Oriental Nippon Development
Back to portfolio
Case studyUS staffing & recruitment business

AI Sales Enablement: Prospect Research in Seconds, Outreach Drafted, Not Sent

Reps built prospect context by hand from job boards and directories, spending hours that weren't conversations. A research portal plus an AI agent now returns decision-makers, verified contacts and pain points, and drafts outreach for the rep to edit.

Industry
Professional Services & Staffing
Region
United States
Engagement
Lab-type ODC (dedicated team)

Outcomes

  • An afternoon of research → seconds
  • Verified contacts & pain points
  • Human-edited outreach

Stack

Python · LLM agents · Industry classification data · Web app · Email integration

Written byOriental Nippon Team

A US staffing and recruitment business starts every sale the same way: find organisations that are hiring, understand what they are hiring for, and find the person who owns that problem.

Each rep did that research by hand, prospect by prospect, across job boards, company directories and industry databases. It took hours, and the quality depended on who did it.

The problem

  • Fragmented sources. Prospect information had to be assembled from scratch every time.
  • Hours lost. Research time came out of selling time.
  • Uneven quality. The same prospect could be approached with very different levels of context.
  • Thin outreach. Emails written from thin context read that way.
BeforeAfter
Prospect researchAn afternoon per prospectSeconds
SourcesSearched one by oneAggregated and verified
Context qualityVaried by repConsistent
OutreachWritten from scratchDrafted by AI, edited and sent by the rep

What we built

  • A prospect research portal

    Open roles, company details, key people, hiring challenges, office locations and contacts, with industry-classification search and location-and-radius filters.

  • A research agent

    Give it an industry, a location and a few parameters. It returns owners and decision-makers, verified contact data, stated interests and likely pain points.

  • Draft outreach

    A personalised first draft for the rep to edit. Never sent automatically.

Technology and the role it played

TechnologyRole in this project
PythonAggregation and verification pipeline
LLM agentsResearch agent and outreach drafting
Industry classification dataSearch by industry and sector
Web applicationThe research portal for reps
Email integrationHand-off of edited drafts to the rep's mailbox

How we delivered it

We watched the reps who converted well and wrote down what "good context" meant to them. That became the specification.

  1. Observe the manual process. What a good rep actually gathers.
  2. Aggregation layer. Sources combined, with verification on contact data.
  3. Search and filtering. Industry classification and geography.
  4. Drafting layer. The rep always edits before sending.
  5. Pilot. A small group of reps, comparing conversation quality, not just time saved.
Delivery timelineIndicative total: 11–16 weeks
  1. 01Process observation

    1–2 wks

    What context a good representative actually gathers

  2. 02Aggregation layer

    4–5 wks

    Sources, verification, company and people records

  3. 03Search and filtering

    2–3 wks

    Industry classification, location and radius

  4. 04Drafting layer

    2–3 wks

    Personalised outreach generation, editing workflow

  5. 05Pilot

    2–3 wks

    Small rep group, conversation quality reviewed

Risks and how we handled them

RiskHow we handled it
Wrong contact detailsVerification on contact data during aggregation
Generic AI emails sent at scaleDrafts only, always edited and sent by the rep
Optimising for speed over qualityPilot compared conversation quality, not just time saved

Outcome

Research that took an afternoon now takes seconds, and reps approach prospects with consistent context instead of whatever they had time to find.

The outreach is drafted, not sent, on purpose. Automated sending scales the wrong thing. A personalised email nobody read before it went out is not personalised.

What the client keeps

  • Portal and agent source, in their own repository
  • Source integrations and verification logic
  • Search configuration
  • Outreach templates

Sources and further reading

  1. Guidelines for human-AI interaction, Microsoft HAX Toolkit
  2. People + AI Guidebook, Google PAIR

Facing something similar? This is the kind of work our AI, Data Science & Automation team delivers every day. Talk to us about your stack.