Task complexity
The number of variants, exceptions, and decisions the solution must support.
Home › Solutions › AI
AI for business
AI for business—agents, processes, and data connected to company systems
We implement Copilot and an AI agent in AMODIT, build agents that support request-for-proposal (RFP) assessment, proposal preparation, and contract review and revision, and develop DataFlow, our product for AI-assisted data preparation. We also build custom solutions with AI agents and integrate them with any system, while applying AI in BI and ERP.
Solutions
AI solutions we have already built
We have completed AI projects for sales, contract work, and IT infrastructure maintenance. We build the agents ourselves on existing language models and connect them to company systems. For each solution, we define the required data, expected output, and person responsible for the next action.
AI in processes
AMODIT Copilot
It guides users through the system, suggests what to address first, and prepares a report on request.
How it works in a process →AI in processes
AI agent in AMODIT
It handles work that cannot be expressed through rigid rules: reading a document, comparing information, and assessing the significance of changes. The employee responsible for the next stage receives the output.
See details →Sales and purchasing
RFP assessment agent
It organizes the RFP materials and prepares an assessment for the team to review.
Explore the RFP assessment agent →Sales
Proposal preparation agent
It prepares proposal material before a sales representative takes over. The sales representative approves the price, scope, and commitments.
Explore the proposal preparation agent →Legal
AI agent for contract review and revision
It reviews a contract and inserts proposed revisions for the lawyer to decide on. It does not approve clauses independently.
Explore the contract review agent →Data
DataFlow: AI-assisted data preparation
Our product combines data from multiple systems into one reporting source. The AI assistant's proposal is validated and reviewed before publication.
Check whether your data is ready for AI →IT infrastructure
Cloud Command Center: agents for infrastructure monitoring and management
They observe system health, detect irregularities, explain their causes, and prepare recommendations for the administrator.
Explore Cloud Command Center →Integrations
AI agents integrated with any system
We build end-to-end solutions with AI agents and connect them to systems the company already uses: AMODIT, ERP, CRM, a data warehouse, and other applications.
How we connect AI to systems →ERP
AI in ERP
We connect the AI functions available in Comarch systems with integrations, process automation, and data preparation beyond standard ERP capabilities for companies operating in Poland.
AI in ERP →Have another task for AI?
We apply the same expertise to build agents tailored to your company's processes.
Describe a task for AI →Sound familiar?
AI prepares the analysis; a specialist reviews and approves it
Each situation shows what consumes time today and what the same work looks like after the change.
Before
An account manager reads a dozen attachments before identifying the requirements for a response.
After
An AI agent compiles the requirements from the attachments, and the account manager reviews them and decides how to respond.
AI agent for RFP assessment
Before
A sales representative searches for similar proposals even though the first draft follows a repeatable structure.
After
An AI agent prepares the first draft from earlier proposals, and the sales representative reviews, refines, and approves it.
AI agent for proposal preparation
Before
A lawyer analyzes the entire contract and prepares revisions manually.
After
An AI agent reviews the contract and inserts proposed revisions, which the lawyer reviews and approves.
AI agent for contract review and revision
Before
An administrator analyzes a series of alerts before finding the event that requires action.
After
AI agents organize the alerts and flag events for review, while the administrator decides how to respond.
AI agents for IT infrastructure
Before
An analyst spends hours tracing the source of an incorrect value in a report.
After
The data path is described, so the analyst traces the value from source to report and sees where the error occurred.
DataFlow: transparent data flows
Before
A process owner manually checks whether the description and tests match the configuration.
After
AI prepares the process description and test cases from the configuration, and the process owner reviews and approves them.
AI-assisted process documentation and testing
Where we start
We do not start with a model. We start with work that slows the team down.
People in many departments still read documents, compare options, reconstruct context, and prepare decision material manually. AI handles the repeatable part of the work and prepares the material for the next decision.
1. InputDocuments, data, and process context: an RFP, proposal, contract, invoice, or system record.
2. AI analysisAI reads, compares, searches company knowledge, and prepares the output.
3. RecommendationA proposed assessment, response, revision, or next step, with supporting reasons.
4. ApprovalThe output goes to the employee designated in the process; nothing moves forward without that person's approval.
5. Action in the systemThe approved output enters the process in AMODIT, ERP, or another application.
AI in processes
AI works where the team already works
In processes run on the AMODIT platform, AI is part of everyday work rather than a separate tool. We select the right functions for the process, configure them, and connect them to company systems.
- Copilot supports navigation, priorities, and reporting.
- AI agent performs tasks that require interpreting content within the process.
- AskAI and agent rules provide additional AI functions in AMODIT processes.
- Knowledge repository stores policies, instructions, and guidelines used by Copilot.
- Model Context Protocol (MCP) makes AMODIT data and workflows available in a selected AI tool according to assigned permissions and configuration.
“AMODIT takes over more work in the process—from reading a document to preparing the next step.”
We automate repeatable work and route the output to the designated role in the process.
Process Automation with AI →Sales, purchasing, legal
Agents for RFPs, proposals, and contracts
These three areas share one trait: extensive reading, analysis, and document preparation based on a repeatable structure. The agent performs this work and gives the team a ready assessment, a proposal draft, or a contract opinion with proposed revisions.
Collect materialsRFP, specification, proposal template, or draft contract.
Analyze materialsThe agent reviews the documents, requirements, and areas that need attention against defined criteria.
Prepare the proposed outputAn assessment, proposal draft, or opinion and revisions inserted into the contract.
Approve the outputA sales representative, purchasing specialist, or lawyer revises and approves it.
Take the next stepSave it in a system, send it, or start the next round of negotiations.
We embed AI in specific purchasing and supplier-selection processes, contract and legal work, and sales and proposal processes.
AI in data
AI needs organized data
Incomplete or inconsistent data leads to incorrect conclusions regardless of model quality. Preparing data, definitions, and access is therefore part of every AI project. We have done this in BI projects for years.
AI in data analytics →DataFlow →AI Business Indicators →MCP and integrations
We connect AI to systems the company already uses
We integrate AI agents with any system. The agent uses data and functions exposed by the systems the organization already works with: AMODIT, ERP, CRM, and a data warehouse. Each agent receives only the permissions required for its tasks.
AI ↔ AMODIT ↔ ERP · CRM · data warehouse · applications
How we implement AI
First the task and accountability. Then the model.
A pilot must lead to an informed decision: scale the solution, change the approach, or stop the work. We compare the outcome with criteria defined before the pilot begins.
01
Problem and decision
We define the process, user, decision, cost of the current problem, and expected result. We do not start by choosing a model.
02
Business case
We assess task frequency, potential value, the cost of error, and whether AI offers an advantage over conventional automation or BI.
03
Data and security
We check data availability and quality, permissions, personal data, ownership, retention, governance, and regulatory requirements.
04
Architecture and cost
We select the model, integrations, knowledge sources, semantic layer, monitoring, and cost-control approach.
05
Pilot
We build a limited scope and measure response quality, process time, usage cost, adoption, and errors.
06
Decision after the pilot
We compare the outcome with the defined criteria and decide whether to scale, change the architecture, or stop the pilot.
07
Post-launch controls
We set permissions, monitoring, error handling, activity history, and the approval path.
08
Maintenance and development
We monitor response quality, costs, incidents, and source changes, and maintain a backlog of further improvements.
Security
AI under human control
We design every solution with visible controls. When AI uses personal, confidential, or sensitive data, we define available data and functions, approval roles, and regulatory requirements before launch.
Limited scope
The agent works only with the data and functions it is authorized to access.
Visible output
The recommendation includes its reasoning and references to the data on which it is based.
Approval point
A decision that carries accountability always belongs to a person.
Regulatory compliance
We run models on Azure, Google Cloud (Vertex AI), and AWS in European Union regions. We retain full control over where data is processed, and the models are not trained on client data. This ensures compliance with the GDPR and the EU AI Act.
Why trust us
Evidence instead of promises
AI in a product used by large organizations
AI functions operate on the AMODIT platform, which we have developed for years and implemented in Enterprise organizations.
Completed AI projects
Agents for contracts, RFPs, proposals, and IT infrastructure monitoring and management: built and operational.
AI in our daily client work
We use AI to generate process documentation and test cases from the live configuration. Our team reviews them.
Experience in AI and technology
Przemysław Sołdacki, Vice President and one of Astrafox's owners, holds a PhD in artificial intelligence and has worked with AI since 2009. AI operates on processes and data that we have organized, integrated, and maintained for clients for years.
AI project pricing
What determines the scope and cost
We define the scope for a specific task, data set, and assigned roles—not for the model name alone.
Data and knowledge sources
Their number, quality, availability, format, and the need to organize them before use.
Integrations
The systems, APIs, permissions, and data-exchange directions required in the process.
Security and oversight
Access, audit, and compliance requirements, as well as roles and approval stages.
Testing and maintenance
The number of control scenarios, quality monitoring, and solution maintenance and development.
The first step toward a quote
Describe one task, the input material, and the person who reviews the output today. Based on this, we will determine what must be verified before preparing the scope.
FAQ
Frequently asked questions about AI
What does Astrafox implement in AI?
AI functions in AMODIT (Copilot, AI agent, AskAI, agent rules, and MCP), custom agents for RFP assessment, proposal preparation, and contract work, AI solutions in BI (DataFlow for data preparation), and AI in ERP. We integrate agents with any client system.
Does an AI agent make decisions on its own?
No. The agent works within a defined task and prepares an output for the role responsible for the next process step.
How should a large company start an AI implementation?
Start with one task: define the required data, expected output, and decision owner. Then check the data, exceptions, integrations, and cost of error. Only then select the product, model, or agent.
Can AI be connected to AMODIT, ERP, and other systems?
Yes. We integrate AI agents with any system, and integrations are part of our offer. In each project, we configure the interfaces, permissions, and connections for the specific system.
What AI experience does Astrafox have?
One of Astrafox's owners holds a PhD in artificial intelligence and has worked with AI since 2009. We combine this expertise with capabilities in processes, data, software, integrations, and cloud.
How should data be prepared for an AI implementation?
Preparing data, definitions, and access is part of the project. The quality of the output depends on it.
How much does an AI agent implementation cost, and how long does it take?
We estimate cost and time based on the task, number of sources, data quality, integrations, and security requirements. We start by describing one scenario and its success criteria. We then provide the scope and cost.
Describe one task for AI
Show us what the team must read, compare, or prepare manually today. We will advise whether an AMODIT function, a ready-made agent, or a solution built for your process is the right fit.
Describe a task for AI