Oriental Nippon Development
Back to portfolio
Case studyUS staffing business

An AI Role-Play Simulator Inside Microsoft Teams for Sales Training

Role-play with a manager was rare and uneven, and new hires practised on real prospects. Our R&D lab built an AI counterpart tuned to the industry's real objections, with live coaching, scoring and manager visibility.

Industry
Professional Services & Staffing
Region
United States
Engagement
R&D Lab (joint innovation)

Outcomes

  • 24/7 practice inside Teams
  • Live coaching & scoring
  • Manager visibility on skill gaps

Stack

Microsoft Teams · LLM agents · Python · Scoring & analytics

Written byOriental Nippon Team

A US staffing business needs its sales reps to handle industry-specific objections in live calls. The usual way to build that skill is role-play with a manager, which depends on the manager's time, the manager's skill, and both people taking it seriously at nine in the morning.

In practice, role-play rarely happened, and new hires often got their first real practice on a real prospect.

The problem

  • Hard to schedule. So sessions were rare.
  • Inconsistent. Quality depended on who played the prospect.
  • Generic scenarios. They did not reproduce the objections reps actually hear in their industries.
  • Slow ramp. New hires practised on live deals.
  • No visibility. Managers could not see who was practising or where they were weak.
BeforeAfter
Practice partnerA manager, when availableAI counterpart, any time
ObjectionsGenericCollected from real calls
FeedbackAfter the session, if at allReal time, with suggested phrasing
Manager viewNoneScores, gaps and engagement per rep

What we built

  • A simulator inside Microsoft Teams

    Practice happens where the team already works.

  • An industry-tuned AI counterpart

    Raises the objections and hesitations reps actually meet.

  • Live coaching

    Feedback and suggested phrasing during the conversation.

  • Scoring and gap analysis

    Per-category scores and exportable transcripts for reps and managers.

One practice session:

Technology and the role it played

TechnologyRole in this project
Microsoft TeamsWhere reps practise, inside their daily tool
LLM conversation agentsPlay the prospect with industry-specific objections
PythonConversation loop, coaching and integrations
Scoring and analyticsPer-category scores, gap analysis and transcripts

How we delivered it

A simulator is only as good as the material behind it, so our R&D lab started by collecting real objections from recorded and remembered calls.

  1. Collect real objections. The raw material for the AI counterpart.
  2. Define "good". Scoring categories agreed with sales leadership before any scoring was built.
  3. Prototype the conversation. Tested with experienced reps, the harshest and most useful critics.
  4. Add scoring and gap analysis. Only once the conversation itself was believable.
  5. Pilot with new hires. Their ramp is the clearest test.
Delivery timelineIndicative total: 12–16 weeks
  1. 01Objection collection and framing

    2 wks

    Real objections, scoring categories, definitions of good

  2. 02Conversation prototype

    3–4 wks

    The loop, tested with experienced reps

  3. 03Coaching layer

    2–3 wks

    Real-time feedback, suggested scripts

  4. 04Scoring and analysis

    2–3 wks

    Per-category scoring, gap analysis, transcript export

  5. 05Pilot

    3–4 wks

    New-hire cohort, engagement visibility for managers

Risks and how we handled them

RiskHow we handled it
An AI prospect that sounds fakeBuilt from real objections and tested with experienced reps
Scores nobody agrees withScoring categories defined with sales leadership first
Low usage after launchPiloted with new hires, whose ramp needs it most

Outcome

Reps can practise at any hour without booking anyone and without risking a live deal. Managers can see who is practising and which weaknesses they are working on.

Whether it shortens the ramp for new hires is the question that matters. The client will answer that better than we can after a full hiring cycle.

What the client keeps

  • The simulator and its source
  • The objection library
  • Scoring definitions and rubric
  • Exported transcript history

Sources and further reading

  1. Microsoft Teams platform overview, Microsoft Learn
  2. Collaborative intelligence: humans and AI are joining forces, Harvard Business Review

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