A $10M federal bid, made submittable
A USACE solicitation for Scott AFB put a ~$10M opportunity on the table — behind a compliance burden that normally makes bids like it too costly to chase. AISE built the tooling that assembled ~3,115 pages of GSA price-book evidence on a holiday deadline, and put MillerKnoll in the running for a contract it could not have pursued before.
The opportunity
The U.S. Army Corps of Engineers (USACE) was soliciting commercial office furniture for Scott Air Force Base (Scott AFB) — a federal opportunity worth roughly $10 million, awarded under LPTA (Lowest Price Technically Acceptable). It is exactly the kind of contract MillerKnoll is positioned to win on product and price, and exactly the kind the business has historically had to weigh against its own capacity to even respond in time.
Federal bids carry a compliance-and-evidence burden that commercial ones don’t. Winning on price is only half of it — every quoted price has to be proven against the contractor’s GSA schedule, in a prescribed order, with a highlighted page from the official GSA price book behind each line, grouped and labeled by the workstation’s item (FID) number. For this solicitation that meant assembling evidence across 108 workstations and 6 bills of material, due before the holiday in a fraction of the normal turnaround. Done by hand, it is a multi-hundred-hour task on top of everything else a proposal this size requires.
“Responding to this was not possible without major intervention or a significant overtime lift from the team. AI and the AISE team made it happen.”
— Michael Pietrangelo, AI Solutions Engineering
What AISE built
The team proved one capability at small scale before committing to the whole job: could AI locate the exact GSA price-book row for a given part, highlight it on the original published page, and label it with the FID it belonged to? The distinction that mattered was evidentiary. A recreated page could show a price, but only a highlight on the original published GSA page stands as proof to a government reviewer. Once the tool highlighted originals directly, the team had a format it could sign off on.
The build then scaled to the full solicitation. Parallel sub-agents located price rows by product family, each verifying its own work against a rendered image of the actual page before anything was assembled. Critically, where a part number did not match cleanly, the tool did not infer a substitute — a deliberate design rule. It returned the mismatches as a gap report, sorted by how many workstations each one touched, with the likely correct part number already worked out for a human to confirm.
“For the sake of accuracy, it is better when we don’t have the agent infer. Given the scale of the task, assumptions can be dangerous. We give it a hard rule to follow that is explicit.”
— AI Solutions Engineering
The benefit to the business
The immediate result was a submittable bid on a deadline that was otherwise not makeable by hand. Hundreds of hours of manual page-pulling collapsed into an assembled, verified response — with no added headcount and no scramble for unplanned overtime. That, on its own, was the difference between competing for a ~$10M federal contract and passing on it.
The larger benefit is capacity. Organizing a federal response into the exact structure a solicitation demands — the step that usually makes these opportunities too expensive to pursue directly — is now something AI can carry. That lets the internal team touch far more RFPs than headcount alone would allow, and that capacity flows outward: the team can back dealers on government deals and offer a level of support that was not practical before.
What’s next
The pilot is becoming a program. MillerKnoll Design Services has formed a cross-functional team to turn it into a durable capability, and the full build moves into development the first week of October 2026. It will also go beyond lowest-price bids: a DFAS Indianapolis opportunity is judged on best value, so the tool has to handle more than one kind of evaluation. Success will be measured the way the business measures bids: win rate before and after, dollars won, and how many dealers take part.
The team is now mapping the process end to end — a process map of where AI plugs into each stage of a federal proposal — so the capability generalizes beyond a single turnaround. Federal is the focus today; the same structure-and-evidence discipline extends naturally to state and local government work as the next scope expansion.