Personal AI agent set up for one person, not for a job title
A chat window in your browser knows nothing about your writing style, your clients or what is urgent this week, so every conversation starts with you explaining the same things again. An agent configured for one person already holds that context and starts at the task.
- 2 weeksfrom the interview to an agent that writes in your voice
- 1 personwe configure it around a named person, not around a job title
- 30 minutesa free call after which you know whether this makes sense in your case
What this actually is
What is a personal AI agent, and why is browser chat not enough?
A language model in a browser tab answers what you type into it. On its own it knows nothing about you beyond what you have just pasted in: not your writing style, not your clients, not which of the twenty open matters is on fire today. It works well on one-off tasks and badly on the work you do every day.
Hence the ritual anyone who has tried it knows. Before you get anything worth sending, you paste the thread, explain who the client is, remind it what tone you use with them. You do it every time, because the previous conversation left nothing behind.
The effect is that the tool saves time on writing and takes it back on explaining. On short tasks it breaks even. On longer ones you are left with text you rewrite anyway, because it reads like it came out of a generator.
- 2 to 3 hours a day on email, summaries and hunting for context
- Drafts that get rewritten because they do not sound like you
- The same client history explained from scratch in every conversation
- What you know about your own matters stays nowhere except in your head
Where the line runs
Chat tool or AI agent: where does the difference actually sit?
A chat tool waits to be told. It answers, it summarises, it prepares a draft, but every move starts with you and ends with your decision. An agent is given a goal and picks the steps itself: it checks the data, runs the operation, comes back with a result. The difference is not how clever the answer sounds. It is who holds the wheel.
For someone running a business that is the whole difference in value. A chat tool helps you do your work faster. An agent does part of that work instead of you, including at hours when you are nowhere near a keyboard. The first shortens your day by a quarter of an hour. The second takes whole tasks out of it.
The same split now runs through the industry: Anthropic describes it in Building effective agents as the difference between a workflow steered from above and a system that decides its own next steps. We build the second kind, and that is why we say agent. This page is about an agent configured for one person. The agent for a whole team, wired into company systems, is BOND.
When the job is specific to one of your processes and spans several systems, we build a custom agent for it as an AI agent development company.
What you get
What an agent configured for one person does
Writes in your voice, not in generator English
It learns from your real correspondence: sentence length, how formal you are, the phrases you never use. The email comes out ready to read and send, not ready to rewrite.
Trained on your own material, not on templates
Knows your clients and your open matters
Who is who, what stage the conversation is at, what you agreed last time. You do not explain it twice, because the context is part of the configuration rather than part of the prompt.
Context stays between conversations
Knows what is urgent this week
Your scope of work and the 5 to 10 matters currently on your desk go into the configuration, so a question about the plan for the day returns your own list instead of an article about time management.
Configured around a person, not around a job title
Reaches company data when it has to
We can connect it to BOND, and then it answers from company systems. Without that connection it works only on your personal context, and that is a complete rollout too.
Runs on its own or alongside the company agent
Your context stays yours
The material we configure the agent on does not reach any external training service. The configuration is dedicated and available only to you.
No training on your data in third party services
And in your case
Will a personal AI agent cope with the work you actually do?
Four sentences come back in every conversation. We answer them before anyone signs anything.
Will it really sound like me?
That is the one thing you can check on the spot. On the free call we show the agent working on a sample context, with none of your data shared. If the text sounds artificial, you see it in the first 5 minutes rather than after signing a contract.
I do not want to hand over my whole inbox.
The configuration needs examples you pick yourself, not access to your mailbox. That material does not reach any external training service, and the configuration is dedicated and available only to you.
I am not technical.
You do not need to be. Your part is one hour of interview and then judging the drafts the agent produces. Configuration, integrations and tuning are on us, and nothing gets installed on your side.
We already have ChatGPT in the company.
Good, because then you know exactly where this work stalls. An agent for one person does not replace the chat tool. It takes away the context explaining you do before every single prompt.
- Clients and sales
- HubSpot, Pipedrive, Salesforce, LiveSpace
- Work and documents
- Google Workspace, Microsoft 365, Slack, Notion, ClickUp
- Data and reporting
- BigQuery, Google Sheets, Google Analytics, SQL databases
- Invoicing and stock
- Fakturownia, Xero, sevDesk, BaseLinker
Browser chat, an off-the-shelf AI tool and a personal AI agent
Three things people use interchangeably in conversation. They differ in what they remember and in what they are allowed to do.
| What we compare | Browser chat | Off-the-shelf AI tool | Agent for one person |
|---|---|---|---|
| Knows your writing style | Only if you paste examples | Usually not, it runs on templates | Yes, from training on your correspondence |
| Remembers between conversations | Within a single session | Remembers settings, not context | Yes, context is part of the configuration |
| Sees company data | No | Rarely, through ready-made plugins | Optionally, through BOND |
| Carries out tasks by itself | No | No, it waits to be told | Yes, within the scope agreed with you |
| Who it is for | One-off tasks | Small repetitive jobs | Business owner, manager |
| Time to launch | Straight away | Straight away | Around 2 weeks |
The columns do not rule each other out. The usual order is browser chat first, then an agent for one person, and once that proves itself, an agent for the whole team.
How we do it
How to build your own AI agent, step by step
3 steps, each one self-contained. After the second you know whether the agent sounds like you, and that is the point where you decide about the rest.
Interview
60 minutes on how you work: what you write most often, to whom, in what tone, what is genuinely urgent and what is merely loud. This is the material a browser chat window never gets.
Configuration and testing
The agent learns on your real examples: emails, summaries, follow-ups. For 2 weeks you compare its drafts with what you would have written yourself, and we tune whatever drifts.
Extension
The scope grows with your work: new contexts, new integrations, an optional connection to company data through BOND. Billed as a monthly retainer you can stop at any point.
How the day changes
What you do before, and what you do after
Before
- Before you ask for an email, you paste the thread and explain who the client is
- The draft gets rewritten because it reads like it came out of a generator
- You assemble the weekly summary by hand out of several places
- The plan for the day lives in your head and in three separate lists
After
- You write one sentence of instruction, the agent already holds the context
- The text comes out in your tone and goes out after a read, not a rewrite
- The summary arrives finished and all it needs is your comment
- A question about priorities returns your own list of matters, not generalities
First useful drafts after 7 days, the voice tuned in by day 14. The retainer keeps it in step with changes in your work.
Before you decide
What you are risking, and what you are not
There is no board to convince here. There is a harsher judge instead: you, a month from now, looking at a tool you stopped opening after 2 weeks. That is how most private AI rollouts end, and that is the real risk in this decision, not the cost.
So the whole process is arranged so that the moment you can walk away falls BEFORE you pay for anything ongoing, rather than a year into a subscription.
- After a 30 minute call you know how it reads in your own writing, and you have paid nothing
- After 14 days of testing you compare the agent's drafts with your own, and only then decide about the retainer
- The retainer is separate from the rollout, so dropping it does not undo what you already have
- The configuration is yours: the description of your style, context and priorities stays with you even if we stop working together
What it costs
What does a personal AI agent cost?
We quote the rollout individually, because the price follows the scope of your work and how much context has to be collected. An agent for someone running 2 projects and an agent for an owner with 5 companies is not the same job.
Maintenance and extension go on a separate monthly retainer, quoted after the rollout. We keep it out of the rollout price on purpose, so it can be dropped once the agent stops being needed. Model usage you pay for directly to the provider, with no margin from us: EUR 0.005 to 0.07 per query at provider list prices.
The quote comes after the free call, not on the spot. A price named before anyone knows the scope would be a guess in one direction or the other.
- Most popular
One to one setup
Quoted individually
after the free call, not from a price list
A self-contained piece of work: interview, configuration and 2 weeks of tuning. It ends with an agent that writes in your voice.
- Interview about the way you work
- Configuration around your clients and your priorities
- Training on your own real material
- 2 weeks of testing and tuning
Monthly retainer
Agreed individually
optional, quoted after the rollout
For people whose scope of work shifts during the year and whose agent has to keep up with it.
- Extending the context the agent holds
- New integrations, including the connection to BOND
- Tuning as the work itself changes
The first 30 minutes are free and end with the agent shown on a sample context, with none of your data shared. If you decide afterwards that this makes no sense in your case, it has cost you nothing.
Frequently asked questions
How is a personal AI agent different from a chat tool?
A chat tool waits to be told and prepares a draft for your decision. An agent is given a goal and takes the steps itself, including when you are not at the keyboard. The first speeds your work up, the second takes whole tasks out of it.
Which AI model is the best one for this?
There is no single answer, because the difference comes from the context the model was given, not from the model itself. The same model with no context about you writes text you rewrite, and with that context writes text you send. Which is why a rollout starts with an interview, not with picking a tool.
How is this different from ChatGPT?
ChatGPT does not know your style, your clients or your priorities, so every conversation starts with you explaining the context. We configure the agent once, for one named person, and that context stays in place between conversations.
How do you build a personal AI agent and how long does it take?
We start with 60 minutes of interview about how you work and what you write most often. Then configuration and the first tests on your own real examples. The whole thing takes around 2 weeks, and anything beyond that goes on a monthly retainer.
Does it work in languages other than English?
Yes. We configure the agent on your own material, so it learns your English, your Polish or your German rather than a translation out of one into another. On the free call we show that on a sample context.
Does my data go into training the model?
No. The material we configure the agent on does not reach any external training service. The configuration is dedicated and available only to you, and you can have it removed at any point.
Can the agent use company data?
Optionally yes, through BOND, the AI agent wired into company systems, which answers from company data in 5 to 10 seconds. Without that connection it works on your personal context. The order we recommend is the agent for one person first, BOND second.
Who is this not for?
For someone who uses AI once a week for a one-off task. A browser chat window is enough in that case, and configuring the context will not pay for itself. We say so on the call instead of selling a rollout.
Start with thirty minutes
On the call we show the agent working on a sample context, with none of your data shared. You tell us what you re-explain to AI every single day, and we tell you whether it can be taken off you and how long that takes.
Calls are run by Paweł Pasterak, the founder of JustAutomate, with no reception and no helpdesk. Which is why there are only as many slots as he has hours in the week.
Book a free consultationPS. If you explain the same context to AI every day, that is not a problem with the tool. It is a missing configuration. The call is free and ends with a straight answer on whether this pays off in your case.