THEME 2: Fix the Work Before Scaling the AI

Core insight

AI will not repair poor work design. It will often make poor work faster.

This theme captures a strong warning from the European Executive Roundtable. Participants recognized the potential of AI, but they also emphasized that AI should not be applied to broken processes, unclear workflows, weak data or poorly defined roles.

One comment captured this sharply: “Don’t bolt on AI on broken work, you will just get faster chaos.” This is one of the strongest statements in the data set because it challenges the assumption that AI automatically improves work. Participant notes suggest that if FM leaders apply AI to badly designed processes, they may simply accelerate confusion, inefficiency, duplication or poor decision-making.

Subthemes, focus & example comments
Subtheme
Focus
Example comments
2.1 AI amplifies existing work design
Weak processes can become faster and more chaotic if AI is applied too early.
“Don’t bolt on AI on broken work, you will just get faster chaos.”
2.2 From pilots to purposeful implementation
AI experimentation needs structure, ownership, learning and follow through.
“95% of the pilots failed”; “For routine AI is a gamechanger…but we are in the exploring phase.”
2.3 Role-based AI capability
AI use should be designed around real FM roles, tasks and decisions.
“AI must be role based to be more impactful”; “Changing skillset”; “Different tasks & competencies.”
2.4 Prompting as work design
Good AI outputs depend on clear context, intent and business logic.
“Write a clear and focused prompt connected with business objectives.”
2.5 Beyond automation
AI should not be reduced to automating current processes. It should support better ways of working.
“Processes are not just automated”; “Remove boring, unproductive tasks.”
Why this matters

Many organizations are experimenting with AI through pilots. This is understandable. AI is evolving quickly, and experimentation can help teams understand what is possible. However, the Roundtable data suggests that experimentation alone is not enough.

The comment that “95% of the pilots failed” points to a wider concern. AI pilots can fail when they are not connected to real business problems, when ownership is unclear, when the data is poor, when workflows are not understood, or when there is no pathway from pilot to adoption.

In FM, this risk is especially important because the work is complex. FM spans buildings, assets, people, services, suppliers, workplace experience, sustainability, risk, compliance and operational continuity. AI cannot be applied effectively unless leaders understand how work is currently done, where friction exists, what decisions matter and where human judgment remains essential.

The data also suggests that prompting should be understood as more than a technical skill. A good prompt depends on knowing the context, the desired outcome, the relevant constraints and the business logic behind the task. In this sense, prompting is part of work design. It requires clarity about what the work is meant to achieve.

Implications for FM leaders

FM leaders should review the work before scaling the technology. This means identifying which processes are suitable for AI support, which need redesign, and which should not be automated or augmented without further consideration.

A practical starting point is to examine workflows through a simple work design lens.

Work design question
Why it matters
What is the current process trying to achieve?
Clarifies the purpose of the work before AI is introduced.
Where does the process create friction, delay, duplication or waste?
Identifies where AI may help and where redesign is needed first.
What decisions are being made, and by whom?
Clarifies where AI can support decision-making and where accountability remains human.
What data does the process rely on?
Tests whether the process is ready for AI support.
What human judgment, experience or relationship management is required?
Prevents overautomation of work that depends on context and trust.
Which tasks are repetitive, low value or suitable for augmentation?
Helps identify practical use cases.
What would success look like if AI improved this process?
Links AI adoption to measurable outcomes.

FM leaders should also avoid treating AI implementation as a series of disconnected pilots. Pilots should be designed with clear ownership, learning goals, evaluation criteria and a route to scale if successful.

A purposeful AI pilot should answer five questions:

Pilot question
Purpose
What problem is the pilot solving?
Keeps experimentation focused.
Who owns the pilot?
Creates accountability.
What workflow is being tested or redesigned?
Connects the pilot to real work.
How will people be involved and supported?
Builds adoption and trust.
What will determine whether the pilot should scale, pause or stop?
Prevents pilots from becoming open-ended experiments.

FM leaders should also make AI capability role specific. The way a facilities director uses AI will be different from the way a maintenance planner, workplace experience manager, sustainability lead, helpdesk operator or supplier manager uses AI. Generic AI training may raise awareness, but it will not be enough to change practice.

Theme 2 summary

The second theme shows that AI cannot be separated from work design. Participants warned against applying AI to broken processes and emphasized the need to understand roles, tasks, prompts, decisions and workflows before scaling AI.

The message for FM leaders is clear: do not bolt AI onto broken work. Redesign the work first, then use AI where it can improve outcomes, remove friction, support better decisions and create value.

International Facility Management Association (IFMA) supports over 26,000 members in 140 countries. Since 1980, IFMA has worked to advance the FM profession through education, events, credentialing, research, networking and knowledge-sharing.