How Many Employees' Work My AI Agent Does

I am often asked whether an AI assistant really does anything, or whether it is just another toy. I will answer honestly, based on my own experience. In my estimation, my AI agent Jarvis does the execution work of 1.5-2 employees. Let me stress straight away: this is my personal estimate, not a measurement or a study. I compare how much work I would previously have had to delegate to people to get done the same volume that I now get done together with my AI agent.

The problem: execution work takes the time needed for management

In a small company, the manager is often also the salesperson, the marketer, the analyst and the assistant all at once. That was true for me too. Each of these tasks is not complicated in itself, but together they fill the whole day.

The result is familiar to many entrepreneurs:

important things get postponed because urgent ones always come first;
proposals are written in a rush and all look the same;
we post on social media only when there is time left;
procurement opportunities slip by because there is no time to read them.

Hiring one or two more people is a logical solution, but for a small company it means significant costs, training and management. That is why I tried a different route.

What it looks like in the company

Before my AI agent, my typical working week looked roughly like this.

Client research. Before a conversation with a potential client, I browsed their website, public registers and news myself. It was useful, but slow.

Proposals. I took the previous proposal, changed the name and a few paragraphs. The personalisation often remained superficial.

Procurement. Procurement documents are long, and to understand whether it was worth applying at all, I had to read the regulations, the technical specification and the requirements for bidders.

Video and social media. Preparing videos and distributing posts across different platforms took separate time that was hard to schedule.

Calendar and emails. Coordinating meetings, reminders, replies to routine messages. Small things that fragment the day.

None of these tasks could simply be dropped. They are all necessary. The only question was who should do them.

The solution: an AI assistant that does the execution work

My AI agent took over a large part of this execution work. Now the same roles look different.

Client research. My AI agent gathers publicly available information about a company and prepares a short summary: what the company does, what challenges they might have, and what is worth asking in the conversation.
Personalised proposals. Based on the research and my notes after the conversation, my AI agent prepares a draft proposal that fits the specific client's situation, rather than just changing the name in a template.
Procurement analysis. My AI agent reads the procurement documents and highlights the key points: requirements, deadlines, risks and whether our expertise is a match.
Video preparation. From an idea or a text, my AI agent helps prepare video material, which I then review.
Social media publishing. One idea is adapted for different platforms, each with its own tone and length.
Calendar and emails. My AI agent prepares draft replies, helps coordinate meetings and reminds me of things that must not be forgotten.

One principle never changes: I approve every email and every publication myself. My AI agent prepares, I decide. Nothing goes out into the world without my approval.

What my AI agent does not replace

It is just as important to say what my AI agent does not do and should not do:

conversations with clients are led by me, because that is where trust is built;
decisions are made by me, because I am responsible for them;
relationships with clients and partners are human work.

That is exactly why I talk about execution work. My AI agent does not replace a person in the company, it frees the person up for what only a person can do.

What it delivers

The biggest change is not that work gets done faster. The biggest change is that I spend more time where I am needed.

More time for conversations. When the research and drafts are ready, I can focus on the client rather than on preparation.
Better quality. Proposals are more personalised, because personalisation no longer requires hours at a document.
Consistency. Social media and emails no longer depend on whether there was time today.
Fewer missed opportunities. I can evaluate procurements even when the week is overloaded.
Control stays with me. Because I approve every outgoing item, the brand voice and accountability do not change.

It is this overall effect that I estimate as the execution work of 1.5-2 employees. Once again: it is my subjective estimate, based on my daily routine. In another company with different processes, it may look different.

Where to start

If you are thinking about a similar assistant for your business, I recommend starting not with the technology, but with your week.

Write down recurring tasks. What happens every week, takes time, but does not require your decision?
Choose one process. For example, client research before conversations or email drafts. Start with one, not everything.
Set the approval procedure. Decide at the outset what the assistant may only prepare and what it may do independently. I chose to approve everything outgoing myself.
Assess honestly. After a few weeks, look at which tasks have really become easier, and only then expand.
Separate the roles clearly. The assistant does the execution work, while the person leads conversations, makes decisions and builds relationships.

If you would like such an assistant for your business, write to me. Together we will assess which of your processes make sense to entrust to AI and where the person must remain at the centre. More about us: ai-labspace.eu.