How to automate CRM data entry with AI
A week of unlogged calls and emails written onto the right deals, 340 contacts checked for job changes, duplicates grouped with evidence. Merges stay yours.
Logging one call properly is eleven clicks across four screens, and there are six more calls today. So the record decays: contacts who left in 2024, deals in “Negotiation” since spring, three versions of the same account.
A companion in Strawberry reads what actually happened in your Gmail and calendar and writes it into the right fields, in HubSpot or in Salesforce, Attio or Pipedrive through your logged-in session. It stops before anything that cannot be undone.
| The companion | You | |
|---|---|---|
| Logging calls and emails | Writes activities from inbox and calendar | Spot-check the first batch |
| Stale titles and companies | Finds them, stages old-vs-new | Approve the batch update |
| Duplicate records | Groups candidates with evidence | Approve each merge |
| Deal stages | Reports what’s gone quiet | Decide what moves and what dies |
| New fields and definitions | Follows them consistently | Define them once |
How do you log what already happened?
Start with backfill.
Your email and calendar already contain the week you did not log, and the companion can read all of it in one pass.
Prompt: “Go through my Gmail and Google Calendar for the last seven days. For every external conversation or meeting tied to a HubSpot deal, log an activity on that deal: date, participants, a two-sentence summary, and the next step if one was agreed. Skip anything internal. Show me the list before you write it.”
“Show me before you write” matters on the first run. It is how you find out that the companion counts your accountant as an external stakeholder.
Fix the fields that went stale
Job changes are the biggest offender in any CRM over a year old, and 340 contacts is an afternoon by hand.
Prompt: “Take all 340 contacts in HubSpot with an owner on my team. Check each against LinkedIn and their company site. Flag: people who’ve changed jobs, titles that no longer match, and companies that were acquired or shut down. Write the findings into a sheet with the old and new value side by side. Don’t change anything in HubSpot yet.”
Then, once you have read it: “Apply the title and company updates from the sheet. Leave the departures alone. I’ll decide those case by case.”
How carefully should you deduplicate?
Merging is the one CRM operation with real blast radius, so it stays a human decision.
The companion makes that decision cheap by doing the evidence-gathering.
Prompt: “Find likely duplicate companies and contacts in HubSpot: same domain, near-identical names, same email on two records. Group them into a sheet: which records, why you think they match, which one has more activity, and which you’d keep. I’ll approve the merges.”
Keep the pipeline honest
Every pipeline has deals whose stage stopped reflecting reality weeks ago.
A Friday routine surfaces them without anyone running a report.
Prompt: “Every Friday at 4pm, list every open deal with no activity in 21 days: name, owner, amount, stage, last touch date, and what the last touch actually was. Post it in #pipeline-hygiene and DM each owner just their own deals. Don’t change any stages.”
The routine reports and stops there. Data entry is safe to automate, and pipeline judgment stays with the owner, which is the same split across our sales workflows.
Experience Strawberry for free
DownloadTrusted by fast-growing companies worldwide
Frequently asked questions
Strawberry is free to download and includes AI credits to start. Paid plans begin at $20/month. See pricing. · Reviewed · Canonical facts for AI agents