Outlook AI: not just smart reading, but acting in your ERP
Summarizing emails with AI? Everyone can do that now. Co-pilot does it, ChatGPT does it, and new tools appear every month. Summarizing, querying, analyzing — it has become a commodity. The question is: what happens next?
Because that's where the real work is. After reading, someone still has to transfer the action items into the ERP system.
Create a document in Synergy. Register a service request in AFAS. Update a status in Business Central.
That part — the acting — is where every generic AI tool stops. And that's exactly where Outlook AI begins.
The bridge that an LLM doesn't have
A language model can summarize a visit report just fine. But it cannot log into your Exact Synergy environment and create a document with the right metadata, type and status. That requires a structured connection to your ERP — and that connection doesn't exist in ChatGPT or Copilot.
At Any2Info, we have been building dataflows for years: structured connectionsthat read from and write to ERP systems like Exact Synergy, AFAS and BusinessCentral. Hundreds of these dataflows run daily at our customers. They form thereliable bridge between business processes and business systems.
With the introduction of tool calling, we now make those same data flows available to AI agents. An agent can invoke a data flow via a tool call and actually perform an action in your ERP. Simply with a (voice) command such as: “Create a new document in Synergy based on the summary and action points.”
That is the difference between an assistant that talks and an assistant that acts.
How does it work in practice
Take a concrete example. A sales manager receives weekly visit reports fromfive field representatives — as Word or PDF attachments by email.
With Outlook AI, the process works like this:
- Select the emails — the manager selects all five emails with visitreports in Outlook.
- Choose the agent — in the Outlook AI panel, he picks "Exact SynergyDocs AI."
- Chat with the agent — the AI reads all attachments and presents anoverview: findings, action items and attention points per representative.
- Give the instruction — "Create a new Synergy document with thesummary and action items." The agent invokes a dataflow through toolcalling that creates the document in Exact Synergy — with the right content,metadata and status.
Up to this point, steps 1 through 3 are what any AI tool can do. Step 4 iswhat's only possible when you've already built the bridge to your ERP. And wehave that bridge.
The result: an approved document in Synergy, ready for follow-up. Five minutesinstead of ninety. No copy-paste, no manual entry, no switching betweensystems.
One approach, dozens of applications
Because each agent has its own prompts, parameters and tool calls, the sameapproach applies to virtually any situation where email triggers an ERPaction:
- Service requests — a customer emails about a malfunction, the agentcreates a service request in your ERP.
- Quote requests — the agent extracts specifications from an RFQ andprepares a draft quote.
- Expense claims — employees send receipts by email (including images, PDF), the agent categorizesand registers them.
- Onboarding — HR receives documents from new employees, the agent checksfor completeness and logs everything.
Every case combines the same building blocks: an AI agent that understands, anddataflows that execute. The intelligence sits in the model. The action sits inthe dataflow. Outlook AI brings them together.
Not another AI tool, but an executing assistant
Let's be honest: the market doesn't need yet another tool that summarizesemails. What organizations need is an AI that doesn't stop at understanding,but follows through into the business system. With the right permissions,within the right workflows and with a complete audit trail.
That's what Outlook AI delivers. Not because we have the smartest languagemodel — we don't, we leave that to OpenAI and Anthropic. But because we alreadyhave the connection to your ERP. And that connection turns a smart model intoan executing assistant.
