AI agents · Warsaw
I build AI agents on your own data: they qualify the lead, act in your CRM, calendar and invoicing, and run on your website, in Telegram and WhatsApp, with an admin panel and metrics. Production-ready, not a toy.
From €890live in 1-2 weeks
Tech stack: OpenAI / GPT · Anthropic / Claude · Google Gemini · Function calling · Vector DB (pgvector / Pinecone) · Firestore · AmoCRM API · Telegram Bot API · Wappi (WhatsApp) · Node.js / Express
The problem
Three places where a business loses leads before a manager even says hello. One agent closes all three.
A client writes in the evening and gets a reply in the morning. By then they have already booked with whoever answered in a minute. Response speed beats price: the agent replies in seconds, at 3 am too.
80% of questions repeat: how much, how long, how do I book. Paying a person to type the same answers is expensive - and they also take vacations and sick days. The agent takes the routine, people keep the deals.
SaaS bots with "pick a menu option" don't know your prices or services - the client gets annoyed and closes the chat. An agent trained on your data answers the actual question, like an employee who memorized everything and never gets tired.
See it live
One conversation, zero manual entry. While the bot chats on Telegram, the deal moves through your CRM pipeline and every answer lands as a field - in real time.
Formats
Three typical scenarios. Not sure which one is yours - tell me on a call where you lose clients. If a simple automation from €550 is enough, I'll say so.
Answers on FAQ and order status and can carry out one action: book a slot, create a lead, tag the deal.
For teams that want one process off their hands
Qualifies the lead, books the appointment, keeps the CRM card updated - like a manager, just around the clock.
For sales teams and client booking
Works with several systems at once: CRM, inventory, calendar, documents. Doesn't just answer - gets things done.
For companies with complex processes
Industries
An agent earns its keep where questions are frequent, repetitive and the first line of support is expensive. These are the sectors where that shows up in the numbers.
First-touch triage: the agent asks about status and documents, matches the procedure and hands the lawyer an already-qualified enquiry.
Booking, visit preparation, prices and doctors. The agent gives no medical advice: anything ambiguous goes to reception.
Matching listings by budget and district, answers on paperwork and mortgages, viewings booked right inside the chat.
Product matching, order status, returns and abandoned carts. The agent knows your stock and won't promise what isn't there.
Booking, check-in, house rules and extras in four languages, including the night hours when the front desk doesn't answer.
Route-based quotes, shipment status, paperwork and repeat orders from regular clients.
Programme selection, schedule, payment and reminders. The agent gets people to the first lesson who would otherwise vanish after enquiring.
First line on repetitive tickets, handing off to a human on anything unusual. The agent passes the full context, so nobody starts from scratch.
Inbound qualification: the agent works through a script, filters out poor fits and books a meeting in the rep's calendar.
What I ship most often, with the real timeline and the real stack. The price is for the build, with no mandatory retainer.
Answers questions from your documents, price lists and contracts with a link to the source, and says it does not know when the answer is not there.
1-2 weeks · GPT-4o-mini, pgvector, site widget, Google Sheets
Asks three or four questions, filters out the ones who do not fit, tags the lead and moves the deal along the CRM pipeline.
2-3 weeks · GPT-4o, function calling, HubSpot or Pipedrive API, Telegram Bot API, n8n
Closes repeat tickets and hands the hard ones to a person with a summary of the conversation and the customer history attached.
1-2 weeks · Claude Haiku, pgvector, site widget, Slack, email
Checks free slots, books, moves and cancels appointments, and sends a reminder a day and an hour before.
2-3 weeks · GPT-4o-mini, Google Calendar API, Calendly, WhatsApp Business API, n8n
Answers product questions, returns order status by number and walks a return through to a ticket in your system.
2-3 weeks · GPT-4o-mini, BaseLinker, WooCommerce or Shopify API, InPost, Instagram DM
Calls back after a missed call, finds out what the caller needs, books a slot and drops the transcript into the CRM.
3-4 weeks · Vapi or Twilio, Deepgram, ElevenLabs, GPT-4o, Google Calendar
Pulls orders, invoices and delivery notes out of the mailbox, reads line items and totals and files them into the accounting system as a draft for a person to approve.
4-6 weeks · Azure Document Intelligence, GPT-4o, Comarch Optima or Subiekt nexo, KSeF API, n8n
Finds the answer inside company procedures, manuals and contracts instead of sending an employee to ask a colleague.
2-3 weeks · Claude Sonnet, pgvector, Google Drive or SharePoint, Slack
Reads CVs, pulls out what the vacancy actually asks about, comes back to the candidate for anything missing and offers an interview slot. It scores nobody and rejects nobody, the recruiter decides.
3-4 weeks · GPT-4o, pgvector, Google Sheets, Calendly, email
The price covers one channel and two or three integrations: every further integration is quoted separately, and model and telephony usage is billed straight to you. The clock starts the day I have the access and the content in hand, and I run no more than two projects at a time.
A closed list: systems I have already connected, or ones with a documented API that your own licence opens up. If yours is missing that is not a refusal, but I check the API before quoting.
When a system has no API, what is left is screen automation: it is fragile and breaks with every update, so I quote it separately and with caveats, and sometimes I say plainly that the project stops there.
A chat that answers from your knowledge base - without hallucinations. Citations, sources, fallback to a human.
Lead qualification with function calling, scenario classification, automatic pipeline movement in your CRM.
Daily lead pulls, email classification, on-brand content generation, report aggregation.
AmoCRM, HubSpot, Pipedrive, Telegram, WhatsApp, Instagram DM, Slack, Google Workspace.
Not a black box. Dashboard with dialogs, tags, response rating, exports and manual overrides.
GPT-5 / Claude / Gemini - we pick per task and budget. Swap without rewriting the project.
Zero risk
Not verbal promises - actual clauses in the contract we sign before work starts.
The price is set before we start and doesn't change along the way. More scenarios than planned - we agree on a new quote first, then I build.
The launch date is in the contract. If a delay is on me, you get a discount for every week it slips.
Chat history stays with you, personal data is handled under GDPR, models run through European endpoints. No "the data went somewhere, who knows where".
Prompts, knowledge base, integrations and access are all registered to you. Switch developers - the agent keeps working without me.
The main risk of hiring one person instead of an agency. Here is the answer before you ask.
From day one I work in your GitHub or GitLab and in your hosting accounts. Nothing is parked on my machine as leverage for the final invoice: you see every commit the day it lands. Prompts, the database schema and the n8n configs sit in the repository next to the code.
You open the OpenAI, Anthropic and Twilio accounts under your own company and the invoices go straight to you. I put no markup on tokens, and usage shows up in your billing rather than in my report. You grant my access and you revoke it, without asking me first.
At handover you get a runbook: how to deploy, where the logs are, what to do when it falls over, which environment variables it needs. Plus a 30 to 40 minute screen recording where I go through the whole system. It is written for a developer who has never seen the project.
Response times are in the contract: 24 hours on working days, and 4 hours for a production incident during working hours, 9 to 18 Warsaw time. I do not run a 24/7 rota and will not promise one person can. If I go silent for more than five working days you terminate without penalty and without needing anything from me: there is nothing to hand over, the code, accounts and documentation have been yours since day one.
None of this costs extra or needs an annex, it is what my standard contract already says.
The process
A typical plan for a chat agent. The multi-tool one goes through the same steps, just with more integrations. Your part takes a couple of hours.
30 minutes on where you lose clients: what people ask, what should happen with a lead, where the agent lives - website, Telegram, WhatsApp. Then a quote and a contract.
I build the base from your price lists, pages and FAQ and tune the tone to your brand. The agent learns to answer like your best manager, not like a bank robot.
CRM, calendar, handover to a human. Then I grill the agent with tricky questions: provocations, typos, off-topic stuff - and tighten up the answers.
We switch it on for real clients. For the first week I watch live conversations and fix the weak spots. You get an admin panel with metrics: chats, leads, conversion.
Pricing
EU agencies: €2,000-8,000
A narrow bot for one job: FAQ, order status or qualification - one channel.
EU agencies: €8,000-20,000
RAG on your data + CRM integration and lead qualification. The most common pick.
EU agencies: €20,000-60,000
An agent that acts, not just answers: multiple tools, integrations and automations.
EU agency prices shown for comparison - the same work at €100-200/hour. On top: API only, €30-80/month at 1000+ dialogs. Final number fixed after a 30-minute call. No mandatory retainer.
Describe the task in two sentences. You get a fixed price, a timeline and a plan. If it isn't a fit for us, we say so straight away and point you elsewhere.
Comparison
The job is the same - answer clients fast. Here is what each option costs over a year.
| Agent by buildbyalex | SaaS builder | One more manager | |
|---|---|---|---|
| Upfront | €890-2 700 one-time | €0-100 | recruiting + onboarding |
| Monthly cost | €10-50 for API | €50-500 subscription | from 5 000 zł + social contributions (ZUS) |
| Works 24/7, no days off | yes | 8 hours, vacations, sick leave | |
| Knows your services and prices | canned replies | yes | |
| Writes to CRM and books clients | rarely and poorly | yes | |
| Languages: PL / EN / UA / RU | limited by plan | whatever they happen to speak | |
| Replies in seconds | yes | when not busy | |
| Code and prompts are yours | no, you rent a platform | the knowledge leaves with them |
The agent doesn't replace your team - it takes over the repetitive 80% so people can work on deals instead of copy-paste.
It depends on the number of integrations and how complex the processes are. A simple FAQ agent is a different league than one wiring together CRM, calendar and payments. I don't pull a number out of thin air - we run a short scoping based on your processes and you get a fixed figure before we start.
Real. Function calling, tools, RAG, dialog classification, backend logic, integrations - that's what separates a working agent from a wrapper.
Yes. I work in line with GDPR, sign a data processing agreement, and can host the knowledge base and conversations on EU servers. Your data is never used to train public models and stays with you.
In a well-built RAG - almost never. I enforce a strict-citation mode: if the answer isn't in the corpus, the agent says 'I don't know' and routes to a human. It's controllable.
No. The agent takes over the repetitive, time-eating tasks - FAQ, lead qualification, order status - and escalates the harder cases to a human with a ready summary. Your team gets time back for the work that actually needs people.
Usually 2-6 months, depending on query volume. We calculate the break-even point before we start - from the number of conversations the agent will take over and the team hours it frees up. If the numbers don't add up, I'll say so straight.
Through the native API. AmoCRM v4, HubSpot, Pipedrive - they all support webhooks and REST. The bot tags, moves through the pipeline, leaves notes. Everything a manager does manually - automatic.
A focused bot - a FAQ assistant or lead qualifier - ships in 1-2 weeks; an agent with RAG and a CRM integration takes 3-4 weeks. Running costs are mostly API: a typical agent on GPT-4o-mini / Gemini Flash sits around €30-80/month at 1000+ dialogs. The date and cost are in writing before we start.
Yes. A single agent handles Polish, Russian, Ukrainian and English, replying in whatever language the customer writes in. The RAG layer is built on your documents and knowledge base, so answers come from your sources - not the model's general knowledge. Your data stays yours and is never used to train anyone else's model.
We work across all of Poland and remotely - Warsaw, Kraków, Wrocław, Gdańsk, Poznań or a smaller town, it makes no difference. Calls, demos and handover happen online; the contract and payment are remote. An in-person meeting in Warsaw is available on request.
A regular bot walks people through buttons and breaks on the first unusual question. An AI agent understands a question phrased in the user's own words, looks the answer up in your documents and can take action: create a deal, book an appointment, issue an invoice. If what you need is a bot that answers from your knowledge base on the site and in messengers, that is a separate page - AI chatbots for business.
That's exactly how I work. First I look at where time leaks: manual re-typing, repetitive replies, sorting email. The build goes into your current tools, with no CRM migration and no retraining the team.
Yes, and the point of doing it with an agent is that the actions follow the customer between channels. Someone asks on WhatsApp, comes back on Telegram, finishes on the site, and there is still one deal in your CRM with one history rather than three cards nobody merges. The agent qualifies, books and updates the record the same way in each channel. The channel mechanics themselves - the WhatsApp Business API, Meta verification, approved templates - are covered on the AI chatbots page.
Yes. A voice agent answers the call, replies out loud, books the caller and creates a CRM record. It's a separate scenario with its own pricing - we size it on the call against your actual call volume.
Dig deeper
What clients say
A few words from founders and teams we've shipped for.
Launched our store in 11 days - multilingual, fast, and already on page one for three of our main keywords.
The Telegram bot handles the first conversation with every lead now. My sales rep only steps in when the deal is hot.
Leads land in our CRM with every field already filled. No more copy-pasting from chats - the pipeline moves itself.
Shipped to the App Store and Google Play in three weeks. The handover docs let my team take it from there without me.
Set up our Google and Meta ads from zero. Cost per lead dropped by about a third in the first month.
He rebuilt the site and six weeks later ChatGPT was recommending us by name. Didn't know that was even possible.
Free 30-minute call. No sales pitch - just a real conversation about what you need and whether we're the right studio to build it.
No sales pitch - just a clear read on your scope.
