Optum Value Connect
AI-guided provider workflows inside Optum's value-based care platform, publicly launched February 2026.
read public news articles
00
problem
Providers and care teams often work across fragmented systems to document care opportunities, build care plans, record notes, coordinate referrals, conduct interventions, and manage medication-related tasks. The result is a workflow that is cognitively heavy, repetitive, and difficult to complete efficiently.
solution
The platform should not feel like another clinical form, dashboard, or chatbot. It should feel like a guided workspace that helps users move from identifying work to completing it with less friction while keeping them in control of high-stakes actions. Unify care-opportunity documentation, care planning, note-taking, referrals, interventions, and medication-related work into one guided workflow. Let AI draft, suggest, summarize, and structure work inside the flow while preserving clear review moments before anything becomes official.
Context
I led design for multiple products inside Optum Value Connect, the company's AI-first platform for value-based care, publicly launched in February 2026. It helps payers and providers identify population health opportunities, prioritize interventions, and route the next action into a care team's existing workflow instead of a separate dashboard. Interface details, internal metrics and unreleased work remain confidential; visuals shown here are generalized or recreated.
My Role
Principal Product Designer. I led design across multiple products inside Value Connect. I directed a team of designers on several of them, and worked as the hands-on designer on others. On my own surfaces I owned information architecture, journey mapping, interaction design, workflow logic, prototyping and handoff. Across the team's, I set the AI interaction standards (how a suggestion is labelled, what a review moment has to look like, when the product is allowed to act without being asked) and held the work to them through critique, pairing and handoff.
Assistance inside the flow, not a chatbot beside it
I rejected the prompt-driven pattern the category was defaulting to. The AI summarizes the chart, transcribes and drafts the outreach record, pre-fills the drug therapy form, proposes care plan goals and generates the clinical note from what the flow has already captured, all in place. There is no separate surface to switch into and no prompt for a clinician to compose.
Provenance: who suggested it stays in the record
The system could draft anything and commit nothing. Every AI output carried its source, and the clinician's confirmation was the only action that wrote to the record. Where a form was auto-filled from something the pharmacist had actually said, it was attributed to the pharmacist, not to the model. Same automation, different provenance. Collapsing the two would have been easier to build and would have quietly cost the product its auditability.
Users and Constraints
Users included providers, care teams, and clinical operations stakeholders. Constraints included healthcare privacy and governance, multiple user roles, dense workflows, enterprise-system limitations, accessibility needs, and the need for AI outputs to remain reviewable and trustworthy.
Key Insight
In a regulated setting the hard question is not how accurate the model is. It is whether a clinician can see where a suggestion came from and reverse it in one step. We designed for attribution and reversibility first, and treated model quality as necessary but not sufficient, because an unattributed correct answer still has to be re-checked by hand, which means it saved nobody any time.
Design Principles
Keep providers in control.
Use AI assistance to reduce burden, not replace judgment.
Make outputs reviewable, editable, and confirmable.
Reduce task switching across related workflows.
Show only what matters now through progressive disclosure.
Impact
Optum publicly reports Value Connect customers seeing 35% lower medical spend, 29% fewer ER visits, 28% fewer preventable readmissions, 17% lower pharmacy costs and 20%+ improvement in care gap closure (Optum, February 2026). Separately, internal workflow analysis during design suggested some complex tasks could be completed up to approximately 80% faster than previous processes.
Reflection / Takeaway
This work reinforced that in regulated environments, the success of AI depends less on novelty and more on whether the interface helps people feel guided, informed, and still fully in control.
On What Is Shown Here
Value Connect's interface is Optum's confidential product work, so the screens are not shown. Optum's own material covers what the platform does and the results it reports: how Optum uses AI to drive affordability and accelerate value-based care, and the Value Connect product page. I am happy to walk through my part of the work in detail in an interview.
