How Fin (Intercom) rebuilt reps’ trust in its outbound account list

What Intercom did

The starting point

Intercom bet the company on Fin, and outbound had to find new accounts.

In a May 2024 email to staff, reported by The Irish Times, CEO Eoghan McCabe wrote: “Winning in this game will require aggression on all fronts, not just with our capital.”³

EM

Eoghan McCabe@eoghan·Mar 2, 2026

Three years ago, our future was looking pretty bleak. I jumped back into the CEO role after a two year hiatus while sick, when we were heading quickly towards negative growth. …

We changed our targets to Fin revenue. … We moved our R&D focus to be nearly 80% on Fin, while that business was still a single digit percent of our revenue.

Excerpt; cuts are marked …Read the post on X

That fall, GTM Ops was launching outbound for the entire sales team, said Alexander DeMoulin, who ran it. Accounts that looked like Intercom’s existing customers had thinned out since 2022, he said, so reps would be calling companies unlike the ones they knew.^5,6a

The problem

Sales worked 19% of the fourth list.

DeMoulin was at a chili cook-off in Sausalito one weekend in September 2024 when a Slack message arrived from his boss’s boss. It said the reps seemed to really dislike their new outbound list, and asked what had happened.^5b His heart sank, he said, and he drove home to fix it.

It was the fourth list in his four years at Intercom.^c The first used whatever data was in Salesforce, the second a cold list imported from outside, and the third good scoring and intent data mapped badly onto Intercom’s own records. Each time, he said, trust fell further.⁵

By his count, sales worked 19% of the fourth list’s accounts. It counted accounts reps were willing to send a single email.⁵ By then, sales would only give ops time for work that clearly helped them hit their number.⁶

In 2025 he said that on earlier lists he had doubted the accounts or the data sources, said nothing, and supported them anyway.⁶ In 2026 he said he had destroyed trust with the sales team three times in a row.⁷

Intercom was good at propensity modeling, DeMoulin said, but after so much lost trust the list had to match what reps thought a good account was. If they only wanted accounts that mention peanut butter on their website, he would find as many as he could.⁶

GTM Ops asked reps to describe the ideal outbound list as if they had a magic wand. It also sat behind reps on ride-alongs and asked what they look for in the 30 to 90 seconds they spend researching an account.^5,6

DeMoulin did not believe Intercom would ever get rid of its duplicate Salesforce records, and would not let that stop the list. He said the trouble came when an account looked great on the list while a duplicate record was already in a relationship with Intercom, and the sales team was strictly monogamous.⁵

Ops promised reps that, while duplicates still existed, no assigned account would hide one.⁶

The redesign

Reps wrote the rules, in their own words.

Some rules took an account off the list; the rest ranked what was left.

The rules that took an account off the list were blunt: a country Intercom never sells to, or too few employees.⁵ A slide he showed at RevFest in June 2025 also flagged any company whose description mentioned “nonprofit”, “agency”, “consulting”, “acquired” or “now part of”.⁸

The ranking rules came from the ride-alongs. A login, sign-up or “my account” button on a company’s website meant a customer portal behind the marketing site, where customers raise support tickets.⁶ GTM Ops had an AI agent visit every account’s website to check, which DeMoulin called a vibe check that ordinary enrichment data could never give.⁵

Every record on the domain was checked

GTM Ops pooled accounts from Salesforce, data providers and intent sources, and collapsed them to one company per Clearbit domain.⁵ For each domain it queried every Salesforce account on it. His slides show the checks as columns, in his own labels: Customer?, RM Owned? (owned by a relationship manager), Sales Strategy?, Open Opps? and Recent Churn, and on the podcast he added activity in the last 60 days.^5,6

Reps still searched Salesforce for duplicates themselves, which DeMoulin said was right, before reaching out.⁶

The result

Sales worked 72% of the new list in week one.

The rebuilt list went out less than a week later.

No salesperson had ever told him a list was really good and that they couldn’t wait to cold outbound to it, he said, until this one.⁶

With bad data no longer the excuse, ops could see reps spending too long on low-propensity inbound deals, cold email copy that wasn’t landing, and buyer personas that no longer fit an AI buying committee.⁵ When accounts sales had picked sat untouched, ops asked why, and sometimes a rep simply had too many inbound deals.⁶

What it took

A weekend of ops work started it, and keeping it right became an engineering job.

The first rebuild ran on GTM Ops time and data Intercom already had. Batches went from monthly to quarterly, and by September 2025 to twice a year, as a systems team took over the website enrichment ops had prototyped.^5,6

On September 10, 2025, DeMoulin posted that he was hiring the first GTM engineer onto his team in San Francisco, to improve GTM processes with AI.⁹

In August 2026 Fin posted a Staff Data Engineer, GTM role in its Data group. It owns account identity, hierarchy, deduplication, sourcing, enrichment and data-quality monitoring, the same problems the weekend rebuild worked around.¹⁰

What to borrow

1. Track how much of the list reps actually work. Check it before meetings or pipeline. Intercom went from reps working 19% of the accounts to 72%.
2. Ask your reps what makes an account worth calling, and turn the answers into rules. Intercom asked reps to describe their dream list and sat with them while they researched accounts. That’s how it learned a login button on a company’s website was a good sign.
3. Check the CRM before a rep gets an account. Rule out anything already a customer, owned by someone else or in an open deal. Intercom’s CRM was never fully clean, so it ran this check on every account, every time.

How to build a version of this in Unify

Build it in Unify

Everything above is what Intercom did. From here on, this is our suggestion for a version in Unify.

Intercom’s GTM Ops joined its data itself: Salesforce, data providers, intent sources and a website scrape. Unify combines your CRM, warehouse and product usage data with enrichment from 40+ sources, with warehouse data synced in through a sync tool or the API. A rep working the list would see an account like this fictional one.

This week, run the session with reps and list the CRM fields that make an account unworkable. The Audience, the AI Agent qualification questions and the Play are built from that session.

did this

Our version in

Rules from repsMagic-wand calls and ride-alongs

✓

A person, before any setupRevOps sits with reps and their manager and writes each exclusion and each rep check as a yes or no question

One company per domainDuplicates collapsed to one key

✓

A list of unworkable fieldsFrom the same session, the CRM fields that mark an account unworkable. A duplicate record can still hide a deal, so the rep’s first Sequence step checks the CRM for another record on the domain

Rules out: can sales work it?Country, size and six Salesforce checks

✓

Audience filters and exclusionsCountry and employee range on the Audience, with exclusions on synced CRM fields for open opportunities, current revenue, churn within 90 days, an owner or strategy, and activity within 60 days

Ranks the rest: how much do we want it?A login button and other rep checks

✓

Agent qualificationEach rep check as an AI Agent qualification question, such as “Does the website have a login or ‘my account’ link?”. Accounts that pass go forward

Assigned in batchesMonthly, then quarterly, then twice a year

✓

A reviewed batch, then one PlayRevOps reviews the results and marks the approved batch in a CRM field. A Play triggers when an account enters that batch Audience, assigns an owner and finds 1 to 10 people per company across up to five personas

Reps check, then reach outSearching Salesforce themselves

✓

SequencesThe first step is an action item to confirm the account is workable. Manual emails, calls and LinkedIn action items follow, each approved by the rep, with delays in days

Feedback both waysRules change before the next batch

✓

Two-way CRM sync and a reviewReps record failure reasons in a CRM field, and RevOps and sales managers edit the Audience and Agent questions at each review. Opportunities are created by reps in the CRM, and Unify reads them