Approve Lusha enrichment field by field before HubSpot saves it

Build me a review desk for CRM data hygiene where I approve enrichment changes instead of letting a script overwrite my contacts. This is deliberately human in the loop and it is the opposite of a scheduled enrichment job: the app never writes to HubSpot on its own, never enriches a record I did not select, and never applies a value I did not approve. The field level approve and reject step is the whole point of the app, so do not collapse it into a single save button that pushes everything back at once.

The app opens on the hygiene queue. Pull it with HubSpot Search Contacts using a filter group that matches contacts where email is empty, or phone is empty, or job title is empty, or lastmodifieddate is older than a staleness threshold I set at the top of the page, defaulting to six months. Page through results one hundred at a time using the cursor in paging.next.after, and request the properties email, firstname, lastname, phone, mobilephone, jobtitle, company, website, the org's LinkedIn URL property, lastmodifieddate and hubspot_owner_id. Each queue row shows the contact name, company, which fields are missing as small chips, how long since the record was last touched, and the owner name. Resolve owner names once per session with HubSpot List Owners and cache the mapping rather than looking one up per row. Let me sort and filter the queue by gap type, by owner, and by staleness, and select rows with checkboxes that keep a running count and a credit estimate for the current selection.

Clean first, dedupe second, enrich last. Before the Enrich button becomes available for a selection, show me a short pre-flight panel listing the rows the app will skip and why. Skip any contact with no usable identifier, meaning no email, no full name plus a company domain that can be derived from the website or the email domain, and no LinkedIn URL, because there is nothing to match on and enriching it would burn a credit for nothing. Group any obvious duplicates inside the selection, such as two rows with the same email or the same name at the same company, and make me pick one or exclude both rather than paying to enrich the same human twice. The pre-flight panel is informational and fast, so I can adjust my selection and re-check it before spending anything.

When I hit Enrich, the app calls Lusha Enrich People (Bulk) in chunks of up to one hundred contacts, because that endpoint caps at one hundred records per request. Build one person object per contact using the best identifier available in this priority order: the existing email, then full name plus company domain, then full name plus LinkedIn URL. Show progress chunk by chunk, keep the results of chunks that succeeded if a later one fails, and give me a retry action for just the failed chunk instead of re-running the whole batch.

Results land in the diff view, which is where I spend my time. Render one card per contact and, inside it, one line per field: email, mobile phone, phone, job title and company by default, plus any other property I configure. Each line has three columns, the value HubSpot currently holds, the value Lusha returned, and my decision. Classify every line as a Fill when the CRM value is empty, a Change when both sides hold a different non empty value, or a Match when they agree. Collapse Matches by default since there is nothing to decide. Approve and reject are per line, with keyboard friendly controls so I can move down a card quickly. You may offer a bulk helper that approves every Fill in the batch, because filling an empty field is the safe case, but every Change has to be decided individually and must never be swept up by a bulk action.

Make the spend auditable per row. Each record Lusha returns carries an isCreditCharged flag, so badge every contact card with whether a credit was actually charged for it, show the charged count and total for the batch, and keep a running total for my session. A contact Lusha had no data for should read as not found rather than as an error, and should still be visible so I can see what my credits did and did not buy.

Saving writes back only what I approved. Collect the approved lines, group them by contact, and push them with HubSpot Batch Update Contacts in chunks of up to one hundred, matched on the HubSpot record id, carrying only the approved properties and leaving every rejected or undecided field untouched. Immediately after each successful chunk, append the approved changes to my Google Sheet with Append Values, one row per approved field change, carrying the timestamp, the contact id, the contact name, the field name, the old CRM value, the new value, whether a credit was charged for that contact, the reviewer name and the batch id. Use USER_ENTERED so timestamps parse as dates. Show a save summary afterwards including anything HubSpot refused, and keep those rows in the view so I can fix and retry them.

Treat a company change as the most suspicious thing in the diff, not the most exciting. Reviewers regularly report enrichment data still showing an employer someone left six to twelve months ago, so when the Lusha company differs from a non empty CRM company, badge the contact as Moved company, show the CRM last modified date right beside it so I can see how old my own value is, and never let that line be approved by a bulk action. I should be able to reject the company change while still approving the email and phone that came back on the same record.

Every row badged as Moved company also gets a Draft follow-up button, and it stays available whether or not I approve the company field, because I may want to reach out before I am willing to overwrite anything. That button kicks off a background agent that reads the contact record, the old company, the new company and the new job title, and writes a short re-intro note of about four to six sentences that acknowledges the move, references the previous relationship and asks for a short conversation, with no hard pitch. The draft comes back into the app and is shown to me first. I can edit it in place, regenerate it, or discard it. Only when I press File task does the app create it with HubSpot Create Task, putting the note in the task body, associating the task with the contact, setting a due date a couple of days out, and assigning it to the record owner resolved from List Owners, falling back to a default owner I choose in settings when the contact has none.

Persist my decisions per contact and per field, keyed to the batch, so I can leave a review half finished and come back to it without losing my place, and so a second reviewer can see what has already been decided and by whom. Keep a batch history view listing past review sessions with the contacts reviewed, fields approved, fields rejected and credits charged, so the app itself answers what was touched and when. Remember rejected suggestions so the same value does not come back and ask again on the next run for that contact. Respect the rate limits by batching rather than firing one request per row: HubSpot search allows about four requests per second and Lusha enrichment allows about twenty five per second, and on a 429 from either side back off and continue.