Agentic idea evaluation, built for a global professional services firm.
A Fortune 500 Global professional services firm cut idea-evaluation time from 1 to 2 weeks down to just 15 to 20 minutes.
~99%Reduction in evaluation time
15 to 20 minPer idea, down from 1 to 2 weeks
100%Of submissions fully scored
01 · Client
A Fortune 500 Global professional services firm.
One of the world’s largest professional services companies, spanning strategy, consulting, technology, and operations, from setting a client’s strategy to building and running what it takes to deliver it.
Annual revenue>$10B
Employees>50,000+
Founded1989
Countries served120+
SectorProfessional services
Function in scopeInternal innovation & idea evaluation
02 · The problem
Where the weeks were going.
A dedicated team manually researched and evaluated every submitted idea, the same categories of work, redone from scratch, every time.
01
Manual research, every time
A dedicated team researched and evaluated every idea manually, redoing the same category of work on each submission.
02
Weeks per idea
Each evaluation took 1 to 2 weeks, regardless of how simple or complex the idea actually was.
03
Inconsistent evaluation quality
Manual assessment across 4 dimensions varied in depth and consistency from reviewer to reviewer.
04
A bottleneck on decision-making
Slow evaluation cycles meant leadership couldn’t move quickly on the ideas that mattered most.
03 · Native limits
What a real fix had to do.
Speeding up part of the manual process wasn’t going to close a multi-week gap. Four things had to hold at once.
01
Score every idea the same way
Apply a consistent, standardized framework across market potential, competitive intensity, scalability and ecosystem fit, not reviewer judgment.
02
Do real research, not templated output
Pull actual market and competitive data for each submission, not generic boilerplate.
03
Compress weeks into minutes
Remove manual research time entirely, not just speed up parts of it.
04
Keep leadership in the loop
Deliver a report leadership reviews and decides on, not an automated approval.
04 · What Lyzr built
Five agents: 4 researchers, 1 summarizer.
Lyzr built an AI-agent workflow that researches each submitted idea against 4 fixed evaluation metrics, then consolidates the findings into a single scored report.
The 4 evaluation metrics
Market potential
Market analysis and market-size estimates.
Competitive intensity
Competitor analysis and the competitive landscape.
Scalability
How far the idea scales.
Ecosystem fit
How the idea fits the surrounding ecosystem.
How the workflow runs
Sequential research
Four AI agents conduct the market and competitive research, one metric at a time.
Summary and scoring
A summary agent consolidates the findings and produces numerical scores across the 4 metrics.
Report
The output is a scored report, downloadable as a PDF.
05 · Architecture
One conversational intake, 4 sequential research calls, 1 report.
An employee submits an idea through a guided conversational intake. The system validates the submission, then triggers the evaluation workflows and a summary step that writes the results back to a central database.
06 · Controls
A structured score, not a number pulled from a black box.
Every report is built to withstand a leadership review.
Real-time validation
The conversational intake validates submissions before research begins, reducing incomplete or ambiguous inputs.
Weighted scoring model
A standardized, weighted framework computes pillar-level scores and a consolidated readiness score, the same way, every time.
Structured, traceable reports
Every report includes market analysis, market-size estimates, competitor analysis and pillar-level scores, not just a final number.
Human decision gate
The Innovation Review Team reviews every report and makes the go or no-go call. Nothing is auto-approved.
Re-evaluation on demand
If an idea is revised, the workflow re-triggers, updating research findings and recalculating scores.
07 · Re-imagined workflow
From a multi-week research assignment to one submission.
Before, an idea got assigned to a research team, who spent 1 to 2 weeks manually gathering the same categories of data every time. After, an employee submits the idea once, and a scored report is ready in 15 to 20 minutes.
Drag the handle to move from Before Lyzr to After Lyzr
Before Lyzr
Six steps, every one by hand, for every idea alike.
Employee shares the ideaInformally, with no standard intake and no queue
Researches market sizeMarket reports, opened and read by hand
Scans competitorsCompetitor sites, checked one at a time
Judges scalabilityBy analyst judgment, against no fixed scale
Checks ecosystem fitInternal notes, gathered from memory
Compiles the score by handIn a spreadsheet, formatted from scratch each time
A score, 1 to 2 weeks laterHowever simple or complex the idea actually was
Varies idea to ideaDepth and consistency shift from reviewer to reviewer
The next idea starts the same 6 steps again
Manual researchOutcomeDead end
After Lyzr
One submission in. 4 agents research it. The team still decides.
Employee submits the ideaOnce, in plain English, through a guided chat
Intake agent validates itCatches incomplete submissions before research starts
Four agents research itMarket potential, competitive intensity, scalability, ecosystem fit
Summary agent scores itOne weighted framework, applied the same way every time
A structured reportMarket sizing, competitor analysis and pillar scores, not just a number
Review team decidesThe go or no-go call stays human. Nothing is auto-approved.
Ready in 15 to 20 minutesThe same evaluation, whoever submitted it
Re-runs on revisionA revised idea re-triggers the workflow and recalculates the scores
A revision re-runs the same workflow
Research agentsScoringHuman decisionResult
08 · Results
What changed, per idea.
Same evaluation, before and after.
MeasureBefore LyzrAfter Lyzr
Evaluation time per idea1 to 2 weeks15 to 20 minutes
Research methodManual, by a dedicated teamAutomated, across 4 sequential agents
ScoringManual, variable by reviewerStandardized, weighted framework
ConsistencyVaries idea to ideaSame framework, every submission
~99%Reduction in evaluation time per idea
4Dimensions researched and scored automatically
5Sequential agents, research to final report
09 · What’s next
Built to handle revisions, not just first drafts.
Ideas rarely stop at a first submission. When an idea is revised, the same workflow re-triggers automatically, updating research and recalculating scores, so a report never goes stale as an idea evolves.
15 to 20 minutes per evaluation, down from 1 to 2 weeks
Got a use case in mind?
Platform, engineers and governance all in. We’ll map your workflow against the same 4 tests this one had to pass.
