Switch from Excel or CSV

This page belongs to the area Switch from another system. The process with your AI is described in Switch with your AI. It applies to Excel, Google Sheets and CSV files from any source.

When to use the file import and when the switchCopy link to heading

If you only have leads, use the file import. It is simpler. The switch is worthwhile if your spreadsheets also contain history: calls, notes, tasks, appointments or opportunities with stage. The file import knows only leads.

What you prepareCopy link to heading

One table or sheet per kind of data. Best like this:

The instructions for saving as XLSX or CSV are in Leadmode at the import under "Where does your data come from?" for Excel and Google Sheets. They have been checked.

If you only have one table, that is fine. Your AI shows you what it finds in it and suggests how it is divided.

How columns become Leadmode objectsCopy link to heading

Several contact persons in one rowCopy link to heading

In the file import and in the switch you can store several contact persons and contact channels per company in two columns: peopleJson (persons) and channelsJson (numbers and emails). Both are JSON lists. An example for a company with two contact persons:

peopleJson:   [{"ref":"a","firstName":"Anna","lastName":"Muster","salutation":"Frau","jobTitle":"Geschäftsführung","primary":true},
               {"ref":"b","firstName":"Ben","lastName":"Beispiel"}]
channelsJson: [{"kind":"phone","value":"+49 30 1234567","label":"Zentrale"},
               {"person":"a","kind":"mobile","value":"+49 171 1234567","label":"Direkt"},
               {"person":"a","kind":"email","value":"anna@beispiel.de"},
               {"person":"b","kind":"phone","value":"+49 30 7654321"}]

• Persons: firstName or lastName (one is enough). Optional are ref (a name to which the channels refer), salutation (Herr, Frau, Divers or Keine Angabe; without an entry the salutation stays unknown and is never guessed from the first name), jobTitle, primary (main contact person; without an entry the first person) and doNotContact.
• Channels: kind is phone, mobile or email, plus value. person refers to the ref of a person; without person the channel belongs to the company. Optional are label, preferred (preferred; without an entry the first per person and kind), enabled and doNotContact.
• Errors: an invalid row names the place, for example invalid_structure:channelsJson[1].kind. Values from the file never appear in the message. Preview and import check by the same rules.

Typical pitfallsCopy link to heading

• No identifier. Without a unique identifier per row, Leadmode cannot later recognise whether an entry has already been imported. If you can, add a column with a running number and do not change it later. If it is missing, your AI forms an identifier from the row. This only holds as long as the table does not change.
• Companies as text. If persons only carry the company name and no identifier, your AI connects via the name. Spellings such as "Müller GmbH" and "Mueller GmbH" may be separated by it. Check the messages in the trial run.
• Several values in one cell. Your AI splits several phone numbers or emails in one cell. Check the result.
• Date formats. "01.02.2026" can be 1 February or 2 January. Tell your AI which format applies. Give times as well, otherwise the order in the history is missing.
• Merged cells and empty rows. Remove them. The first row must contain the column headings.
• Several sheets. For the file import exactly one sheet must be filled. In the switch your AI reads the file itself. One sheet per kind or one CSV per kind is simplest.
• Free-text statuses. The more different spellings, the longer the mapping. Tidy up duplicate spellings beforehand, if you can.
• The future. Calls and completed tasks in the future are rejected.

What does not come alongCopy link to heading

• Formulas, charts, colours and comments. The value of the cell counts.
• Files linked in cells, and attachments.
• Emails as a mailbox and call recordings.

AcceptanceCopy link to heading

1. Count the rows of your tables and give the numbers to your AI as "expected".
2. Have the reconciliation report shown to you. Each difference must have a reason.
3. Open three to five companies you know well. Check history, tasks, owner and stage.
4. Check that blocks arrived and that no email went out.
5. Complete the order. If you do not like the result, take it back and start again.

See alsoCopy link to heading

• Your terms, our terms
• Import and Export