Build it with AI
Turn targeting, research, outreach, and next actions into one prospecting workflow.
Create a system that keeps account research, contact context, outreach status, and next steps attached to the same prospect.
Introduction
What a prospecting system has to hold.
Most teams end up with a prospecting system the same way: research in browser tabs, a target list in a sheet, and outreach in a separate tool that never sees either. It holds while one person owns a territory end to end and can remember what they learned about each account. It breaks the moment two people work the same account list, because the research one did is invisible to the other and gets repeated within the month.
The reason an account was targeted lives in the head of whoever added it, so the message that eventually goes out is generic. There is no durable place for what was learned about an account, no link between a research finding and the outreach it justified, and no record that a company was already worked and declined nine months ago.
What follows covers building a prospecting system: the records it holds (target accounts, contacts, the research behind why each was chosen, outreach state, and the next step), the systems it reads (LinkedIn data sources and Gmail), and what it does not fix.
The problem
Research in one tool, outreach in another, memory in neither.
Sales engagement platforms model the sequence and treat the account as an attribute of a contact, which is backwards for anything sold to a committee. Notes in a CRM are durable but unstructured, so the research exists and is not findable at the moment it would change what somebody writes.
The records are target accounts, contacts, the research behind why each was chosen, outreach state, and the next step, and the authoritative copy of most of them already lives in LinkedIn data sources or Gmail. The prospect list in the sequencing tool and the account list in the CRM drift apart within weeks, and the reconciliation is discovered when a customer receives a cold outreach email.
The cost is not the inconvenience: volume goes up, relevance goes down, and the domain reputation pays for it.
You're likely here because
- Somebody researches an account that a colleague researched last quarter
- The reason an account was targeted lives in the head of whoever added it, so the message that eventually goes out is generic.
- When it is wrong, volume goes up, relevance goes down, and the domain reputation pays for it
What gets built
Launch builds it, Grow operates it.
Built in Launch
- • Account workspace
- • Prospect queues
- • Research views
Operated through Grow
- • Targeting
- • Outbound sequences
- • Reply handling
Systems it reads
- • LinkedIn data sources
- • Gmail
- • HubSpot
The record model
What an account workspace holds.
- Account
- The unit of work, not the contact. Anything sold to a committee is worked at the account level, and a system organised around individuals loses the thread between two people at the same company.
- Research findings with provenance
- What was learned and where it came from. A claim in an outbound email that cannot be traced is a claim that eventually gets sent to the person it describes and is wrong.
- Prior touches across channels
- Every previous attempt, including the ones from two years ago. Working an account that declined nine months ago without knowing is a different conversation from a genuine first approach.
- Suppression state
- Derived live from the CRM — existing customer, open opportunity, recent closed-lost — rather than from a list exported last month. This one field prevents the mistakes that cost accounts.
- Account tier
- Because uniform research requirements across a large list are the most common reason a prospecting process is abandoned in its second month.
- Contacts and their role
- Held under the account with the role in the buying decision, so coverage is a question about the committee rather than a count of email addresses.
- Next step and its owner
- On the account, so a territory handover transfers a position rather than a list of company names.
How it runs
From scattered tabs to one account workspace.
Step 01
Describe what a prospecting system has to do
Define the account rather than the contact as the unit of work, and say what has to be known before outreach starts. That single modelling decision is most of the difference between a prospecting system and a mail merge.
Step 02
Connect the systems of record
The CRM supplies existing relationships and closed-lost history, email supplies what has already been sent, and the calendar supplies whether anyone has ever met. Reading them prevents the three most damaging outreach mistakes.
Step 03
Build the operating surface
An account workspace holding research, contacts, prior touches, current status, and next step — with the outreach queue generated from account state rather than maintained as a separate list.
Step 04
Start narrow
One account workspace for the current target list, showing prior contact history from the CRM and email. Even with no outreach automation at all, this prevents the mistakes that cost accounts.
Step 05
Route the exceptions
An account that becomes a customer, opens a support case, or enters an active opportunity drops out of the prospecting queue automatically rather than by somebody remembering to remove it.
Step 06
Measure reply rate on accounts where the message references specific research, against those where it does not
Measure meetings booked per hundred accounts genuinely worked, not per email sent. Volume metrics reward the behaviour that makes prospecting worse.
Implementation path
Building prospecting around the account, not the message.
- 01
Define the account tier model and what research is required at each tier. Prospecting systems fail most often because every account gets the same treatment regardless of what it is worth.
- 02
Baseline meetings booked per hundred accounts worked, and the share of outreach that went to a company already in an active relationship. The second number is usually not zero.
- 03
Connect the CRM read-first and confirm the system sees every existing customer and open opportunity before a single message is drafted.
- 04
Work one tier of accounts through the new workspace for a full cycle before extending it. The tier with the most research per account is the one where the workspace pays for itself fastest.
- 05
Build the narrowest useful version first: an account workspace where the research and the outreach state sit on the same record.
- 06
Defining the tier model and the research requirement per tier is a half-day and shapes everything else. Connecting the CRM read-first so suppression is live is the non-negotiable technical step and should happen before any message is drafted. One tier working through the new workspace within two weeks; extend once meetings per hundred accounts worked has a baseline to compare against.
- 07
After the workspace has removed duplicate research, add closed-lost reasons feeding back into targeting — the accounts that said no for a structural reason should stop being worked, and almost no team removes them. Sequencing integration comes after that, driven from account state rather than from a list.
Controls
Controls that matter.
Control 01
Suppression against existing customers, open opportunities, and recent closed-lost enforced before any outreach is drafted
Control 02
Every outbound message staged for human review, with the research it was drawn from visible in the same view
Control 03
Research provenance recorded, so a claim in an email can be traced to where it came from before it is sent to the person it describes
Examples
Three things that stop being re-done.
The cold email to a current customer
Suppression enforced from live CRM state rather than from a list exported last month is the difference between an embarrassing message and one that never gets drafted.
The account worked twice
Prior research and prior outreach attached to the account mean the second person picks up where the first stopped, including the reason the account said no, rather than starting from a blank tab.
The territory handover
What was learned about each account is a property of the account rather than of the person who learned it, so a handover transfers understanding rather than just a list of company names.
How it goes wrong
Three ways prospecting systems decay.
Personalisation is generated at volume, and the research field is filled with a plausible sentence nobody verified.
Require provenance on any claim that appears in an outbound message. Generated personalisation that cannot be traced to a source is how a prospect receives a confidently wrong statement about their own company.
Activity metrics go up, meetings do not, and the team is measured on emails sent.
Measure meetings per hundred accounts genuinely worked, with worked defined in advance. Volume metrics reward exactly the behaviour that makes prospecting less effective, and they do it while looking like progress.
Suppression runs off a list exported at the start of the quarter, and a customer receives a cold email in week six.
Read suppression live from the CRM at draft time. A static suppression list is correct on the day it is made and degrades continuously, and the failure is only ever discovered by the customer.
Limitations and considerations
What better prospecting infrastructure will not achieve.
- Better infrastructure does not fix a target list aimed at the wrong companies. It will make that clear faster, which is worth having and is not the same as a solution.
- Enrichment and firmographic data are frequently stale, particularly on headcount and funding. Building outreach that quotes an enriched figure back to a prospect is a reliable way to sound automated.
- Contacting people at work is regulated differently across jurisdictions, and the rules turn on how the contact data was obtained. Record the source; take the permission question to whoever owns it.
- If the product sells to an individual rather than a committee, an account workspace is the wrong shape and a lead tracker is the right one. If the target list is small enough that one person holds it, the coordination problem this solves does not exist yet.
- Connector coverage varies: LinkedIn data sources, Gmail, HubSpot are representative rather than guaranteed, and the fields exposed depend on your workspace permissions.
FAQ
Build a prospecting system with AI: common questions.
Does this replace our sequencing tool?
Not necessarily. The common arrangement is that the workspace owns account state and research while the existing tool sends, with suppression driven from live account state rather than from a static list. That alone removes most of the damage sequencing tools do.
Should AI write the outreach?
It can draft from research that is genuinely attached to the account, which is a different thing from generating plausible personalisation. The staging step matters more than the drafting: a message that references a real finding and was read by a human before sending is the whole point.
How do we stop the research being busywork?
Tier the accounts and require research only where the deal size justifies it. Uniform research requirements across a large list are the most common reason a prospecting process is abandoned in its second month.
What counts as an account genuinely worked?
Define it before you measure it — typically research completed plus a multi-touch attempt across at least two channels. Left undefined, the denominator inflates until the conversion metric means nothing.
What should the first version contain?
An account workspace where the research and the outreach state sit on the same record. Everything else waits until that one is genuinely used.
How will we know whether it worked?
Measure reply rate on accounts where the message references specific research, against those where it does not against the baseline taken before anything changed.
Start with ARIA
Ask ARIA to build it.
Describe the website, application, workflow, or operating surface you need. ARIA plans, connects, builds, tests, and keeps refining it — inside the permissions you set.
- ARIA acts only through the systems and permissions you connect.
- Connections use scoped credentials you can change or revoke.
- Actions are recorded, and consequential ones can require approval.
Start here
Build a prospecting system around the process you actually run.
Make the account the unit of work, connect the CRM before anything is drafted, and measure meetings per account worked rather than emails sent.