.png)
Every organization has that one process. You know the one. It's structured, documented, governed, and still painfully manual.
In the Dompalooza 2026 breakout session, From Data to Deals: AI That Negotiates, Dan Stagnitta, strategic sourcing leader, and Taylor Will, senior information management consultant, shared how Elevance Health transformed their hotel RFP negotiation workflow by embedding intelligence in enterprise workflows rather than treating AI as a separate initiative. Their approach offers a repeatable pattern for any team wrestling with high-volume, decision-heavy processes.
The results speak for themselves. A process that once took several hours from upload to output now finishes in about 20 minutes. But the bigger story is how they got there and what you can apply to your own workflows.
Target the right workflow first
The Elevance Health team deliberately avoided what Dan called "AI as a blanket statement." Instead, they identified a specific process where repetition was the bottleneck: an annual hotel RFP involving 300 to 500 properties, 717 unique data fields per bid, and multiple rounds of negotiation.
"What we ended up discovering and what we ended up building was much bigger than that," Dan explained. "It's a repeatable AI pattern for those structured, decision-heavy workflows."
The key insight? Find the workflow where the logic is definable, the steps are known, but scale makes repetition the enemy. Then embed the decision rules directly into that workflow so the repeatable parts run automatically while humans focus on exceptions and strategy.
How to identify your target workflow
Look for processes that share these characteristics:
- High volume: Hundreds or thousands of similar decisions per cycle
- Rules-based: Clear criteria exist for most decisions, even if they live in someone's head
- Time-consuming: The process stretches longer than it should because of manual review
- Expensive: Skilled people spend hours on repetitive tasks instead of judgment calls
Once you find that process, resist the urge to automate everything. As Dan put it: "The goal is not automating just for the sake of automation, but discovering and knowing where automation belongs and where it does not."
Unify data and validation in one workflow
Taylor described how the team brought together bid data across 717 fields, historical stay data, benchmarking information, and quality and safety scores into a single workflow. But the real breakthrough was treating data validation as an embedded step rather than a separate cleanup phase.
"You can't expect one person to monitor 700 fields 500 times a year and see that...this rate field actually has the letter A in it," Taylor noted.
The workflow now uses automated filters to remove invalid bids immediately, such as missing rates, nonsense values like $999,000 per night, or undefined seasons. Alerts notify the team within minutes when a bid gets rejected, so they can request corrections before the bad data corrupts downstream calculations.
Building validation into your workflow
This approach turns data quality from a post-mortem problem into a proactive guardrail. Here's how you can apply this approach in your own workflows:
- Identify your critical fields: Which inputs, if wrong, would skew your entire analysis?
- Set automated filters: Define acceptable ranges and formats for those fields
- Create immediate alerts: Notify the right people when validation fails
- Block downstream processing: Don't let bad data flow into benchmarking or scoring
Before this cleanup, troubleshooting a single bad bid could take a week. Now, issues surface immediately and get resolved before they cascade.
Simplify pipelines to accelerate cycle time
The original workflow that Elevance Health targeted had evolved over years, with 12 to 20 intertwined dataflows where the output of the eighth flow fed back into the third. Cleaning it up consolidated everything into a cleaner setup and moved ownership to Taylor and Dan's team. The result was an app that did its work in 20 minutes (compared to the previous workflow, which took several hours to go from upload to final product).
That speed matters because hotel negotiations involve multiple rounds. Faster turnaround means more iterations, better deals, and less time spent waiting on the system.
Steps to simplify your own pipelines
If your workflow has accumulated complexity over time, consider these actions:
- Map your current flows: Document every data flow, input, and output
- Identify circular dependencies: Look for outputs that feed back into earlier steps
- Consolidate where possible: Reduce the number of separate flows
- Move ownership closer to the business: Reduce dependency on external teams for troubleshooting
Speed isn't your only goal. Simpler pipelines mean faster debugging, easier maintenance, and more confidence in your outputs.
Keep humans in the loop for final decisions
Throughout the workflow, AI performs the first-pass review: Comparing bids against historical data, benchmarking against market rates, scoring properties on weighted criteria, and surfacing recommended counters. But the human always makes the final call.
"So the automation is powering the process, but the people are powering the decision," Dan explained.
When Dan reviews a market, he sees flags indicating where counters are needed and green checkmarks where bids meet targets. He can accept the AI's recommendation with a single click or override it based on context the system doesn't have, like relationship history or strategic priorities.
Designing human-in-the-loop controls
This pattern preserves judgment for the decisions that matter most. To implement it:
- Surface recommendations clearly: Show what the system suggests and why
- Make overrides easy: One click to accept, simple editing to modify
- Track changes: Maintain a log so you can review decisions and refine rules over time
- Preserve context: Display the information humans need to make informed overrides
After deals close, saved scenarios feed datasets that compare agreed rates against actual bookings. Alerts fire when paid rates don't match expected rates, with context about availability, room type, or blackout periods.
Key takeaways from Elevance Health's breakout
The Elevance Health team demonstrated that embedding intelligence in enterprise workflows produces better results than bolting AI onto the side of existing processes. Their approach offers lessons you can apply regardless of your tools or industry:
- Start with the right process: Target structured, governed workflows where repetition drives cycle time
- Embed validation as a workflow step: Catch bad data before it corrupts downstream analysis
- Simplify your pipelines: Reduce complexity to accelerate turnaround and improve maintainability
- Keep humans in control: Automate the repetitive review, preserve judgment for exceptions and strategy
- Extend intelligence downstream: Connect negotiated outcomes to operational monitoring
The pattern works because it respects both what machines do well (consistent, high-volume comparison) and what humans do well (contextual judgment, strategic thinking). You don't need to automate everything. You need to automate the right things.
Ready to see how this workflow operates in practice? Watch the full breakout from Domo to explore the complete implementation and discover how to apply these principles to your own high-volume, decision-heavy processes.
.png)
