SAP Connect Days for AI + Data Leaders · Houston · September 2026
Healthcare and Life Sciences SAP BTP SAP Build SAP Build Work Zone Featured
How Solventum went from Signavio process mining to Joule agents in order-to-cash
Delivered 2026
Signavio showed blocked orders sitting 69 hours before anyone touched them. Solventum designed the fix with its CSRs, then built a worklist and three Joule agents to close the gap.
By the numbers
-
69 hours
To clear a blocked order, measured in SAP Signavio. About 50 percent longer than standard.
-
18 percent
Of orders blocked at the header, benchmarked in Signavio against industry
-
4 weeks
Design sessions with the CSRs who work the orders
-
3 Joule agents
Fulfillment Status, Customer Material Info Record, Phase In / Phase Out
Before
Swivel-chair triage across three systems
- One of roughly 200 customer service reps started every day re-checking the same blocked orders by hand.
- Credit and pricing blocks landed in an analyst queue that only grew. Every pricing change meant another manual touch.
- An order leader could not answer where an order was, or why, without three systems and a phone call.
- A blocked order took 69 hours to clear against a 46-hour standard, about 50 percent longer.
After
Exceptions surfaced, agents doing the lookups
- CSRs work the exceptions the consolidated worklist and the agents surface. No more morning sweep.
- One SAP Build app on BTP replaces the swivel chair, reached from a tile in SAP Build Work Zone.
- Order leaders ask Joule where an order is and why, and get a cited answer in under a minute.
- Next: the analyst's block queue arrives triaged before she opens it.
“Most enterprises pick AI use cases by intuition. Solventum measured first.”
Why this matters
This is the pattern we think most SAP shops should copy. Mine the process in Signavio, pick the use case from the data, design with the people who do the work, then build the agent on BTP behind a governed gateway. Solventum did it in order-to-cash and took it to the SAP stage with us.
The challenge
Solventum is the health care business 3M spun out in April 2024. It trades on the NYSE as SOLV, is headquartered in St. Paul, Minnesota, and stood up its own SAP S/4HANA landscape as part of the separation. That gave the ERP and Apps Transformation Office a rare clean slate, and a question most enterprises answer by gut: where should AI go first?
Order-to-cash was the obvious candidate, but obvious is not evidence. Roughly 200 customer service reps started every day re-checking the same blocked orders by hand. Credit and pricing blocks piled into an analyst queue that only grew. An order leader could not say where an order was, or why, without three systems and a phone call. Nobody had a number for what that cost.
What we did
Phase 1 was SAP Signavio Process Intelligence, not a workshop vote. Mindset mined Solventum's order-to-cash flow out of S/4HANA and benchmarked it against industry. The data pointed at blocked orders: 18 percent of orders hit a header block, and a blocked order took 69 hours to clear against a 46-hour standard, about 50 percent longer. That is the number the rest of the program was aimed at.
Next came four weeks of design sessions with the CSRs themselves. The people who would live with the result helped shape it, which is where adoption starts.
Then two builds on SAP BTP, kept deliberately outside the system that takes the order so they could be built and tested without touching it. The first is the Consolidated Blocked Worklist, one SAP Build app (CAP and Fiori) reached from a tile in SAP Build Work Zone, replacing the swivel-chair triage across screens. The second is a set of three Joule agents built in Joule Studio, each aimed at a bottleneck the mining found: Fulfillment Status, Customer Material Info Record, and Phase In / Phase Out. They deploy into Solventum's own BTP tenant and reason against live business data through a governed agent gateway. Identity-gated, logged, nothing copied.
The outcomes
The order-to-cash gap is now a measured, benchmarked number instead of a feeling, and the fix is built to that number. CSRs get a single worklist of exceptions instead of a morning sweep. Order leaders can ask Joule where an order is and why, and get a cited answer in under a minute. The agents are landing in Solventum's BTP now, with a triaged block queue for the pricing and credit analysts next on the list.
Just as important, Solventum has an operating model for the agents that follow: security and identity involved from day one, a defined intake and prioritization process, and a named owner for each agent from idea through support.
Janessa Collins, Solventum's Vice President of the ERP and Apps Transformation Office, and Tarik Hujdurovic of Mindset presented the whole journey together at SAP Connect Days for AI + Data Leaders in Houston on September 2, 2026. The deck is below.
How it fits together
Measure first. Design with the people who do the work. Then build the agent.
The order matters. Every later step at Solventum points back to a number the mining produced.
Context
SAP Signavio process mining
Order-to-cash mined out of S/4HANA and benchmarked against industry. Blocked orders came out on top.
69 hours · 18% of orders
Design
Four weeks with the CSRs
The reps who work the orders shaped the worklist and the agents. Adoption starts here, not at go-live.
~200 CSRs · Design Thinking
Build
A worklist and three Joule agents
One SAP Build app on BTP, plus Fulfillment Status, Customer Material Info Record, and Phase In / Phase Out agents from Joule Studio.
SAP Build · Joule Studio · SAP BTP
Govern
An agent gateway
Identity-gated and logged. Agents reason over live business data and copy nothing.
Live S/4HANA data · nothing copied
Solventum's own checklist for the agents that follow: bring security and identity in on day one, define intake and prioritization, name an owner for every agent from idea through support, and decide up front how value gets measured.
What we built
-
69 hours vs a 46-hour standard
Order-to-cash process mining in SAP Signavio
Phase 1 mapped the flow out of S/4HANA and benchmarked it against industry. Blocked orders came out as the top improvement candidate.
-
Replaces swivel-chair triage
Consolidated Blocked Worklist
One SAP Build app (CAP and Fiori on BTP) that pulls delivery, credit, billing and rejection blocks into a single prioritized list, reached from a Build Work Zone tile.
-
Joule agent
Fulfillment Status agent
Reads order and delivery information together to answer where a sales order stands.
-
Joule agent
Customer Material Info Record agent
Looks up customer material info records, and can create and maintain them.
-
Joule agent
Phase In / Phase Out agent
Confirms a replacement material is active and sellable, finds the sales orders and CMIR records that need updating, makes the update, blocks anything incomplete or mispriced, and produces a summary of changes.
-
Govern
Governed agent gateway
Agents reason against live business data through an identity-gated, logged gateway. Nothing is copied out of S/4HANA.
A look at the work
Presented publicly
We took this work to the SAP stage.
Solventum picked its first AI use case from the data, not a workshop vote. SAP Signavio mining of order-to-cash showed blocked orders sitting 69 hours before anyone touched them, about 50 percent longer than the 46-hour standard. Four weeks of design sessions with the CSRs followed, then a consolidated blocked-order worklist on SAP Build and three Joule agents deploying into Solventum's own BTP behind a governed agent gateway.
