Lisbon, Portugal
Development Services
Contact us
ENG
Follow us
How we built an AI agent for OmniTrade Brasil that processes 1,200 construction-project CSV files per month, applies configurable scoring rules, and pushes qualified leads directly into HubSpot — turning a multi-day manual screening process into a few autonomous hours.
1,200 CSV files per month from the prospecting database, dropped into a monitored folder.
Configurable exclusion rules and scoring weights — tuned by the customer, no engineering required.
Public information on companies involved, queried per record to enrich the score with external context.
Combines CSV fields, business criteria, and public research to produce a score 0–100 with a logged rationale — then routes to one of three outcomes automatically.
OmniTrade Brasil belongs to an international group leading the engineering and construction materials market. Responsible for industrial, logistics, and sports projects, the prospecting process relied on an external project database — where data arrived in large volumes but was not optimized for sales action. All screening, entry, and qualification in the CRM depended entirely on manual labour.
About 1,200 CSV files containing information on new construction projects were imported each month. Analysts would review every record, determine whether the companies involved were relevant, and then enter those leads into HubSpot. The process took several days, lacked consistent criteria, and was impossible to scale.
The sales team was hampered by outdated prospecting methods, relying on a database that provided quantity rather than quality. All the work was done manually — making the process slow, inconsistent across analysts, and impossible to scale without proportional headcount.
Without consistent qualification criteria, every analyst applied their own judgement — producing inconsistent leads, multi-day cycles, and no audit trail of why a record made it to CRM or didn't.
Every lead had to pass through a person before reaching CRM. With volume scaling, this became the structural bottleneck — making it impossible to scale without proportional headcount growth.
The agent monitors a Google Drive folder for new CSV files and processes every one automatically. Configurable negative criteria (irrelevant project types, excluded regions) are applied first, before any analysis — cutting workload before scoring runs.
The agent assigns a score from 0 to 100 based on business criteria — combining CSV-field analysis with public research on the companies involved. Three outcomes follow automatically: high score (≥75) goes to HubSpot, medium score gets emailed for review and tagged "UnderReview", low score is discarded with a logged reason.
Before insertion, the agent verifies the company doesn't already exist. Each lead created in HubSpot includes its score, the key criteria that generated it, and contextual notes about the project. A complete execution log records every cycle: total processed, classifications, insertions, duplicates, and errors.
Monitors a Google Drive folder and automatically processes every new CSV uploaded by the team.
Negative criteria — irrelevant project types, excluded regions — applied before analysis. Customer controls the rules.
Score combines CSV-field analysis with public research on the companies involved. Tunable weights against business criteria.
High (≥75) auto-added to HubSpot. Medium triggers human review with UnderReview tag. Low discarded with logged reason.
Verifies whether the company already exists in HubSpot before adding it — prevents pollution of the CRM.
Each lead created includes its score, the key criteria that generated it, and contextual information about the project — sales sees the why, not just the what.
Complete record of each cycle: total processed, classifications, insertions, duplicates, and errors — full auditability.
A process that used to take days now takes a few hours and runs autonomously — with no analyst time spent on manual filtering.
About 10% of processed leads are classified as hot leads — delivered to the team ready for action, not waiting in a queue.
Less noise in the pipeline, more sales focus. The same channels yield better-qualified opportunities because filtering is consistent.
The team is freed from repetitive screening tasks, with time for what really matters: selling and closing deals.
Built on a Drive → AI → CRM pipeline, with HubSpot as the destination CRM and configurable scoring criteria the customer controls. No new tools introduced — the agent connects to systems the team was already using.
We offer a no-obligation introductory and assessment session. We'll identify the problem, the conditions for success, and the right approach for your organisation.