defined tasks with a growing degree of autonomy.

The Lünendonk study, “Agentic AI: from copilot to autopilot”, examines how far organisations have progressed along this path. Based on a survey of CIOs, IT leaders and business unit managers in German-speaking countries, it shows where AI is already being used, where companies are investing next and what is required to move AI agents into productive use at scale.

Key findings at a glance

Trends you should keep an eye on

Scaling tops the agenda: 96% expect AI to boost efficiency and reduce costs.

AI adoption continues to advance: 43 percent of companies plan to deploy individual AI agents, and 96 percent expect AI to increase efficiency and reduce costs. Yet only one in four AI projects successfully transitions from proof of concept to production. As the number and complexity of AI applications grow, building the structures to move them reliably into production and scale them across the enterprise is becoming increasingly important.

This is the approach Siemens Healthineers is taking with Cerebra. Together with Reply, the company has created an AI agent factory that provides teams with a shared environment for developing, deploying and scaling specialised agents. Reusable components and established standards allow new agents to build on a proven platform instead of starting from scratch with every use case.

AI Agents open new paths for legacy modernization: 89% see strong potential.

Existing IT landscapes are emerging as a particularly relevant field for agentic AI. 89 percent of respondents see significant potential in using AI agents to automate the analysis and documentation of legacy systems. Agents can help organisations understand mature software landscapes faster, creating a stronger basis for maintenance and modernisation.

For Audi, Reply developed a multi-agent system that draws on years of architecture documentation, tickets and process knowledge from the "Used Car Platform". Specialised agents analyse existing code, support implementation tasks and conduct quality reviews, making accumulated knowledge directly available within the software development process.

Data and platforms provide the foundation for AI at scale: 98% are investing in data quality and integration.

Scaling AI depends on the foundations underneath it. 95 percent of organisations identify data quality and availability as a key challenge, while 96 percent see a need to improve their technological prerequisites.

Investment priorities reflect this: over the next two years, 98 percent plan to improve data quality and integration, while 94 percent intend to invest in platforms and tools for AI development and orchestration. The focus is therefore moving towards reusable data, technology and operating environments that can support AI applications beyond individual projects.

Governance brings structure to enterprise AI: 62% plan a central catalogue of approved AI agents.

As the number and autonomy of AI agents increase, organisations need mechanisms to control which technologies and agents can be used, how they are evaluated and where they can operate. At the same time, companies want to retain technological flexibility: 88 percent consider dependence on individual AI providers a risk.

Central governance is gaining importance as a result. 62 percent plan to establish a central catalogue of approved AI agents, compared with just 11 percent that have such an approach in place today. Reply addresses these requirements through provider-independent architectures and governance mechanisms for AI models and agents. The Reply Model Factory, for example, provides a framework for managing models throughout their lifecycle, including evaluation, versioning, quality controls and governance, helping organisations scale AI while retaining control over their technology landscape.

Business value becomes the measure success: 51% expect AI initiatives to achieve full payback within a year.

AI initiatives are facing growing pressure to demonstrate their economic value early. 51 percent of executives expect planned or ongoing AI initiatives to achieve full payback within twelve months, while eight out of ten organisations already use defined KPIs to assess cost-effectiveness.

Reply therefore connects AI development with business value from the outset: use cases are prioritised according to their expected impact, measurable KPIs are defined early, and performance and costs are continuously evaluated in production. This creates a basis for deciding which AI applications should be scaled, optimised or reconsidered.

All results

Benchmark your AI roadmap
against the market

The full report shows how AI maturity, deployment scenarios and investment priorities vary by industry and company size. Explore where other organisations stand, which capabilities they are building and what this means for the production-grade deployment of AI agents.

Achieve success with Reply

From individual AI use cases to scalable agentic systems

Reply helps organisations turn AI into a scalable enterprise capability. Our end-to-end approach connects business value, technology and governance: from identifying high-impact use cases and defining measurable KPIs to engineering specialised agents, establishing the required data and platform architecture, integrating agents securely into enterprise processes, and continuously measuring and optimising their performance in production.

Case Study

How Audi is redefining software development for the “Used Car Platform”

Comsysto Reply built Audi’s “Used Car Platform” and now enhances it with a multi-agent workflow, letting developers focus on progress instead of process.

Case Study

Munich Airport: Computer Vision turns cameras into a smart control system

47 million passengers a year, thousands of whom rely on wheelchairs: Axulus Reply delivered an AI-based solution that tracks where one is available in real time.

Case Study

Siemens Healthineers leverages “Cerebra”: Knowledge platform and AI agent factory in one

“Cerebra” transforms complex data into actionable insights for marketing and sales, while also enabling the rapid creation of new AI agents.

Turn the study insights into your AI roadmap

Find out where your organisation stands, which use cases offer the greatest potential and what it takes to scale AI agents securely and economically across your business.

Lünendonk

Lünendonk & Hossenfelder has been analysing European business-to-business (B2B) service markets since 1983. The market researchers focus on the following sectors: management and IT consulting, auditing, tax and legal consulting, facility management and maintenance as well as personnel services (temporary employment, staffing). For decades, the market research and consulting company has been publishing the "Lünendonk lists and studies", which are regarded as market barometers.

Agentic AI:from copilot to autopilot

Status quo, potential, and next steps from the perspective of executives

Trends you should keep an eye on

Scaling tops the agenda: 96% expect AI to boost efficiency and reduce costs.

AI Agents open new paths for legacy modernization: 89% see strong potential.

Data and platforms provide the foundation for AI at scale: 98% are investing in data quality and integration.

Governance brings structure to enterprise AI: 62% plan a central catalogue of approved AI agents.

Business value becomes the measure success: 51% expect AI initiatives to achieve full payback within a year.

From individual AI use cases to scalable agentic systems

How Audi is redefining software development for the “Used Car Platform”

Munich Airport: Computer Vision turns cameras into a smart control system

Siemens Healthineers leverages “Cerebra”: Knowledge platform and AI agent factory in one

Lünendonk

Agentic AI:
from copilot to autopilot