The Agentic Revenue CompanyRevenue as a Service
Blog

ERP integration for AI agents: the data sales really needs

Which ERP data AI agents need for repeat business, and how synchronisation, legacy data migration and completeness checks make sure agents act correctly.

An agent suggests approaching a regular customer about a reorder. But the customer placed an order a few days ago. Another customer receives a friendly e-mail even though a block is recorded for them in the ERP system. And during a call, your colleague asks for a contact person who left the company long ago. Three mistakes, one common cause: patchy ERP integration.

None of these mistakes is down to artificial intelligence. They are down to missing or outdated data. That is why how well agents are connected to the ERP system determines the quality of your sales work more than any language model.

What data AI agents in sales really need

Repeat business lives on order history: who ordered what, and when will the need come round again? Those answers sit in the ERP system, not in a mailbox or a call note. An agent without this data can only guess – and a guessing agent in customer contact does more harm than none at all.

On top of that, there is a familiar problem in knowledge work: according to an estimate by the McKinsey Global Institute, knowledge workers spend nearly 20% of their working week searching for internal information or for colleagues who can help (McKinsey Global Institute, The social economy, July 2012). The estimate is older and covers knowledge work in general, not just sales. The idea behind it applies to agents just as much: whatever they first have to search for is not at hand when it matters.

A good integration does not fetch everything, but the right things. For agents in sales, that mainly means:

  • customers with their master data,
  • contact persons at those companies,
  • blocks recorded in the ERP system,
  • orders with their line items.

Line items matter more than they seem. Only with them does an agent know what a customer actually ordered last time. That is why the call briefing before a call shows the line items of the last order. And only real orders count towards purchase cycles and demand, not quotes. How each customer’s purchase cycle is derived from them is described in recognising the purchase cycle.

What ERP integration for AI agents has to deliver

Importing data once is not enough. What matters is that it stays current and that bad data does not trigger any action. In the ERP integration of the REVAGO system, several rules take care of this:

  • Synchronisation daily and on request. A new order affects the open cases, so that a customer is ideally not reminded about an order they have just placed.
  • Completeness check on import. The system checks whether the data is complete as soon as it is imported.
  • Newer data beats older data. An outdated retrieval never overwrites a more recent state.
  • Duplicates are cleaned up, so that the same customer is not approached twice.

A change of system can be handled as well: legacy data from a predecessor system can be migrated. Anything that cannot be assigned unambiguously goes onto a clarification list – and is decided by a person, not silently guessed. You can read more about checks before every action in the post on CRM data quality.

Human decisions survive the import

A common mistake with integrations: the next import overwrites what people have deliberately decided. If your team has marked a contact person as not relevant, the next synchronisation must not bring them back into outreach.

That is why human decisions survive the import in the REVAGO system. To make this work, separate status models keep apart what tends to blur in everyday work: the status of a company, a person’s employment at that company and the state of a use case. A contact person leaving does not mean the customer is inactive – and a closed use case does not mean the relationship has ended. That is human-in-the-loop at the data level: the machine updates, the person has the final say.

Your ERP system, your mandate

There is no universal integration with every ERP system, and we do not promise one. We set up the integration per mandate for the ERP system you have agreed with us. Which data your agents need, which blocks apply and how legacy data is to be handled are things we clarify up front.

This is how the building blocks fit together: in AI-first coaching, you find out whether and how AI agents can secure and grow your revenue. During implementation, we build your agents and connect them to the agreed systems. In operation, we keep the integration and synchronisation running on an ongoing basis. How a clear mandate turns into a running process is the subject of Go: from pilot to operation. You will find the use case under repeat business and the capabilities under analyse customer data and prepare briefings.

Your next step

Before interfaces come into it, there is another question: which customer decisions could your sales team make better today if the data from your ERP system were available at the right moment? If you would like to clarify that for your company, check your REVAGO potential.


← All articles