Human + AI Assistants: Building a Hybrid EA Stack That Actually Scales Your Week
Authored by: Filip Pesek
There was a period in early 2024 when every founder I spoke to asked the same question: should I replace my assistant with AI?
ChatGPT could draft emails. Scheduling tools could manage calendars. Transcription services could summarize meetings in seconds. The logic seemed obvious – why pay a human when algorithms can handle it?
A year and a half later, most of those founders are back to hiring humans. Not because AI failed, but because they discovered something important: AI is exceptional at processing information and terrible at understanding context. The real advantage comes from combining both.
At DonnaPro, the virtual assistant agency I founded for European CEOs and founders, we have built our entire model around this principle. Our Executive Assistants are real people from across the EU, trained extensively on AI tools. The combination delivers what neither can achieve alone.
Here is how we built that hybrid stack – and three lessons from the process that any founder can apply.
Let AI Handle the Mechanical Layer
The first mistake founders make with AI is trying to automate judgment. They feed client emails into ChatGPT and ask it to respond. The output is grammatically perfect and contextually tone-deaf. A long-term investor gets the same register as a cold inbound lead. A sensitive board matter gets summarized with the same weight as a routine status update.
AI earns its keep when you point it at the mechanical work that sits underneath the judgment calls. At DonnaPro, we have built custom automation workflows using self-hosted n8n, Claude, and various MCP integrations that handle the repetitive infrastructure our EAs used to spend hours on.
One concrete example: when a potential client books a sales call with us, automations run immediately in the background. The system researches the person, pulls insights from LinkedIn, surfaces recent company news, and assembles a complete briefing document for our team. That process used to take 30 to 40 minutes of manual research per lead. Now it happens before anyone opens their laptop.
We use the same approach for onboarding. Fathom records and transcribes every sales call. Our n8n workflows then extract relevant client information from those transcripts and prefill about 80% of the onboarding questionnaire automatically. The client opens the form and most of it is already done – they just confirm details and add what is missing. That saves roughly an hour of manual work on our side per new client and at least 30 minutes of the client’s time.
Similar automations handle invoice routing to billing accounts, preparation of HR documents and demo tasks for candidates, and ongoing SEO analysis for our website. None of these workflows make decisions. They prepare the ground so that humans can make better decisions faster.
Keep Humans on Everything That Touches a Relationship
This is where most AI-first approaches fall apart.
Consider managing a CEO’s calendar. An AI scheduling tool can find open slots and send invitations. But it cannot determine that a particular investor meeting deserves 90 minutes instead of 30, or that two specific calls should be separated by prep time, or that Thursday afternoons are when this founder does their best strategic thinking and should be kept clear.
That kind of understanding comes from working closely with someone over weeks and months. It requires reading patterns, remembering preferences, and exercising judgment that no model can replicate – because it depends on a relationship, not a dataset.
Our rule at DonnaPro is simple: anything that touches a client relationship stays human. AI can draft a routine email, but the EA reviews and personalizes it before sending. AI can transcribe a meeting, but the EA synthesizes what actually matters based on context that exists nowhere in the transcript. AI can compile research, but the EA decides what the founder needs to see and how to frame it.
The client should never feel the AI. They should only feel remarkably efficient human.
Train Your Team to Use AI Strategically, Not Reflexively
The hardest part of building a hybrid stack is not the technology. It is teaching people where to draw the line.
When we first introduced AI tools to our team, the instinct was to automate as much as possible. If AI could do it, why would a human bother? But we are an executive assistant agency whose entire value proposition is human connection. When clients started receiving responses that felt slightly off – slightly too polished, slightly generic – we knew we had gone too far in the wrong places.
Now AI training is built into our onboarding from day one. Every EA learns which tools to use, for which tasks, and – critically – where to stop. They learn to use AI for first-draft correspondence, data compilation, research synthesis, and transcript processing. They learn not to use it for anything where tone, nuance, or relationship context matters.
The distinction sounds obvious but it is surprisingly difficult to maintain in practice. The temptation to let AI handle “just one more thing” is constant. Discipline around that boundary is what separates a hybrid model that enhances trust from one that erodes it.
The Future Is Not AI or Human. It Is AI-Powered Humans.
The businesses gaining the most from AI right now are not replacing people with algorithms. They are equipping skilled humans with AI capabilities – combining technological speed with human understanding.
For founders still debating whether to hire an assistant or subscribe to another AI tool, my answer is the same: do both. Use AI to eliminate the mechanical friction in your week. Use a human to manage everything that requires judgment, context, and trust.
The competitive advantage in the next few years will not belong to whoever has the best AI. It will belong to whoever uses it most strategically – and knows exactly where to let a human take over.
Filip Pesek is the Founder and CEO of DonnaPro (donnapro.com), a managed executive assistant service built exclusively for CEOs and founders across Europe. Filip specializes in building systems that combine AI efficiency with human judgment to help business leaders reclaim their time and focus on strategic growth. Before DonnaPro, he founded Spark, a marketing agency, and authored the bestselling book Pisma za Leona.