What Building 20+ AI Systems for One Client Taught Us About Enterprise AI
Before InBusiness existed as a standalone unit, we had one client: Botilia. Over two years, we built more than 20 AI systems for them. Covering analytics, search, CRM, marketing automation, and customer service. Not as a consulting exercise. As the people responsible for making it work.
That experience shaped everything about how we operate today. This article is a candid account of what we learned.
The first lesson: don’t start with the technology
Our first instinct was to build the most technically impressive system we could. An analytics platform with every feature, every integration, every dashboard. The result was a tool that did everything and that nobody used.
The problem wasn’t the technology. The problem was that we hadn’t mapped Botilia’s actual workflows before building. The marketing team didn’t need 47 dashboards. They needed three numbers, updated daily, with a recommended action attached. That insight became the foundation of InsightFlow, our analytics tool that returns action plans, not just charts.
Every project we’ve done since starts the same way: map the process first, understand where the pain actually is, then build only what’s needed. That insight became InsightFlow, and we wrote separately about how it pairs with the enterprise AI search engine we built for the same client.
The second lesson: AI that doesn’t integrate is AI that dies
We built a beautiful customer service agent for Botilia. It handled inquiries, resolved tickets, escalated correctly. But for the first month, the support team kept switching back to their old tools. Why? Because the agent lived in a separate tab. It didn’t connect to their existing ticketing system, their CRM, or their knowledge base.
Integration isn’t a feature. It’s the entire point. When we rebuilt the agent to live inside the tools the team already used, adoption went from 15% to 80% in two weeks. No training sessions needed. No change management workshops. Just put the AI where the work already happens.
The third lesson: custom beats generic, every time
We tried using off-the-shelf AI tools early on. Every time, we hit the same wall: the tool was built for a general use case, but Botilia’s processes were specific. The CRM had custom fields the AI couldn’t read. The search tool didn’t understand Greek product names. The analytics platform assumed data structures that didn’t match reality.
So we built our own stack. Not because we wanted to, because we had to. InsightFlow, our enterprise search tool, the AI CRM that replaced gut-feeling marketing with predictive segmentation, the service agent: these all emerged from solving Botilia’s specific problems with tools that didn’t exist yet.
Now, when we work with new clients, we don’t start from zero. We start from a toolkit that’s already been tested in production across 20+ deployments. We adapt it to their processes, their data, their team structure. That’s fundamentally different from either buying SaaS or hiring a consultancy to build from scratch, we wrote a separate piece on the economics.
The fourth lesson: the first win has to be fast
Botilia’s leadership gave us room to experiment. But that room had a clock on it. If we hadn’t delivered a visible result within the first six weeks, the project would have been deprioritized.
We learned to structure every engagement around a quick first win. Not the most impactful project, the most visible one. At Botilia, that was automating a weekly reporting process that took their analytics team an entire day. We replaced it with an automated InsightFlow report that generated in under a minute. The time savings were modest in the grand scheme. But the signal was unmistakable: this works.
That early win bought us the credibility to tackle harder problems: predictive customer segmentation, automated content generation, the full IR service agent we’re deploying now.
The fifth lesson: AI needs an owner inside the client
The projects that scaled at Botilia all had one thing in common: an internal champion. Someone who understood the AI well enough to defend it in meetings, escalate issues to us, and push for adoption within their team.
The projects that stalled were the ones where nobody inside the organization felt responsible. We’d deliver the system, train the team, and then watch usage flatline because there was no one keeping the momentum going.
Now we build this into our process. Before we start building, we identify who the internal owner will be. If there isn’t one, we work with leadership to designate one. It’s a prerequisite, not an afterthought.
What this means for how we work today
Every principle we follow at InBusiness traces back to something we learned at Botilia:
Map first, build second. We spend the first weeks understanding your processes, not writing code.
Integration over isolation. We build AI that lives inside your existing tools, not beside them.
Custom on proven foundations. We don’t start from zero, and we don’t sell you a generic product. We adapt our battle-tested stack to your reality.
Quick first win. We pick the project that creates visible value fastest, then use that momentum to go deeper.
Internal ownership. We need a champion on your side. We’ll help you find one.
If you’re considering AI for your enterprise and want to talk to a team that’s done this 20+ times for real, not in theory. We should talk. Book a discovery call at inbusiness.gr.