The solutions arm of MillerKnoll’s Lean AI.
AI Solutions Engineering is the solutions arm of MillerKnoll’s Lean AI team. Where the data-science side builds models, we take a business problem all the way to a production system — with agents, with automation flows, and often with both. AI is reinvigorating how the business looks for automation and brings problems forward; deciding whether the answer is AI, legacy automation, or a blend is exactly what this team is for.
One team, two arms
Lean AI is 11 people, reporting up through Chris Miller & Derek Torrey. It splits into a data-science side and a solutions side — this is the solutions side.
Data Science
Shabnam Nazmi · Senior Manager, Data Science
The data science arm — forecasting, predictive and machine-learning models, and the statistical analysis that turns enterprise data into decisions.
AI Solutions Engineering
Michael Pietrangelo · Director, AI Enablement & Knowledge
The solutions arm — we take a business problem all the way to an enterprise-grade, production system.
A small team, on purpose
Two internal engineers with deep roots in the systems the business already runs on — plus a lead, and two interns who carried real flagship work this summer.
Michael Pietrangelo
LeadDirector, AI Enablement & Knowledge
Leads AI Solutions Engineering and holds the business relationships. Keeps the backlog orderly and sets project priorities in partnership with Nate Mackovjak, Senior Program Manager, AI & RPA.
Shelby Corbitt
since Feb 16, 2026Staff AI Solutions Engineer
Our lead architect — she designs the systems the rest of the team executes and builds on. An internal candidate from Enterprise Apps, where she administered Atlassian, Microsoft 365, and much of the contact-center stack (Genesys); that deep systems knowledge is what lets us bridge into legacy platforms. She also mentors the summer interns.
Trevor Cline
since Jul 13, 2026AI Solutions Engineer
An internal candidate from Information Security, where he led MillerKnoll’s Identity & Access Management team from its inception. A long history with the business and the systems it runs on.
Neha & Anna — mentored by Shelby Corbitt
Two summer interns, mentored by Shelby, moving flagship work forward — their names are on repos like SAM and the IT Support agents. As the team was still forming, they were pivotal to our capacity, and to delivering more to the business.
Where they are now: Anna stayed on part time as of August 7, 2026, while she finishes her senior year at Calvin College. Neha returned to Michigan State for her junior year, and we hope she rejoins us for another internship next summer.
Reusable components, and the right tool for the job
Being lean isn’t a constraint — it’s the strategy. Build once, reuse everywhere, and only reach for AI when AI is genuinely the best answer.
The right tool, not the loudest
Not every problem needs the AI hammer. We work closely with the RPA team and reach for legacy automation, AI, or a blend of both — whichever serves the business best on cost and reliability. Choosing correctly is our job, not a foregone conclusion.
Reusable components first
Being lean works because we build once and reuse. Connectors that bridge into legacy systems get used again and again — so when a team comes to us, we already have the plumbing to get them the data they need.
AI reasons; flows keep it lean
Deterministic flows access data and shrink it before it ever reaches a model — AutoBid narrows precedent to the top 12 rather than dumping thousands of options into an LLM. Less cost, less noise, and the model focuses on the actual decision.
From mini-agent to enterprise-grade
Anyone can stand up a small agent. The friction is the data — uploading a file, re-running a prompt, exporting a report. We solve that programmatically, then scale it, harden it, and deploy it across a wider base.
This is why we’re Solutions Engineering
We aren’t tied to one vendor. We find the right solution for the work, then engineer it, and we move a build when a better option appears.
Microsoft Copilot Studio
Classic and skills-based agents where people already work, in Teams and M365 Copilot. Most of the portfolio: TracE, DJ, Oscar, SAM, AI Intake, Helix Intake, Experience Compass.
Claude Managed Agents
AutoBid runs here in production. It moved from Copilot Studio at about 30% lower cost, and the Copilot Studio build stays in step as a twin. The SKU Attributer is the next build.
Power Automate
The flows that trigger agents, shrink data before a model sees it, and carry API contracts, like AutoBid’s Product Portal path and Arthur Morgan’s inbox.
Azure Functions
For when a flow can’t go fast enough. The Data Assistant moved to a Function and went from 47–150 seconds per answer to about 6.
Atlassian Rovo
Agents native to Jira and Confluence. The Story Readiness pair checks whether a story is ready, right where the team plans its work.
Claude Code skills
Packaged capabilities teammates run on their own machines, like Knowledge Harvesting, which turns a shared mailbox into documentation.
Amazon Web Services
Hosting for agents and the front ends people use them through. With AWS alongside Azure, we build on whichever cloud a team already runs on.
Snowflake Cortex & Google AI Studio
AI next to the governed data in Snowflake, and Gemini-based builds where they fit best. First builds start next month.
Enablement is a product too
Every build leaves something behind that makes the next one faster, for us and for anyone else building at MillerKnoll.
How to build on each platform and promote DEV → TEST → PROD safely, written from real builds: Copilot Studio, Claude Managed Agents, Azure Functions, Rovo, solution export and import.
Daily follow-alongs plus step-by-step guides: clone a repo, create an agent, edit one, replatform one. New engineers ship in their first weeks.
42 skills built across the AISE portfolio plus 61 curated from Microsoft, copied verbatim from the agents that use them and free to reuse.
Knowledge Harvesting is packaged so a teammate can run it against their own shared mailbox. Scaffolding, sprint-planning and executive-summary skills do the same for delivery work.
We build the way we tell others to
AI Solutions Engineering runs agentic engineering practices end to end. We use AI to maintain our repositories, to help develop the solutions themselves, to keep our standard governance documents current, and to scaffold new agents. Even intake is an agent — the AI Intake Coordinator guides a project in before a human ever triages it. The through-line across all of it is the same: reduce human struggle.
A team that formed this year
The team forms
AI Solutions Engineering stands up as the solutions arm of Lean AI.
Shelby Corbitt joins
Staff AI Solutions Engineer, from Enterprise Apps — Atlassian, Microsoft 365, and the contact-center stack.
Two interns aboard
Neha and Anna join for the summer (through Aug 7), mentored by Shelby, on flagship builds.
Trevor Cline joins
AI Solutions Engineer, from Information Security — the IAM lead he built from the ground up.
Anna stays on part time
As the internship wraps, Anna continues with the team part time while she finishes her senior year at Calvin College.