AutoBid / Decisions Engine
From a $200K buy decision to a specials decision engine in production — 91.7% agreement with the quoters in its first 30 days.
The story, in brief.
A vendor wanted ~$200–240K a year to evaluate MillerKnoll specials with AI. The Lean AI team built it internally instead, and it went into production on August 22, 2026. In its first 30 days it evaluated 5,700 real dealer requests, agreed with the quoters’ actual decisions 91.7% of the time, and helped cut quote response time from about 72 hours to 26.1.
How it’s built.
One Copilot Studio agent, grounded purely in the Confluence Specials Knowledge space — instructions and Confluence, nothing else. Its job was to vet the knowledge and prove an internal build could scratch the surface and pass the smell test.
Steps through 8 generations — from Stage 0 to Stage 7.
What AI actually does here.
Knowledge audit at scale
Read 3,300+ Specials articles to surface rule conflicts, complexity-label mismatches, trends, and gaps — distilled into the de-facto spec and the determinism-first article template.
Pricing & upcharge analysis
Analyzed a two-month, ~9,349-line Jarvis quote export to validate the 10% rule, the ~82%-Simple distribution, and the baseline upcharges — the controls and pricing-inference ground truth.
Evaluation engineering
Built control-question evaluation suites — grown from an original 40, to 100 at UAT, to 400 today — targeting nearly 1,000 by end of July to test against the real breadth of Specials volume, not a sample.
Built with the business, not for it.
The engine is only as good as the documented rules and ground truth behind it. These are the partners who supplied the knowledge, the data, and the real-world testing that made it trustworthy.
Nate Zylstra
The business-side owner and day-to-day partner. Owns the Specials Knowledge base and the Jarvis quoting pipeline that is the project’s ground truth; authored the approved-bid-price logic, brokered the Product API credentials, and supplied the pricing and complexity rules plus the Jarvis quote set used as the eval baseline.
Jack Clark
The original POC user and one of the first three TEST testers. Authored and edited dozens of deterministic Specials KB articles (including the hinged-modesty rule the worked example lands on), confirmed the 10% department-policy upcharge documented nowhere in Confluence, and caught the upcharge-base bug.
Nate Mackovjak
Program oversight and connector across the project. Drove the push to embed the engine in Product Portal and was first to name real-time base pricing as the Phase-3 requirement, brokering the cross-team moves that unblocked the build.