Lead research and qualification
By Toolspoke
Finds companies that match your ideal customer profile and scores each one, with the evidence and source link behind it.
Needs a Toolspoke account. You choose which project it installs into.
- Firecrawl + HubSpot + Google Sheets
- Required integrations
- Runs when you ask
- No schedule of its own
- Write the list
- What its last step does
- Some actions can delete
- 3 of 34 actions in its tools
On this page
The short version
- What it is
- A ready-made Toolspoke agent that turns an ideal customer profile into a ranked list of companies, each with its evidence.
- What it reads
- Public pages on each company's site: home, product, pricing, careers and blog.
- Where it writes
- Rows in a Google Sheet you choose, best fit first.
- How it scores
- Each of your criteria is met, partly met or not found, with the page that says so.
- How often
- When you ask. It has no schedule of its own.
- What it never does
- Send or draft outreach, sign in anywhere, collect personal contact details, or write to your CRM.
- What you need
- A Toolspoke account, a Firecrawl key, a Google account for the Sheet and a HubSpot private app token.
What Lead research and qualification does
Lead research and qualification is an AI sales research agent. You give it an ideal customer profile and a number of leads, and it hands back a ranked list of companies with the evidence for each one's score.
Prospect research is slow because it is reading. Someone opens a company's site, finds the pricing page, checks the careers page for hiring, looks for a launch, and writes a note. Doing that for fifty companies is a week of a salesperson's time that is not spent selling.
A score from an AI tool does not fix this on its own. A lead marked 82 gives the salesperson nothing to say and nothing to check, so they redo the research to trust it. The value is the evidence.
Lead research and qualification keeps the evidence. Every criterion it marks as met comes with a line from a public page, the link to that page and the date it was read. A criterion it cannot find is marked not found, never guessed.
In practice, you get a spreadsheet you can act on:
- Companies ranked by how well they fit, highest first
- A source link and read date behind every judgement
- Buying signals and a likely use case for each company
- Companies already in HubSpot skipped before you see them
How it works
Lead research and qualification works in four steps: it reads your profile, finds and checks candidates, scores them against your criteria and writes the list. All of it comes from the agent's written instructions, which you can read and edit after you install it.
Step 1: Find candidates
It searches for companies that show the problem in their own words, not only ones that fit on paper.
Step 2: Check your CRM
It searches HubSpot by domain and name and skips anyone already there.
Step 3: Score with evidence
It reads each company's public pages and marks every criterion met, partly met or not found.
Step 4: Write the list
It adds one row per company to your Sheet, best fit first.
It reads your profile first
The profile comes from your request: industry, size, geography, the problem your product solves and any signals that matter. If the criteria are missing or contradict each other, it asks once before it searches. A search run on a vague profile produces a long list of poor fits.
It checks the CRM before the web
A company that is already a customer, an open deal or marked as declined is not a new lead. The agent searches HubSpot companies and contacts by domain and name first and names the companies it skipped, with the reason. HubSpot is only read: nothing is created or changed.
It reads a handful of the right pages
For each candidate it reads the home page, the product or pricing page, the careers page, the blog or changelog and any customer stories. It does not crawl the whole site. A few relevant pages answer the criteria, and a crawl costs more and adds noise.
It scores each criterion on its own
Each criterion gets met, partly met or not found, with one line of evidence, the source link and the date it was read. The total comes last. A buying signal it inferred, not read, is labelled as inferred. A short list with evidence is better than a long one padded to the number you asked for. The limits are listed under what it will never do.
What the list looks like
The result is one Google Sheet. Each company is a row, ranked by score, with the reason it fits, the evidence with links, the likely use case and what could not be found. A short message after the run names the three to contact first.
Lead list, run of 12
12 companies researched against your profile. 4 were skipped because they are already in HubSpot.
Contact first (3)
- Larkspur Software scores 9 of 10. Hiring three sales reps (careers page), sells per seat to agencies (pricing page), and launched an API last month (changelog).
- Brightway Systems scores 8 of 10. Customer stories name two of your competitors. Team size is not published, so that criterion is marked not found.
Worth a look (5)
- Northwind Labs scores 6 of 10. Right industry and size, but the pages show no sign of the problem you solve.
- Keystone Cloud scores 5 of 10. Fits on size and region. The blog suggests they are about to change tools, which is marked as inferred.
Skipped
4 companies skipped: 3 already customers or open deals in HubSpot, 1 marked declined
- Contact first
The highest scores, where most criteria are met and the pages show a buying signal. Each has its evidence, so the person reaching out can say something specific.
- Worth a look
Companies that fit on paper but where the evidence is thinner or a key criterion was not found. Useful for a second pass or a nurture list.
- Skipped
Companies already in HubSpot, and any that failed a hard criterion. They are counted and named, so you can see nothing was dropped silently.
Why the list may be shorter than you asked for
If fewer companies fit than you asked for, Lead research and qualification returns the ones that do and says so. It does not pad the list to reach a number, because a weak lead in the middle of a strong list costs a salesperson an hour.
If a site will not load or blocks automated reading, the company is listed with that gap named and the agent moves on. The gap is information, not an error to hide.
Who uses Lead research and qualification
Lead research and qualification suits anyone who has to find the right companies before anyone can sell to them. The examples use made-up names.
A founder doing their own outbound
Mira Okafor sells a developer tool and has no sales team. She describes her ideal customer once and gets twenty researched companies with a reason each, instead of spending Sunday on tabs.
An agency building a client's target list
Tomás Reyes runs outbound for a client and needs a list the client trusts. Every row carries its source link, so the client can open the page and check the claim in seconds.
A sales team keeping a CRM clean
Jo Adeyemi leads sales at a thirty-person company. The agent checks HubSpot first, so the team never works a company that is already a customer or was marked declined.
A B2B SaaS marketer sizing a segment
Sana Haddad is testing a new market. She asks for a small run against a draft profile and reads the not found cells, which show which criteria public pages can never answer.
What Lead research and qualification connects to
Lead research and qualification connects to three tools: it reads public pages with Firecrawl, checks HubSpot for duplicates and writes the result to Google Sheets. You connect each one yourself, and Toolspoke stores the connection encrypted, so the agent never sees the credential.
Firecrawl
RequiredExtract clean, LLM-ready content from a web page.
Authentication: API key
HubSpot
RequiredSearch and update contacts, companies, deals, and notes in the CRM.
Authentication: Private app access token
Google Sheets
RequiredRead and write spreadsheet ranges, append rows, add sheets, and create new spreadsheets.
Authentication: Sign in with Google
| Tool | What Lead research and qualification does | What it cannot do |
|---|---|---|
| Firecrawl | Searches the web and reads company pages (search, map, scrape). | Sign in to anything or read a page behind a login. It can also start a crawl, but its instructions say to read a handful of pages, not whole sites. |
| HubSpot | Searches companies, contacts and deals to find leads you already have (search_companies, search_contacts, search_deals). | Delete records: the connector has no delete action. It can create contacts, deals and notes, but this agent's instructions forbid writing to the CRM. |
| Google Sheets | Adds one row per company to the end of your sheet (append_rows). | Delete a spreadsheet. The connector can overwrite or clear a range, and the agent is told to only add rows, never to change or clear one. |
That the agent only reads HubSpot and only appends to the Sheet is an instruction to the agent, not a lock. If you want a lock, a project can switch individual actions off. Turn off the HubSpot write actions and the Sheets overwrite and clear actions, and no agent in the project can call them. The access control guide explains how.
The activity log records each call. Scraped page text is not kept, but the rows the agent writes to your Sheet are.
What Lead research and qualification will never do
Lead research and qualification is told never to send, draft or schedule any message. It researches and ranks, and a person does the outreach.
- Send, draft or schedule any email or message.
- Sign up for anything, fill in a form, or read a page behind a login or a paywall.
- Collect personal email addresses or phone numbers, or buy or scrape contact lists.
- Score a company on a guess.
- Write to the CRM: no contacts, deals or notes.
Why it does not collect personal contact details
It uses public company pages only, and it does not read social networks that forbid automated access. It does not collect personal email addresses or phone numbers, and it does not buy or scrape contact lists. If a company publishes a work contact itself, that is fine. Otherwise it names the role to contact.
What it does when the evidence is not on a page
It does not score a company on a guess. If a criterion is not on a public page, it is marked not found, not met. A buying signal it inferred from something else is labelled as inferred.
Why you may still want a lock
No instruction is a perfect lock. If you want certainty, switch off the HubSpot write actions in your project, so those calls are refused whatever a web page says.
If something goes wrong
If a site will not load or blocks automated reading, list the company with the gap named and move on. If fewer companies fit than were asked for, return the ones that do and say so. A short list with evidence is better than a long one padded to the number.
These are written into Lead research and qualification's instructions. They guide the agent, but they are not a permission. What it is allowed to touch is set by the access you grant.
Lead research and qualification researches and ranks. Who to contact, what to say and every message that is sent stay with a person.
When Lead research and qualification runs
Lead research and qualification has no schedule. It runs when you ask, with a profile and a number of leads, and it stops when the list is written.
- You give it a profile and how many leads you want.
- It checks HubSpot, reads the public pages and writes the rows.
- It finishes with a short message naming the three to contact first.
- If a site will not load, the company is listed with the gap named, and the run carries on.
Each run uses model credit from your workspace balance, and it grows with the number of companies you ask for. Start with ten, check the evidence, then ask for more. The tool calls themselves are free.
Your copy is an ordinary agent in your project. You can run it on demand, give it a schedule if you want a fresh list every week, and edit its instructions. The agent's page lists recent runs and their cost.
How to set up lead research
To start researching leads, add Lead research and qualification to a Toolspoke project, connect Firecrawl, HubSpot and Google Sheets, and give it your profile. It takes a few minutes.
Step 1: Create a Toolspoke account or sign in
Choose Use this agent. If you are new, you will be asked to sign up first, and you come straight back to this agent afterwards.
Step 2: Add it to a project
Pick the project it should live in. You get your own editable copy, and you can see the three tools it brings before anything connects.
Step 3: Connect Firecrawl
Paste a Firecrawl API key. It is stored encrypted and the agent never sees it.
Step 4: Connect HubSpot
Create a private app in HubSpot with the CRM read scopes for contacts, companies and deals, and paste its access token. It does not need any write scope for this agent.
Step 5: Connect Google Sheets and pick a sheet
Choose Sign in with Google and name the sheet the rows go to. Give the first row your column headings.
Step 6: Describe your ideal customer
Tell the agent your industry, company size, region, the problem you solve and the signals that matter, then ask for ten leads. Read the evidence before you ask for more.
Make it fit your market
Lead research and qualification's instructions are plain English and you can edit them after you install it. Installing copies the agent, so your edits stay with your copy. Open the agent in Toolspoke and tell the builder what to change, and it rewrites the instructions.
Paste any sentence below into the builder chat, or say the same thing in your own words.
Count hiring for a sales role as three points and every other criterion as one, and show the weights in the sheet header.
Skip any company outside the United Kingdom or with fewer than 20 staff, and list them under Skipped with the reason.
For each company, check the blog and changelog for a launch in the last 90 days and put it in a Recent launch column with the link.
Use the columns Company, Website, Score, Fit, Evidence, Use case, Signal and Read date, in that order.
Run every Monday at 8 AM with the same profile, and only add companies that are not already in the sheet.
This changes when the agent wakes up, so the builder edits its schedule, not the instructions.
None of these edits needs more access or touches the Never rules. They change what Lead research and qualification is told, not what the connections allow.
How it compares to other options
Lead research and qualification beats the alternatives when you need to trust and check why a company fits, and loses to them when you only need a raw list of names.
| Compared on | Lead research and qualification | A contact database | A generic AI chat | Researching it yourself |
|---|---|---|---|---|
| Shows why a company fits | Yes, a source link per criterion | Filters, no reasons | Sometimes, often unsourced | Yes |
| Marks what it could not find | Yes, as not found | Blank fields | Often fills the gap | Yes |
| Skips companies you already have | Yes, checks HubSpot first | No | No | If you remember |
| Uses your own criteria | Yes, in plain English | Fixed filters | Yes | Yes |
| Collects personal contact details | No, told not to | Yes, often bought | Depends on the prompt | Up to you |
| Time for fifty companies | One run | Minutes | Long, by hand | Days |
| Setup effort | Connect three accounts | A subscription | None | None |
A contact database is the better choice when you need volume and verified emails, and are comfortable with how the data was collected. This agent does not collect contact details, so the two work well together: the database finds names and the agent explains which of them fit.
A generic AI chat can research a company too. The difference is that it has no sheet to write to, no CRM to check and no rule that a missing fact stays missing. Researching it yourself is still the right call for a short list of accounts that matter a great deal.
Frequently asked questions
How does the agent find companies?
It searches the web with Firecrawl using your profile and the problem your product solves, then reads each candidate's public pages. It prefers companies that describe the problem in their own words over ones that only fit on paper.
Does it send emails or messages to the leads?
No. It is told never to send, draft or schedule any message. It hands you a ranked list with evidence, and a person does the outreach.
Where does the list go?
Into a Google Sheet you choose, one row per company, best fit first. It only adds rows to the end and is told never to overwrite or clear one that is already there.
Does it write to HubSpot?
No. HubSpot is searched so the agent can skip companies you already have. The HubSpot connection can create contacts, deals and notes, but this agent is told not to. If you want certainty, switch those actions off in your project (how access works).
Does it work with Salesforce instead of HubSpot?
Not as shipped, because it checks HubSpot. Toolspoke has a Salesforce toolkit, so you can ask the agent builder for a Salesforce version of the same duplicate check.
Can it use LinkedIn?
No. It reads public company pages only and does not sign in anywhere or read sites that forbid automated reading. A person can add what they learn from LinkedIn to the sheet.
How do I know a score is right?
Open the link beside it. Every criterion marked met has a line from a public page, the link and the date it was read. A criterion with no evidence is marked not found, not guessed.
Is Lead research and qualification free?
It has no price of its own, and every Toolspoke plan includes agents. Tool calls cost nothing, but each run draws on your usage balance, and a bigger list costs more. Pricing has the details.
Try it on your own market
Connect Firecrawl, HubSpot and Google Sheets, describe your ideal customer and ask for ten leads. You will have a ranked sheet with the evidence for each one.