Automate prospect research so every call starts informed
The ten minutes a rep spends on each account before a call (what it does, its rating, branches, hiring, size) done for the day's accounts before the rep logs on.
Part of the guide AI employee for sales prospecting
You can automate prospect research by giving an AI employee the list of accounts a rep will call today and letting it gather, before the working day starts, what the rep would otherwise look up by hand: what the business does, where it is and whether it is open, its rating, how many branches it has, whether it is hiring, how big it is and who to ask for. It then writes a few lines per account and sends them as one morning note.
The work being replaced is familiar to anyone in sales: open the website, check the map listing, skim the reviews, search for job adverts, look the company up in a contact tool, then try to remember all of it ten calls later. This article lists the questions worth answering before a call, shows which of them a flow can answer reliably, walks through a made-up morning, and covers the traps.
What a rep actually checks before dialling
Ask a few experienced reps what they look at and the list is short and consistent. Each item has a reason behind it:
- Is the business still trading, and where? Calling a closed shop or the wrong branch ends the conversation before it starts.
- What do customers think of it? A rating that has slipped, or a run of complaints about waiting times, is an opening for many products.
- How big is it? One site or twelve, a team of five or fifty. It changes who you ask for and what you offer.
- Is it growing? New branches and open job adverts both say money is being spent.
- Who decides? A name and a title to ask for, rather than "whoever handles this".
Which questions a flow can answer, and from where
| Question | Block that answers it | What comes back |
|---|---|---|
| Still trading, address, phone, website | Map business scan, looking up each place itself | Opening status, address, phone, website, rating |
| How many branches | Nearby places with Only this brand, counted | Branches of the brand within the area you set |
| Hiring now | Job postings, matched to your accounts | Job titles and links for accounts that are hiring |
| Company size and who decides | Apollo, from your own account | Employees, industry, founding year, people by job title |
| Anything in the news | News, from feeds or news pages you choose | Stories with words you set, such as an account's name |
| The two lines the rep reads | xMap LLM | A short brief and a suggested opening question |
How each pre-call question maps onto a block in xMap Flows.
The first and last rows are the core. The others depend on what you sell: a software firm cares about hiring, a shopfitter about new branches, a supplier to restaurants about ratings.
Where the day's accounts come from
The research only helps if it runs on the accounts the rep will really call. Three common setups:
- A table in My data or a sheet in Google Sheets that the rep or manager fills with today's accounts the evening before.
- A saved list of results the team chose to work this week.
- Yesterday's new leads from another AI employee, handed over each morning.
Whatever the source, keep it short. A brief on eight accounts gets read. A brief on eighty gets skimmed and filed.
A made-up morning
This example is invented. A rep at a firm in Toronto that sells kitchen maintenance contracts has six restaurants on today's call list, typed into a sheet the night before. At 07:00 an AI employee runs:
- My data reads the six rows from the sheet.
- Map business scan looks up each restaurant by name near its address: one has closed since the list was made.
- Nearby places, limited to each restaurant's brand, counts its branches in the city: two belong to small groups with three and four sites.
- Apollo adds company size and asks for owner and operations manager at each.
- xMap LLM, set to the whole list, writes one paragraph per restaurant and an opening question, and notes the closed one.
- An Email reaches the rep at 07:10.
The note reads something like: "Two of six are part of small groups, so ask about all their sites. The one rated lowest has many ratings and an average well under the others; ask how breakdowns affect service. One has closed: take it off the list." Ten minutes of tab-switching per account became one email read over coffee.
Adding hiring and news without drowning the brief
Hiring and news are the signals reps value most and the ones most likely to add noise. Run Job postings for the rep's city and the roles that matter to you, then use Match with my table to keep only postings from companies on the account list. Do the same with News: read a few local business feeds and keep only stories naming an account. Both then feed the same AI step, which mentions them only when they exist.
Traps that make automated research worse than none
- Writing a brief from nothing. If the lookups find nothing about an account, the brief should say so, not fill the space. Tell the AI to write "nothing found" rather than guess, or add Check the work to refuse claims the data does not support.
- Researching the whole territory every day. Research is for the accounts being called. A territory-wide sweep belongs in a separate weekly flow.
- Long briefs. Ask for three lines per account at most. A rep reading between calls needs the one fact that changes the call.
- Old lists. If the call list is not updated, the brief repeats yesterday's. Make updating the sheet part of the evening routine.
Setting it up as an AI employee
In order: Schedule on weekdays before the rep starts, My data for the day's accounts, Map business scan to confirm each place, Nearby places for branch counts where it matters, Apollo for size and people, xMap LLM writing the whole-list brief, and Email to the rep (or Slack). What lands each morning is one message: a short paragraph per account, a question to open with, and a clear flag on anything that has closed or changed. Start from Morning call list with decision makers and call notes, or Changes at my key accounts, with what to do for a weekly version on a fixed account list. The sales prospecting guide shows how this fits with the AI employees that find the leads in the first place.
Questions
How long before the rep's day should it run?
Fifteen to thirty minutes is plenty for a short list. Each account is looked up once per run, so a short list finishes quickly.
Can it research accounts in my CRM?
Yes, if the day's accounts reach the flow: a table, a sheet, or a CRM lookup block. CRM lookup can also say whether each company is already a customer.
Will the brief invent facts if it finds nothing?
It should not, and you can make sure. Tell xMap LLM to say when nothing was found, and add Check the work with a rule that every claim must come from the item's fields.
Is reading reviews part of it?
The rating and the number of reviews come with each place. Review texts are not read, so the brief speaks of how customers rate the place, not what individual reviews say.