The bottleneck
Every system people type data into day after day gets messy. Slowly but surely, and each in its own way:
- In your CRM: a mobile number in the landline field, the same person twice because they email from two addresses, a contact named after their email address.
- In your ERP: a customer or supplier that exists three times under a slightly different name, missing VAT numbers, addresses in all sorts of formats.
- In your PIM: products without dimensions or weight, "piece", "pc." and "pce" mixed together, descriptions only half translated, items in the wrong category.
Small each time. Until you want to set up a campaign, send a correct invoice, feed your webshop or catalogue, or pull a report. Then that mess really costs time, and errors travel all the way to your customers. Nobody enjoys cleaning it up by hand, and a separate cleanup tool per system means yet another licence each time.
How the platform solves it
- Look first, touch nothing: AI reads your records and shows, per type of problem, what would change, with concrete examples. You decide whether it is right.
- Your rules, in plain language: which field something belongs in, which format or unit you use, when two records are the same customer or the same product. No script someone has to write first, just clear agreements.
- Missing data from your own sources: a number from an email signature or your call history, a VAT number or address from an invoice, product specifications from your supplier's datasheet or price list.
- Duplicates merged: whatever exists two or three times becomes one record. What was unique to each version is kept.
- Doubt goes to a human: conflicting or unreadable data is set aside for a decision. Better empty than wrong.
Then turn off the tap
A one-off cleanup is nice, but if nothing else changes, the same mess simply creeps back in. That is why we record the rules in one place, and every process that writes to your systems uses those same rules: a new lead in your CRM, a new customer or supplier in your ERP, a new product arriving in your PIM. Before anything is written, it is checked. And when a process touches an existing record again, it cleans up the data already there along the way.
That way data quality is no longer a project you redo every year, but something that stays right on its own. If your systems are already connected, as in a 360° view of every customer, those integrations go through the same check, and an error no longer travels from one system to the next.
How we did it ourselves
We first applied this to our own CRM (HubSpot), when we were building an invite list for an AI workshop and ran into numbers in the wrong field and duplicate contacts. The result: almost two thousand contacts updated, more than half again as many filled-in mobile numbers because they finally sit in the right field, and dozens of contacts with a number or real name from an email signature or call. The full story is in Cleaning up your CRM with Claude, and keeping it clean.
The same approach works for customer and supplier records in an ERP or product data in a PIM. Only the rules and the sources differ.
Systems involved
- The system you want to clean up: your CRM (for example HubSpot or Teamleader), your ERP or accounting package (for example Exact Online) or your PIM
- Your own sources to fill in data: mailbox, telephony, invoices, supplier catalogues and datasheets
- An AI model like Claude, securely connected to those systems (ChatGPT, Copilot or Gemini work too)
- Recorded data quality rules that check every write action
What it delivers
- Data you can build on: for campaigns, invoicing, your webshop and reports
- No more duplicates, without losing any data
- Missing data filled in from your own sources, without an external database
- No separate cleanup tool or licence per system
- New entries that follow the same rules, so it stays clean
Related services
- Connect: the integrations between your CRM, ERP, PIM and other systems.
- AI: AI that works securely with your own business data.