Industrial Technology Research & Marketing Intelligence Global

Analyst research automation, built for a global industrial technology leader.

A global leader in industrial technology and infrastructure, with ~$68B in annual revenue, used Lyzr to replace manual, browser-by-browser analyst research with an AI agent that builds and maintains a complete, standardized analyst contact database.

No manual researchBuild analyst lists without searching site by site
No duplicate recordsAdd new analysts without cleaning the database afterward
No stale databasesKeep analyst information current as roles and firms change
01 · Client

A global leader in industrial technology and infrastructure.

A global technology and infrastructure company with businesses spanning digital systems, energy, mobility, and industrial solutions, serving customers across industries and markets worldwide.

RegionGlobal
SectorIndustrial technology & infrastructure
Annual revenue>$10B/year
Employees>10,000+
Founded1910
Function in scopeGlobal Analyst & Advisor Relations
02 · The problem

Where the analyst database kept falling behind.

Analysts were researched manually, one website and one LinkedIn profile at a time, a process that couldn’t keep the database complete, current, or consistent.

01

Manual research, one analyst at a time

Analyst research was done one profile at a time across websites and LinkedIn, with no faster path to a complete picture.

02

Records went stale quickly

Even once researched, contact details didn’t stay accurate for long, and nothing kept the database current as analysts changed roles or firms.

03

No standard set of fields

Records were incomplete and inconsistent, missing the basic information, like title, firm, and contact details, needed to act on them.

04

Targeted outreach became guesswork

Without complete, consistent records, building a scalable, targeted list for event invitations was slow and unreliable.

03 · Why not native

What had to change to scale analyst research.

Speeding up the same manual process wasn’t going to close this gap. Four things had to hold at once.

01

Cover the sources that matter, automatically

Reach every relevant public analyst site and social profile, without a person manually opening each one.

02

Extract the same fields, every time

Every analyst record needed the same core details: name, title, firm, email and source, so the database could be searched, verified and acted on consistently.

03

Catch duplicates before they pile up

Every new record had to be checked against the existing database before being added, so the same analyst wasn’t repeatedly captured across research runs.

04

Stay inside a clear, safe boundary

No paywalled or login-protected sites, no CAPTCHA bypassing, and no automated outreach, only what’s publicly and legitimately available.

04 · What Lyzr built

An agent that finds, extracts, and keeps the database up to date.

Lyzr built an agent that identifies relevant industry analysts, extracts their publicly available contact information, and inserts clean, de-duplicated records directly into the company’s own database.

Multi-source discovery
Scrapes preferred public analyst sites and social profiles to identify relevant analysts.
Structured extraction
Pulls a standard set of fields for every analyst: name, title, firm, email, and source URL.
Duplicate detection
Checks new records against what’s already in the database before anything is added.
Scheduled or on-demand runs
Can run on a set schedule or be triggered manually, with logging and monitoring throughout.
05 · Architecture

From a request to a clean database record, in five layers.

A request moves through orchestration, research and extraction before a single record is ever written.

System architecture
Frontend, FastAPI backend and MongoDB, plus the external Apollo.io, LinkedIn and Lyzr Agent Studio services, with twelve labeled calls from request to a written record.
Frontend AR Team User Next.js Web App Login / Register Upload CSV / XLSX Dashboard Job Detail + Export 1 2 10 11 12 Backend FastAPI Backend API Routers /auth /jobs /contacts- enriched JWT auth · owner check Enrichment Pipeline Apollo fetch LinkedIn check status updates background task Retry + JSON timeouts · 5xx · invalid JSON Lyzr Batch Pool 5 per batch · 3 in parallel MongoDB Users Jobs Companies Contacts 3 9 4 7 External services Apollo.io API Organizations bulk_enrich 10 domains / call People Search api_search People Match bulk_match work + personal people paged 10 at a time · 2s pause between pages LinkedIn public profile URL Lyzr Agent Studio Research-Domain Agent records[] prompt and model are configured in Lyzr Studio, not in the repo 4 5 6 8
1POST /auth/login
2POST /jobs {rows}
3Save job
4Upsert companies
4Find firms
5Find people, title filter and emails
6Check LinkedIn profiles
7Insert contacts
85 contacts per call, records JSON
9Update fields
10Poll GET /jobs/{id}
11GET /contacts-enriched
12Rerun AI enrichment
06 · Controls & governance

Built to stay inside a clear boundary.

Each run follows a defined set of rules and approved sources.

Public sources only
The agent never accesses paywalled or login-protected sites, and never attempts to bypass a CAPTCHA.
No automated outreach
The agent researches and populates records; it never sends an email or makes contact on the company’s behalf.
Duplicate detection
Every new record is checked against the existing database before it’s added, keeping the dataset clean as it grows.
Logged and monitored
Every run, scheduled or on-demand, is logged and monitored, so activity is always traceable.
Standard fields, every time
Every record follows the same structure, so nothing depends on who or what happened to capture it.
07 · Re-imagined workflow

From browser-by-browser research to one standing database.

Before: analysts were researched manually, one at a time, across websites and LinkedIn. After: the agent runs on a schedule or on demand, and clean, structured records are ready in the database automatically.

08 · Results

What changed for the analyst relations team.

The same database the team always needed, now built and kept current automatically.

MeasureBeforeAfter
ResearchManual, one analyst and one site at a timeAutomated, across every preferred source
Record structureIncomplete, inconsistent fieldsThe same five fields, every record
FreshnessWent stale quickly, with no way to refresh at scaleKept current through scheduled or on-demand runs
DuplicatesNo systematic way to catch themChecked automatically, before anything is added
TraceabilityUnclear where a record came fromEvery record traced back to its exact public source
ScaleLimited by how much one team could manually researchRuns continuously, without added headcount
09 · What’s next

Built to extend to other relationship databases.

The same discovery-and-extraction pattern (find, extract, de-duplicate, write) isn’t specific to analysts. The same approach can extend to other relationship databases the team needs to build and keep current.

12 weeks from discovery to a live, production system

Got a use case in mind?

Platform, engineers and governance all in. We will map your workflow against the same four tests this one had to pass.