Analysis

An AI agent connected to your ERP – what it can do, where it fails and what drives the cost

This article is for owners, COOs, CFOs and IT managers in trading and manufacturing companies considering connecting an AI agent to their ERP, CRM or warehouse management system. An agent that only answers questions and an agent that can change an order’s status are two different projects. We show how they differ, where they deliver results, why it is usually the data rather than the model that fails, and what determines the cost.

Author: Kacper Włodarczyk, CEO of ALGORCOMPPublished: August 23, 2026Reading time: 12 min readAI / AI agentsFor: Mid-sized company
An AI agent connected to your ERP – what it can do, where it fails and what drives the cost

Three levels of agent access to the ERP and what each one changes

A conversation about an AI agent in the ERP becomes concrete once someone asks exactly what it is allowed to do with the data. Read, propose and write are three different projects under one name. The table below compares their risks and the safeguards they need.

With read access, the agent answers in sentences instead of tables: what the status of an order is, how much is left in stock, when the last invoice was issued to a customer. The main risk is showing data to someone who should not see it. This is the cheapest option to start with.

With proposals, the agent prepares a document but does not approve it: a draft order, a draft correction, a preliminary match of payments to invoices. A person checks the prepared entry and approves it with one click. This level usually gives the best balance of benefit and risk.

With write access, the agent changes data in the system itself, e.g. it creates a customer record or changes an order’s status. This requires limits, data checks before saving, a log of operations and the ability to roll back changes. The projects that cause the most trouble usually jumped straight from read to write.

Levels of agent access to the ERP and the safeguards required
LevelWhat the agent doesMain riskSafeguards needed
ReadAnswers questions about operational dataShowing data beyond the user’s permissionsThe same permissions as in the ERP, showing only data available to the person asking, logging the questions asked
ProposePrepares a document for approvalApproving on autopilot, without readingHighlighting fields filled in by the agent, flagging items the agent is unsure of, spot checks
WriteChanges data in the system itselfA series of errors made faster than anyone can reactLimits on the number and value of operations, checks before saving, a log, rollback, tests after every system change

Five processes where an agent in the ERP really works

The processes that work best are those where people switch between systems or retype data from documents in varying formats. The list below shows five such cases.

A simple estimate helps gauge the scale. Suppose 10 people in customer service each spend 30 minutes a day checking statuses in two systems. That is 300 minutes, or 5 hours a day and about 100 hours a month over 20 working days.

  • Statuses and deadlines: “What’s happening with the delivery for customer X?” usually means looking in two or three systems. An agent with read access answers in one place and gives the source of every figure.
  • Taking orders: orders arrive by email, as PDFs or in spreadsheets, each time in a different layout. The agent matches the lines to item codes in the ERP, flags ambiguous ones and creates a draft order for the salesperson to approve.
  • Consistency checks: the agent compares data across systems and reports discrepancies, e.g. a customer with a different tax ID (NIP) in the CRM and the ERP, or a delivery with no warehouse document. It changes nothing; it only shows where to look.
  • Matching payments to invoices: rules handle typical transfers, and the agent helps with the rest. It reads the transfer reference, recognises a shortened company name and a single payment covering several invoices, then proposes a match for approval.
  • Preparing for a customer meeting: the agent gathers purchase history, open orders, overdue receivables and recent complaints into a single note before the meeting.
An AI agent connected to your ERP – what it can do, where it fails and what drives the cost

Where it fails: several definitions of one concept

The most dangerous fault of an agent in the ERP does not look like a fault. The agent does not refuse to answer or invent a customer; it gives a figure calculated correctly, but according to a different definition from the one the person asking had in mind. It also gives no sign that it had to choose at all.

The question “how many active customers do we have?” has three answers if the system allows them to be counted in three ways. It might be a flag in the customer record, the absence of an end date for the relationship, or a purchase in the last 12 months.

The second form of the problem is duplicate item codes. If the same product has three codes in the item master because different people entered it over the years, the agent may create orders for the wrong one. A person makes such a mistake once; an agent repeats it across hundreds of documents.

The third is free-text fields: order remarks, call notes, comments on complaints. Commercial arrangements are often recorded only there, with no expiry date. The agent may draw a conclusion from them that was true two years ago.

Reviewing the data before implementation – four questions

A data review sounds like a separate consulting project. In a mid-sized company, it usually takes a few days of work by someone who knows the data, and the four questions in the list below, asked for every area the agent will touch.

The last question brings the most surprises. In many ERP systems, permissions are assigned to the program’s screens, not to the data. An agent that reads data directly from the database then bypasses them completely, so the permissions have to be recreated in the agent’s settings.

  • Definitions: for every concept the agent uses, e.g. active customer, available stock, overdue receivable, write down one binding definition and the field it is calculated from. If the department cannot point to a single one, the process owner must decide.
  • Completeness: what share of records have the fields the agent relies on filled in. A customer list in which every third customer has no industry assigned is not fit for dividing customers into groups.
  • Duplicates and coding: how many customers appear more than once, how many item codes describe the same product, and whether units of measure are consistent.
  • Permissions: who may see the data the agent will return, and whether this can be reproduced outside the ERP screens.
Diagram of three levels of AI agent access to an ERP system: read, propose and write

If the system has three definitions of an “active customer”, the agent will pick one of them – and do so faster than anyone notices it picked the wrong one.

Safeguards: limits, rollback and division of responsibility

An agent with write access needs safeguards that a traditional integration did not. An integration does exactly what it was programmed to do, whereas an agent makes decisions based on content. That is why the three mechanisms in the list below are a precondition, not an extra.

An example of a limit: the agent may create 20 documents an hour, not 2,000, and will not approve an operation above a set amount on its own. A limit does not mean distrust of the model. Every system makes errors in series, and a limit turns a series into a single problem to fix.

  • Limits on the number and value of operations, set by the process owner and recorded in the agent’s configuration.
  • Log and rollback: every operation is marked in the ERP as performed by the agent, with a record of the basis for the decision. Without this, you cannot trace the cause of an error or roll back a series of documents.
  • A written division of responsibility: one page setting out the agent’s and people’s decisions, amount thresholds and the person who reviews the results every week. Its absence becomes obvious at the first serious error.

What drives the cost of an agent in the ERP

The price depends above all on the number of systems the agent connects to, and whether it is allowed to write to them. The AI model and the number of users usually matter less. The table below lists the main cost components and ways to limit them.

Maintenance matters more here than in an ordinary IT project. An ERP update, a new field in a master record or a change to permissions all require the connection to be adjusted and retested. An agent with no maintenance budget usually stops handling some cases quietly within a year.

In a typical implementation, the cost of the AI model itself is a smaller item than maintaining the connections. It grows quickly with poor decisions, though: overly long instructions, retrieving many documents for every question or using the most expensive model for simple document-type recognition.

Cost components of an AI agent connected to an ERP
ComponentWhat it depends onHow to limit it
Connections to systemsNumber of systems, availability of a data exchange interface, write accessStart with one system and read access only
Reviewing and cleaning dataNumber of inconsistent definitions, duplicates and empty fieldsLimit the first stage to one process and a few concepts
SafeguardsLevel of access, audit requirements, number of thresholdsStay at the proposal level until accuracy is stable
Licences and model usageNumber of questions, length of instructions, choice of model; in Copilot Studio – credits for the whole companyA cheaper model for simple tasks, short instructions, retrieving only the data needed
MaintenanceHow often the ERP and processes changeTests after every ERP update, a responsible person on the company side

An old ERP with no data exchange interface – what then

A data exchange interface (API) is an orderly way for one program to fetch data from another and save data to it, much like a counter in a public office with a set procedure. Modern ERP systems have one; older ones often do not. An agent will still work, but differently and at a higher cost.

That leaves three routes. Reading directly from the database is fast, but the permissions must be recreated outside the ERP. An intermediate layer – a separate program that exposes selected data to the agent – is safer, but it has to be built and maintained. Automating clicks on the program’s screens (RPA) is the most fragile, because any change to a screen’s layout breaks it.

Such connections are increasingly built using the MCP standard, which lets you connect the same system to different AI tools without writing a separate connection for each. It does not solve the data or permissions problem, though – that still has to be designed.

Where to start, and when an agent in the ERP will not pay off

Start with one process at the read level. Choose the process where people most often switch between systems to answer a single question – usually order statuses or product availability. An agent that only reads quickly shows the real state of the data before you decide on write access.

An agent will not pay off if the data is in such a state that people themselves do not trust the ERP reports. Nor does it make sense if the ERP vendor offers a built-in assistant that covers your needs. First check what your licence already includes, and only then consider your own agent for the gaps.

  • Step 1: choose a process and write down the five concepts the agent will use most often.
  • Step 2: review the data using the four questions in this article.
  • Step 3: launch the agent with read access for one group of employees and measure how many questions it handles and how often an answer needs correcting.
  • Step 4: add draft documents for approval. Consider write access only for a narrow group of low-value operations.

Related topics in the knowledge base

Read more about connecting AI to company systems

FAQ

AI agents in the ERP – frequently asked questions

Questions that management boards and IT departments of mid-sized companies usually ask after their first conversation about an agent connected to the ERP.

Isn’t the ERP vendor’s built-in AI assistant enough?
Sometimes it is, and then it is the cheapest route. ERP vendors are increasingly adding AI assistants to their systems. Their limitation is usually that they see only data from one system, while employees’ questions often also concern the CRM, the warehouse or email.
Can the agent work on data we do not want to send to the cloud?
Yes. The model can run on your own infrastructure or in a dedicated cloud environment with control over where data is stored. This raises the cost and needs people to maintain it, so it is worth limiting this option to data that really requires it, e.g. the commercial terms of key customers.
Can the agent replace Power BI reports?
It will not replace them, but it will complement them. A report shows fixed indicators in a proven way, while the agent answers ad hoc questions. Both should use the same definitions, otherwise the board will get two different figures for the same question.
Who in the company should be responsible for such an agent?
The process owner on the business side, e.g. the customer service manager or the COO – not just the IT department. They set the definitions, thresholds and approval rules. IT is responsible for the connections, permissions and security.
Is an agent with write access to the ERP a high-risk system under the AI Act?
Usually not. Managing orders, stock levels or settlements is not listed in Annex III of the AI Act. However, since 2 February 2025 companies have been required to take measures to build their employees’ knowledge of AI, and GDPR applies where personal data is involved.

About this page

About the author

Kacper Włodarczyk — CEO of ALGORCOMP

Kacper Włodarczyk

CEO of ALGORCOMP

Kacper specializes in implementing Microsoft 365 Copilot, Copilot Studio, Power Platform (Power Automate, Power Apps, SharePoint) and AI agents for mid-sized B2B companies in Poland. He leads dozens of projects covering AI strategy, Power Platform governance, document workflow automation and sales processes. In his writing he focuses on the practical side of AI adoption in organizations — from the first proof of concept to company-wide rollout — with particular attention to data security, compliance (GDPR, NIS2, the AI Act) and return on investment.

Meet the team

Check whether your data is ready for an AI agent

We will go through the five most important concepts in your ERP with you and assess whether an agent can work on them safely – and if not, what needs sorting out first.

Featured

Related articles