A message we received recently, lightly paraphrased:

"We're a nonprofit looking for support clearing our Salesforce data of duplicates and making sure everything is up to date — basically a wee declutter. We'd also love to know how to use AI with our Salesforce and Pardot systems."

If that sounds like your organisation, you're in good company. Almost every team that's used a CRM for more than two years has the same problem: duplicate contacts, dead records, fields nobody fills in, and a nagging feeling that the data can't be trusted. And now everyone's asking the second question too — "where does AI fit into this?"

Here's how we'd approach both, in plain language.

Why your CRM got messy (it's not your fault)

CRM data decays naturally. People change jobs, emails bounce, someone imports a spreadsheet without checking for duplicates, a form on your website creates a new contact instead of updating an existing one. Industry estimates put B2B contact data decay at 20–30% per year. You didn't do anything wrong — you just didn't have a system fighting the decay.

The mistake most teams make is treating cleanup as a one-off event. You "declutter", it looks great for three months, and then it's a mess again. A proper fix has two parts: the cleanup, and the plumbing that stops it from happening again.

Part 1: The declutter

Here's the order we work in, and the order we'd suggest even if you do it yourself:

1. Decide what "clean" means before touching anything

Which fields actually matter to your team? Which record types do you genuinely use? Half of most cleanups is deleting fields and processes that nobody has used since 2021. Less structure, maintained well, beats lots of structure ignored.

2. Back everything up

Before any merging or deleting, export a full backup. Deduplication done wrong destroys data — merged records can't easily be un-merged.

3. Handle duplicates with rules, not eyeballs

Salesforce has built-in duplicate and matching rules, and tools like Cloudingo or DemandTools can bulk-merge based on criteria you set (same email, similar name + same organisation, etc.). The important work isn't running the tool — it's deciding the rules: which record wins? What happens to conflicting field values? Who's the surviving record's owner? Get those decisions written down first.

4. Deal with the dead weight

Contacts with hard-bounced emails, no activity in 2+ years, and no donations or open opportunities probably shouldn't be cluttering your reports (or your Pardot email limits — you may be paying for mailable contacts you can't actually mail).

5. Fix it at the source

This is the step everyone skips. If duplicates came from your website forms, fix the forms to update-not-create. If they came from event spreadsheet imports, build an import process with matching rules. Otherwise you'll be doing this again next year.

Part 2: Where AI actually helps with Salesforce and Pardot

Now the fun part. Once your data is trustworthy, AI has genuinely useful applications — and a few overhyped ones. Honest rundown:

Genuinely useful today

Approach with caution

DIY or get help?

Honest answer: if you have under ~5,000 records and someone on the team with a spare week, you can do most of the declutter yourself with Salesforce's native tools and the steps above.

Get help when: you have tens of thousands of records, complex objects (donations, memberships, grants), integrations feeding data in from multiple sources, or you want the "fix it at the source" plumbing and AI layer built properly. That's systems work, and it's what we do.

Either way, if you'd like a second opinion on the state of your CRM — or a realistic take on what AI could do with your setup — get in touch. We're happy to tell you if it's a DIY job.