FlowsBook a callSign inCreate free accountSign up
AI employees and automation

An AI agent for company research: a plan, the digging, then a report

Before a pitch or a partnership, an AI agent can plan what to find out about a company, work through each step with its tools and hand you a report with its sources.

Part of the guide AI employees

By the xMap team · · 5 min read

An AI agent for company research takes a question such as what should I know about this company before Thursday's meeting?, writes itself a to-do list, works through it step by step with the tools you give it, and ends with a report. In xMap Flows that is the AI agent with How it works set to Work through a big task, tools such as Browser, News, Job postings and Nearby places linked to it, and an Email or Make a report after it.

It is for account executives, partnership managers, investors and consultants who prepare for meetings with companies they do not know well. The article covers what the agent does at each stage, an example, what it will not do, and how to read its working.

Why a plan beats a single question

Asking a chatbot about a company gives you one answer from what it remembers. Research is several questions in a row, where each answer changes the next: find their branches, then see which opened recently, then check whether they are hiring for those sites, then read what the press said. A big task agent keeps that thread. It writes the plan first, keeps notes between steps, can add steps when something it finds deserves a closer look, and stops at the limit you set.

The tools that suit company research

  • Browser opens the company's public website the way a person would, searches it, clicks through and brings back the fields you ask for, such as their list of locations or their product lines. It does not buy, book, sign up or type personal details, and it obeys a site's rules for automated visitors.
  • News reads business news feeds or news pages and keeps only the stories about the company.
  • Job postings shows what the company is hiring for and where, a fair sign of where it is growing.
  • Nearby places counts its branches in a city, or its competitors around one of its sites, from xMap's places database.
  • Apollo, with your own Apollo account, adds the company's size and industry and the people in the roles you name.

Write {{agent: the company or place to look at}} in a tool's setting so the agent fills it on each call.

A worked example (made up)

A packaging supplier in Monterrey has a meeting with the purchasing director of a growing bakery chain in Mexico City. The account executive sets up the research the day before.

  1. Write the task in the agent's instructions: prepare a one-page brief on the bakery chain named in the task: how many branches in Mexico City and which opened in the last year, what they are hiring for, recent news, how their branches are rated, three questions for a packaging supplier to ask; give the source of each fact.
  2. Link Browser (start page: the chain's website), Nearby places (bakeries, brand set to {{agent}}), Job postings and News to the Tools port.
  3. Set Most steps to 12.
  4. Add Make a report after the agent so the brief arrives as a PDF.
  5. Press Run once.

The agent's plan has seven steps. It reads the store list on the website, counts branches with Nearby places and notes four that opened in the last year where the data has opening dates, finds two postings for bakery supervisors in new districts, reads one news story about a new production site, and compares ratings across branches. It then writes the brief: what it found, where each fact came from, and the questions. The run panel shows the step it is on while it works, such as Step 4 of 7.

What it will not do

  • It will not invent facts to fill the brief. When a tool finds nothing, it says so; tell it in the instructions to leave a section empty rather than guess.
  • It reads public sources and your own accounts only. Private filings, paid databases and anything behind a sign-in are out of reach unless you connect them.
  • It stops at Most steps, or when the run's credits run out, and still reports what it has. A test run stops after three steps and says so.

Reading how it worked

Open the run and press See how it answered. The plan shows as a checklist: done, nothing found, or not reached. Each step opens to show the tools it used, what it gave them and what came back, its notes and the time it took. If a fact in the brief looks wrong, this is where you find which step produced it.

Mistakes to avoid

  • A task with no shape. Research this company sends the agent wandering. Name the sections you want and how long the brief should be.
  • Too few steps. Five steps is not enough for a company with many sites. Start at 10 to 15 and look at where it stopped.
  • Every tool at once. Link only the tools the question needs; each extra tool is one more thing to decide on at every step.
  • Trusting a count as complete. Branch counts come from what the data lists. Say xMap's places database lists, not they have.

For research you repeat on a list of key accounts every month, the changes at my key accounts and watch my target accounts templates follow each company on a timetable. Companies hiring transformation leaders, briefed by an AI agent shows a brief written per company. The sales prospecting guide puts this research in the wider prospecting routine.

Questions

How long does a research task take?

Usually a few minutes. A long task carries on in the next parts of the run, about an hour at most, and the run shows the step it is on.

Can it read a company's website?

Yes, with the Browser block: it opens public pages, searches and clicks, and brings back what you ask for. It does not buy, book, sign up or enter personal details.

Can I run it on a list of companies?

Yes. Put the companies in a table in My data and use a flow that runs the research once per company, or ask the agent to work through the list in one task.

Who can see the agent's notes?

Only the flow's owner and your organization's admins. Keys and tokens are always hidden.

Templates that do this

Read next

More on ai employees and automation · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.