"We should do something with AI" is the sentence most projects start with, and the reason many end as a pilot. Start with the tool and you go looking for a problem afterwards. It works the other way round: find the process that eats hours or costs enquiries every week, and build a system exactly there. The following five don't come from studies but from my own operation. The page AI marketing use cases (German) lists all eight; here are the five that pay off fastest.
1. Form to CRM: no more enquiries in an inbox
The bottleneck: Enquiries land in an inbox, get retyped, forwarded, forgotten. Nobody knows which channel they came through.
The system: Every enquiry lands in the CRM with source, context and status, including automatic pre-qualification. Sales works from a list, not an inbox.
Why it pays off right away: It is the only case on this list that rescues revenue directly. An enquiry answered on Monday instead of Saturday evening has often already gone to a competitor. Setup: a few days.
2. Reporting that writes itself
The bottleneck: Every Friday someone copies numbers from ad accounts, analytics and CRM into a presentation. Three to four hours, every week, and in the end the report still doesn't answer the management's question.
The system: Numbers from channels, CRM and website flow automatically into a report that says three things: what did it achieve, what does an enquiry cost, where do we turn things down?
Why it pays off right away: The saved time is there from week one. The bigger effect comes afterwards: if you see weekly what a qualified enquiry costs per channel, you switch off the wrong campaigns. How much money sits there is covered in Wasted spend: 8 typical budget eaters.
3. A content system with approval
The bottleneck: Everyone knows regular content works, nobody has time for it. The blog has three articles from last year.
The system: Editorial plan, AI drafts in two languages, human approval, automatic publishing including the sitemap. This blog is made exactly that way: I plan the topics, the drafts come from the system, I approve them, and only then does an article go live.
Why it pays off right away: Half a day per article becomes twenty minutes of approval. The approval is what matters: without it you get interchangeable text, with it a channel that has a stance and gets quoted by Google and AI systems.
4. Proposal automation
The bottleneck: A good proposal costs a day. So many prospects get an off-the-shelf PDF, and it shows.
The system: At the push of a button a lead becomes an individual proposal page in the recipient's own design, with an analysis of their starting point, the approach and matching references. A person checks and adds instead of starting from zero.
Why it pays off right away: What used to take a day takes minutes, and a proposal that shows the recipient's own situation gets read. It pays for itself with the first deal won.
5. Lead radar: signals instead of cold outreach
The bottleneck: In many companies, winning new customers means buying lists and calling through them. Lots of effort, few hits.
The system: A system watches relevant sources, for example job postings that point to a need, scores every signal with AI for fit and priority, and files qualified leads with reasoning in the CRM. Noise becomes a sorted list, daily, without manual searching.
Why it pays off: Sales talks to companies that have a reason right now. This case takes longest to show a first result because a sales cycle sits in between, which is why it is number five.
What the five have in common
None of these cases is "AI for the sake of AI". Each replaces recurring manual work at a point where mistakes cost money, and each keeps a person in the loop: approval before publishing, review before the proposal, a decision before the call. That also explains why the approach works at scale: at Engel & Völkers, AI-supported campaign work produced 90 percent more qualified leads at 30 percent lower cost per acquisition. Not because the AI was smarter than the team, but because the team could test more variants.
Three rules so it doesn't stay a pilot
- Start at the bottleneck. Not with the most exciting tool, but with the process that costs the most hours or enquiries every week.
- Define approvals. Who checks what before it goes out? Systems without approval get switched off after the first mistake.
- Anchor it in the daily routine. A system that requires opening a separate tool doesn't get used. It belongs where the team already works: in the CRM, the inbox, the calendar.
The honest conclusion
"Right away" means weeks here, not days, and not every case fits every company. If you get ten enquiries a month you don't need a lead radar, you need a form that writes into the CRM first. Start with one, measure it in hours or euros, and only take the next when the first runs in daily work. If you want to know how your website looks to AI systems in the meantime: the AI visibility check takes 30 seconds.