Agents

Companies Must Prioritize Organizational Preparation Over Technical Deployment for AI Success

November 18, 2026 · 3 min read

A futuristic car driving on a bridge over a body of water
TSD Studio / Unsplash

Successful AI integration in companies depends more on supporting and accompanying employees through the transformation than on technical aspects.

Many organizations face a dilemma where AI experience is fragmented. Some employees develop their own applications while others are unsure which tools are permitted. This creates a gap between executives expecting higher productivity and employees tasked with verifying AI results without fully understanding the technology's limits. These issues cannot be solved by purchasing additional software licenses.

Cosima von Kries of addhucate observes that companies often implement AI functions in existing systems before preparing organizationally. This leads to delayed questions regarding data structures, governance, access rights, and appropriate inputs. AI is also seen as a way to manage personnel shortages caused by demographic change.

Julian Staub of Lufthansa Industry Solutions warns against generic ideas of AI readiness, noting that a software company has different requirements than a medium sized business or one with many blue collar workers. He argues that companies must identify their specific gaps in processes, data, and competencies before acting.

Kai Waehner of Kestra emphasizes that individual AI use does not equal organizational capability. He distinguishes between personal knowledge, such as a prompt in a chat window, and company knowledge, which consists of versioned and approved workflows. To achieve this, business departments should describe the logic while IT handles operations.

Markus Böhm of Mimacom suggests using experienced AI users as multipliers to teach teams through concrete cases, provided they are given dedicated time. He asserts that management must lead by example and ensure all employees have regulated access to tools suited to their roles.

Steffen Märkl of Cloudera suggests integrating AI directly into daily professional tasks to avoid employees turning to unapproved tools due to bureaucratic hurdles. He states that AI agents should only act independently when decisions are traceable and reversible.

Regarding the workforce, Böhm and Staub suggest involving affected departments in developing solutions to address concerns about job losses. Waehner notes that employees often fear their knowledge being extracted for automation without their involvement in the design process.

Responsibility must be reassigned as AI takes over tasks. Böhm calls for a responsible person from the business department for every agent. Matthias Reichle of Avanade suggests that human approval remains necessary for critical decisions with high potential damage, while lower risk tasks can rely on monitoring and auditability.

Tobias Metsch of Axis warns that human oversight must not be a formality. Supervisors need sufficient information, qualification, and time to intervene. Märkl adds that a human in the loop is mandatory when systems move from information preparation to writing actions on production data or sensitive transactions.

Concerns extend beyond job loss to the fear of not keeping pace with development. Böhm notes that entry level employees might lose learning opportunities if AI takes over simple tasks. Metsch also mentions concerns from works councils regarding the use of AI for employee surveillance.