Skip to content
Tulina
Book a demo
← Processes

Qualify inbound leads

AI lead scoring for inbound leads: each one matched to its legal entity, scored on real headcount and filed accounts, and written back with the reason.

Book a demo
Salesforce→INSEE Sirene→INPI→BODACC
Starts
New lead created
Steps
5
Checks and gates
0
A person decides
Never, runs alone

Trigger · Event

New lead created

Each new lead starts the run as it arrives, so it is already qualified before a rep opens it.

Salesforce

Claude runs the rest autonomously

  1. 1

    AI collects leads since the last run.

    Salesforce

    Salesforce

    Every lead created since the last run comes in with its custom fields already attached.

  2. 2

    AI resolves each lead's legal entity.

    INSEE Sirene

    INSEE Sirene

    Each lead is tied to its registered company on activity, location and headcount, not the name alone, or left unresolved rather than guessed.

  3. 3

    AI gathers filings and legal events.

    INPI

    INPI

    Each matched company comes back with its identity, directors, latest filed accounts and any recent legal event.

    BODACC

    BODACC

  4. 4

    AI scores each lead on your rules.

    Each lead gets a score from the thresholds your team already uses — size, revenue, sector, solvency signals.

  5. 5

    AI updates the lead with its reasoning.

    Salesforce

    Salesforce

    The lead carries the enriched fields and a note explaining its score, and a re-run updates it rather than creating a duplicate.

Inbound leads arrive as a name, an email and a guess. Someone then spends ten minutes per lead confirming the company is real, big enough and still solvent. More often nobody does, and the pipeline fills with noise.

Both halves of lead qualification already exist. Salesforce holds the lead; the French registries hold the truth about the company. The work is the join, and the join is mechanical, which makes it a good job for an AI agent with access to both.

What you get

A rep opens a lead that already carries its SIREN, real headcount, latest filed revenue and any recent legal event.

What each inbound lead is checked against

A run starts from every lead created since the previous one, pulled in a single query. The agent reads your lead schema first, so custom fields arrive in that same pass instead of costing a second lookup per lead.

Each lead is then resolved to its legal entity, a SIREN, through INSEE Sirene. Every company that resolves gets two separate lookups:

  • INPI for identity, current directors and the latest filed accounts, which is where real headcount and revenue come from.
  • BODACC for legal announcements published since the company was last checked: an ownership change, an insolvency proceeding, a change of officer.

They are two registries answering two questions, so they stay two lookups, each batched by SIREN list where the tool supports it.

What makes AI lead qualification hold up

Matching on a name alone is what makes enrichment untrustworthy: two companies share a trading name, one is dormant. Filtering on activity code, department and headcount alongside the name resolves the entity instead of guessing it. When nothing matches with confidence, the lead stays unresolved and the run reports it. A wrong SIREN would put another company's accounts on the record, which is worse than an empty field.

Writing back by upsert on the SIREN means the process can run every hour without ever creating a second copy of anything. It writes only the fields it resolved in that run, and never blanks a value because one pass came back without it.

And the note on the record matters more than it sounds: a rep who can see why a lead scored 82 will trust the score. One who cannot, will not.

Scoring on the thresholds you already use

The score comes from the rules your team already qualifies on: company size, revenue, sector, and the solvency signals the legal announcements surface. The agent keeps each input next to the result, so the note can show which filed figure or which recent event moved a lead up or down.

Where it applies

The registries in this process are French, so as written it qualifies leads from companies registered in France. The method carries over to another CRM, or to a country with its own public company registry, as long as the tools can do the same four things: resolve the entity on more than its name, read its filings, check for recent legal events, and write the result back with its reasoning.

The prompt

The real prompt behind the steps above, detailed enough that an agent with the same tools can run it as written. The Run buttons copy the full version, which also carries the steps for connecting those tools.

# Qualify inbound leads against official company data

You have read/write access to Salesforce and tools that query INSEE's
SIRENE company registry, INPI's company-filing registry, and BODACC's
legal-announcements registry. Run this for new leads.


## 1. Collect leads since the last run

- Query Salesforce for leads created since the last run. Read your own
  lead schema first so custom fields come along in the same pass,
  rather than a second query per lead.
- Batch this — one query for the whole window, not one per lead.

## 2. Resolve each lead's legal entity

- For each lead, match company name and location to a SIREN using a
  search that filters on activity code, department, and headcount
  alongside the name — never name alone. Two companies sharing a
  trading name is common enough that a name-only match produces wrong
  matches often enough to matter, especially when one candidate is
  dormant.
- If nothing matches confidently, leave the lead unresolved rather than
  guessing — a wrong SIREN is worse than no SIREN.

## 3. Gather filings and legal events

- For each resolved SIREN, fetch identity, current directors, and the
  latest filed accounts (headcount, revenue) from the company registry.
- Separately, check the legal-announcements registry for any event
  published since the company was last checked — ownership change,
  insolvency, officer change.
- These are two different lookups against two different registries;
  batch each by SIREN list where the tool supports it.

## 4. Score each lead on your rules

- Apply your team's existing qualification thresholds — company size,
  revenue, sector, solvency signals from the legal-events check.
- Keep the score's inputs, not just the number, so you can explain it
  in step 5.

## 5. Update the lead with its reasoning

- Upsert the enriched fields onto the lead, keyed on SIREN — so a
  re-run of this same lead updates the existing fields rather than
  creating a duplicate lead or a duplicate note.
- Leave the score's reasoning as a note on the record — a rep who can
  see *why* a lead scored 82 trusts the score; a bare number doesn't
  earn that.
- Only write fields you've actually resolved this run; never blank an
  existing field because this pass didn't find a value.

## Output

Report: how many leads processed, how many resolved to a SIREN, how
many scored above your qualification threshold, and any lead that
failed to resolve and why.

Why run it with Tulina

A prompt helps one person. A process moves a team.

Pasted into a chat, the prompt above works once, for whoever pasted it. In Tulina, "Qualify inbound leads" becomes a process the team shares, that keeps its history, runs on its own, and stays inside the same security perimeter as the tools it calls.

  • The whole team works on the same one

    PersonalTeamOrganisation

    A process lives where it belongs: yours, your team's or the whole organisation's. Whoever runs it can improve it, and everyone starts from the best version instead of explaining the job again.

  • Every edit is a version

    v1v2v3

    Each change is a new version with the previous one kept, so a bad edit is undone by restoring an earlier one. Every run records the version it followed, and what a run teaches goes back into the process.

  • It runs without you

    schedulewebhookon demand

    A hosted agent runs the process on Tulina's own infrastructure, on a schedule or when an event arrives, and keeps going when you close the chat. Nothing to host. Every run is tracked to its outcome.

  • Inside the same security perimeter

    vaultrolesrun log

    Credentials stay in the vault, never in the prompt. Roles decide who may run what, every call is logged, and a step a person has to approve waits for that person.

Questions about AI lead qualification

Yes — an agent with access to official company registries can resolve each inbound lead to its actual legal entity, pull real headcount and filed revenue, and score it against your own thresholds before a rep ever opens it.

By matching on more than the name — activity code, location, and headcount together — since two companies can share a trading name while one of them is dormant.

Yes, and it should — a bare score number doesn't earn a rep's trust, but a note showing the inputs behind an 82 does.

It's left unresolved rather than forced — a wrong match is worse than no match, so an ambiguous case gets flagged for a human instead of guessed.

Also on Salesforce

Other processes that run mainly on Salesforce. The tools they share with this one are lit.

See Tulina in a 30-minute demo.

Book a demo