No benefit found, no mapping fee
If my recommendation does not identify a concrete place where AI saves time or costs, or improves the work, you do not pay for the map.
AI agents for individuals and businesses
Choose AI that genuinely helps. And learn how to manage its agents.
I assess how you work, select suitable tools and design AI agents that you can actually use and control. From a personal agent for professional work to a managed team with memory, roles and output verification. I teach practical ways to reduce errors through human gates and multi-stage verification. When a larger process automation is worthwhile, I involve specialist Pavel Šolc.
- AI tool selection and real value
- An AI agent for an individual or a business
- Agent management, orchestration and verification
- Larger business automation
No generic lecture and no pressure to buy another tool.
A guarantee of concrete value
You do not pay for enthusiasm about AI. You pay for a decision you can use.
I will not sell you a tool you do not need
Sometimes the best answer is to use a tool you already pay for more effectively. Sometimes AI is not suitable. I will say both directly.
You see a sample deliverable first
Before you pay, I send you an anonymised sample page from a real recommendation so you know what you will receive.
Proven practice
What I have built with AI and use in practice
Concrete websites, applications, professional deliverables and agent workflows for clients and my own organisations.
My own AI agent team
I run a multi-agent working environment every day. The agents have different roles, memory, communication channels and verification gates. I study their behaviour, failure modes, continuity and how people should manage such systems.
- work divided between different models and specialisations
- memory, context handovers and continuity checks
- multi-stage verification of professional outputs
SME Union — website and AI radars
The Czech SME Union website, a Funding Radar drawing on multiple sources, a Legislation Radar, a legislative calendar and a publishing workflow. The systems are created and operated with a managed team of AI agents.
View SME UnionCustom work for clients
I use agent workflows even when the result is not a chatbot, but a professional document or a complete application:
- grant applications, business plans and legal supporting materials
- an application for fire-safety documentation including drawings
- applications for grant processing and property valuation
- e-commerce automation: product discovery, automatic listing and sales
- design and testing of agent workflows for professional and operational work
AI Act, without the panic
A free practical EU AI Act self-check for small businesses. It produces a prioritised plan and editable templates without registration.
Open the self-checkKontinuita — Agent Continuity Research
An experimental infrastructure for long-term research into the memory, identity, behaviour and relationships of AI agents.
View KontinuitaMy YouTube channel — Radka Harantova
Videos created largely with AI — from research and scripts to Czech and English versions. A practical example of a managed agent workflow.
Open the channelWhat I do for you
From the right tool to a managed team of agents
I start with the simplest solution that produces a result. I build a more complex agent architecture only where there is a real reason.
AI tool selection and real value
I assess how you work, what should stay with people and where a standard AI tool or an agent can help. You get a concrete choice of tools, a first workflow and an estimate of the benefit — not a list of fashionable apps.
An AI agent for an individual or a business
I design and configure an agent for specific professional work: research, decision support, documents, grants, business assignments or internal knowledge. You also learn how to manage and check the agent.
Agent management, orchestration and verification
I design roles, memory, handovers, guardrails and control gates. You learn when to rely on an agent, when to stop it and how to have important outputs independently verified.
Larger business automation
If the analysis shows that several systems must be connected and a complete process automated, I deliver the implementation with Pavel Šolc, a developer and specialist in large-scale business automation.
Practical agent-management tips
More reliable outputs without unnecessary supervision
The goal is not to prohibit everything. Agents work best when they have enough freedom inside clear guardrails and their outputs are checked using a few simple rules.
Give the agent guardrails, not a straitjacket
It can work independently, but define what it may access, what it may send and where it must wait for approval.
Require a source for every number and fact
For amounts, dates, names, legal provisions and other facts, require a link to a specific source. Do not use unsupported data.
Run web research as of today
Put the current date in the task, require up-to-date searching and ask the agent to state both the source date and the time of verification.
Broaden the sources and filters
Do not let the agent stop at the first results. Specify several source types, languages, time periods, regions and conditions before comparing findings.
Separate creation from checking
One agent creates the output, another verifies facts against the sources, and a third checks the brief, completeness and contradictions.
Use people only at important gates
An agent can work freely inside its guardrails. A person approves sending, payment, signing, publishing or another irreversible step.
Practical rule: an agent must not merely write “verified”. It should show the specific source, its date and how the check was performed.
What can be automated
The list can never be complete. These examples help you imagine what may be possible in your own work:
- extracting data from invoices and receipts
- matching payments with invoices
- issuing invoices and reminders
- processing enquiries and preparing offers
- orders and inventory monitoring
- shift planning
- service jobs
- meeting notes and task assignment
- generating documents from templates
- sorting and answering emails
- booking and support chatbots
- publishing to websites and social media
- newsletters
- marketing and campaigns
- data analysis and reports
- document classification
- CV screening and employee onboarding
- e-commerce: product discovery, listing and sales
…and many more in operations, sales, administration and marketing. I have also delivered an e-commerce workflow from automatic product discovery through listing to sale. I prepare larger implementations with Pavel Šolc, a specialist in business process automation. I lead the agent design, system behaviour and quality control; Pavel handles integrations and technical operations.
How I work
Three steps to a prototype you can test yourself
These are not three separate invoices. They are the working process from understanding your task to the first functional version.
- 1
Mapping — 90 minutes
We go through the real work, decisions, tools, sensitive data and the places where time or attention is being lost.
- 2
Design and control rules
I determine where a standard tool is enough, where one agent helps and where several agents are justified. I design roles, permissions, human gates, verification, costs and expected savings.
- 3
A prototype for you to test
You receive the first working version on your real task. You test the output, controls and limits before the solution is tuned for everyday use.
What happens after you test the prototype
If the prototype makes sense, I tune the agent based on your experience, set up multi-stage verification and hand over the finished solution. The custom-agent package also includes training in how to use, check and safely manage the agent.
If you only want a recommendation without building an agent, you can order the standalone map. It does not commit you to anything else.
About me
Radomíra Harantová
I specialise in AI agents: how to design them, divide work between them, hand over context, stay in control of the result and recognise when an output needs verification. I do not merely build a multi-agent environment — I work in one every day and study its behaviour.
I previously ran my own IT company, worked in public administration and now lead the Czech Association of Small and Medium-Sized Enterprises. I have worked with business owners for more than twenty years and understand their decisions from the inside. That is why I start with one question: where can AI genuinely return time, attention or money?
- I run my own team of agents with roles, memory, working channels and verification gates
- I research continuity, failure modes, relationships and the management of agent systems
- I build professional workflows from research to verified output — not only simple chatbots
- I have run an IT company and lead a business association, so I know business from the inside
- I have experience in public administration, grants and European funds
Pricing
Standalone packages, not three paid steps
Choose the deliverable you need. The three steps above only show how a prototype is created — they are not three separate line items.
The most common route: mapping → prototype → tuned agent and training.
A standalone map costs CZK 7,900. The custom-agent package includes the whole process, including mapping. If you have already bought the map from me, I deduct its price from the package.
Individual AI agent consultation
CZK 2,900
90 minutes
For individuals, business owners and specialists. We use your real task to choose suitable AI tools and set up the first practical agent workflow.
AI tools and agents map
CZK 7,900
standalone deliverable
An interview, analysis of your work and a written plan: which tools and agents make sense, who should manage them and where the real savings are. If we continue with a custom agent, I deduct the mapping fee.
Custom AI agent
from CZK 16,900
from mapping to handover
Includes mapping, design, a prototype to test, one tuning round, control mechanisms and training. You know the exact scope and price in advance.
Team workshop
from CZK 9,900
3 hours, up to 8 people
Hands-on work by role: selecting tools, assigning tasks to agents, checking outputs and using AI safely. I tailor the content to your situation.
AI Act readiness
CZK 14,900
starter package
Overview of the AI systems you use, an initial risk classification, internal rules and basic team training. This is not legal advice.
Orchestration and larger automations
On request
based on scope
Multiple agents, memory, task handovers, verification gates and integrations with company systems. I deliver larger process automations with Pavel Šolc.
I am not VAT-registered, so the listed prices are final. The introductory 30 minutes are free. For custom agents and larger implementations, you receive a clear scope, price and method of checking the result in advance.
Tell me what you want to achieve with AI
During a free 30-minute call, we will find out whether you need better tool selection, one agent or a complete agent team. If I am not the right path for you, I will say so.
Privacy and personal data
I process the personal data you provide — such as your name, email address or telephone number — solely to handle your enquiry and communicate with you. When you send an email or book a call, the data may be technically processed by the email or booking service provider. I do not use it for marketing without your consent and keep it only for as long as necessary. You may request access, correction or deletion using the contact details above. This website does not use tracking cookies. The data controller is Radomíra Harantová, company ID 70562849.