Key Takeaways
- AI pays off where work is text-heavy: Programming, writing and customer service are the areas where companies see the largest productivity gains from AI.
- Information drives investment: Simply learning more about AI and about what competitors are doing measurably increases a company's willingness to invest.
- Expectation is not conviction: Around 61 percent of companies already use AI, yet their belief in their own profit from it lags behind their general expectations.
- Fast code needs a security process: Tools such as Lovable build working software from plain language, but speed alone does not answer the security question.
- Meetings can become lasting knowledge: AI meeting agents turn conversations into documented results instead of forgotten notes.
- Structure beats heroism: Automation only frees entrepreneurs from the hamster wheel when the underlying processes are clearly defined.
Introduction
Does AI really make companies more productive, or is the promised gain mostly wishful thinking? A current study from the ifo Institute gives a clear answer for three areas: programming, writing and customer service. This is where the jumps in productivity are largest today.
At the same time, the picture inside companies is mixed. Around 61 percent already use AI in some form. Many expect higher profits from it – but far fewer are actually convinced that their own business will earn more. That gap between expectation and conviction is the real story of this week.
In this article I look at what the findings mean for decision-makers, why information is such a strong driver of investment, and how the topics of the past weeks – secure AI code, meeting agents and structure – fit into the same picture. One principle runs through all of it: technology is rarely the problem. People and processes usually are.
Where AI Delivers the Biggest Productivity Gains
The ifo study "Firms' AI Productivity Beliefs and Adoption Decisions" names three fields where companies see AI paying off most: programming, writing and customer service. That is not a coincidence. All three consist largely of language – code, text and conversation – and that is exactly what today's AI models handle best.
For a managing director, this is a useful filter. Instead of asking "Where could we use AI somewhere?", ask "Where do my people spend the most hours producing or reworking text?" Typical candidates are quotations, customer emails, documentation, internal reports, support tickets and small software tools. These are the places where an AI assistant can take over the first draft and leave the human with review and judgement.
The important word is "draft". Productivity gains come from shifting human effort from writing to checking, not from removing the human altogether. Teams that understand this get faster results with fewer errors. Teams that expect finished work at the push of a button tend to be disappointed and give up early.
Why Information Is the Strongest Driver of Investment
One finding from the study deserves particular attention: new information about AI and about competitors increases a company's willingness to invest. In other words, many firms are not hesitating because AI does not work for them. They are hesitating because they do not know enough about how it could work for them.
That is good news, because a knowledge gap is much easier to close than a technology gap. A short workshop, a well-chosen example from the same industry or a frank conversation with a peer often does more than another pilot project. When decision-makers see how a comparable company saves time in service or documentation, the abstract question "Should we do AI?" turns into the concrete question "Which process do we start with?"
Relevance beats spectacle here. A flashy demo impresses for an afternoon. A realistic example that matches your own processes changes budgets.
The Gap Between Expectation and Conviction
Why do so many companies expect profit from AI but doubt their own share of it? From my work with executives, I see three typical reasons. First, results are rarely measured. If nobody records how long a task took before and after, the benefit stays a feeling. Second, AI is often introduced as a tool rather than as a change in the way work is done. Third, employees receive access but no guidance, so usage stays shallow.
The remedy is simple, though not effortless. Pick one process, measure it, introduce AI with clear instructions and measure again. Conviction grows from experience and visible results – not from press releases about what AI might do one day.
Secure Code, Lasting Knowledge and Clear Structures
The topics of recent weeks fit neatly into this picture. AI-generated software is the most visible example of the programming gains. Tools such as Lovable create working applications from plain-language instructions in a very short time. The open question is security. Speed is impressive, but code that nobody has checked can contain vulnerabilities. A structured review process – security by design rather than security as an afterthought – is what turns fast prototypes into reliable applications.
Meeting agents show the writing gains in a different form. Many meetings end without lasting results because nobody documents decisions and tasks. AI agents can turn conversations into structured, searchable knowledge, so information does not disappear when the call ends.
And then there is structure. Many entrepreneurs still run their business as heroes who personally hold everything together. Automation can relieve them – but only when processes are clearly defined. Automating chaos simply produces faster chaos. The same applies to regulation: the EU AI Act requires companies to train employees in the use of AI, and those who organise processes and training early meet the requirement with far less stress.
Mentioned Tools & Resources
- ifo Institute study: "Firms' AI Productivity Beliefs and Adoption Decisions" – read the publication.
- Lovable: An AI tool that builds working software from plain-language instructions; best combined with a structured security review.
- AI meeting agents: Tools that turn meeting content into documented, searchable knowledge.
- EU AI Act: The European regulation that, among other things, requires companies to train employees in the use of AI.
- TJ's Technology Tuesday: A weekly 15-minute digital impulse for executives, every Tuesday from 12:30 to 12:45 (CET), with recordings available on LinkedIn.
Frequently Asked Questions
Which business areas benefit most from AI?
According to the ifo study, the largest productivity gains currently appear in programming, writing and customer service. All three are language-heavy, which is where today's AI models are strongest. The prerequisite is a clean integration into existing workflows.
Why are profit expectations from AI higher than actual conviction?
Many companies expect more from AI than they currently trust it to deliver for themselves. Conviction usually grows only with hands-on experience and measurable results. Measuring a process before and after introducing AI is the fastest way to close that gap.
How does information increase willingness to invest in AI?
When companies learn how AI is used and what competitors are already doing, they recognise their own potential more clearly. The study shows that such information alone raises the willingness to invest in technology and training.
Is code generated by tools like Lovable insecure?
Not automatically, but it can contain vulnerabilities if nobody checks it. The speed of these tools is impressive. A structured review process and security requirements defined from the start are what make the resulting software reliable.
How can AI help to follow up on meetings?
AI meeting agents turn conversations into structured notes, decisions and tasks. Information is no longer lost after the call, results stay traceable and knowledge builds up over time instead of disappearing.
Why does structure matter more than heroism for entrepreneurs?
Companies that depend on individual heroes are fragile and exhausting to run. Clear structures relieve leaders and employees alike. Automation only works when processes are precisely defined – otherwise it just speeds up the chaos.
Conclusion
The message of this week is encouraging. AI delivers real productivity gains, especially in programming, writing and customer service. What holds many companies back is not the technology but missing information, missing measurement and missing structure.
My suggestion: choose one text-heavy process in your company, measure how long it takes today and test an AI assistant on it for four weeks. As a personal IT coach for executives, I support companies with keynotes, training and C-level sparring on iPad, Microsoft 365 and AI – so that the gain from AI shows up in your own figures, not only in the studies.


