Data & AI
for Healthcare.
A 90-minute working session — what we're seeing, what good looks like, and how to think about the road ahead.
Learning Session · 2026
Today's session
A 90-minute working session.
Introductions
Meet the team you'll be working with.
Setting the stage
What we've heard so far, and what we hope you leave with.
Lessons from the edge
Eight principles from enterprises adopting AI in production.
The platform landscape
Where the market is converging — and what no platform solves for you.
Payer use cases & discussion
Where AI is creating value across plan operations and member experience.
Path forward
Your questions, our reactions, what might come next.
With you today
Senior practitioners. Direct access.
No rotating consultants.
Andy Sweet
Overall Practice Leader
Executive sponsor. 30+ years leading enterprise tech transformations and modernization programs.
Ben Titmus
AI Data, Platforms
& Infrastructure Lead
Modern data solutions development across regulated and consumer industries.
Jake Barger
Director, AI Innovation
& Agentic Engineering
ML, agentic systems, and semantic layer architecture. Deep technical advisor across engagements.
Anna Nguyen
Senior Data Architect
AI Data Platforms
Deep Microsoft and Fabric expertise. Architecting cloud-native data foundations.
Setting the stage
What we've heard — and what we hope you leave with.
What we've heard
What we hope you leave with
Lessons from the edge
Eight principles from enterprises in production with AI.
01
Think big. Start small.
Anchor on outcomes, not AI novelty.
02
Cross-functional wins
Embed domain experts with engineers.
03
Thinking vs. doing
AI reasons. Systems execute. Don't blur them.
04
Governance = speed
Clear rules accelerate adoption. Ambiguity stalls it.
05
Design for change
Models change faster than projects.
06
Hands-on first
Leaders use AI to lead AI.
07
Agents = digital workers
Onboard, supervise, retrain.
08
Leadership sets pace
Permission, expectation, momentum.
The pattern
AI doesn't fix bad data. It amplifies it.
of AI projects never reach production
Not because of models. Because of data foundations.
of GenAI pilots show no measurable P&L impact
MIT Project NANDA — the gap is data readiness, not intelligence.
of agentic AI projects abandoned by 2027
Gartner — without governed semantic foundations.
The pattern is consistent across industries: AI projects fail at the data layer, not the model layer.
Platform landscape
The platforms are converging.
Different philosophies — same conclusion. Every major platform is building a semantic layer.
Microsoft Fabric IQ
Top-down ontologyFull business ontology — entities, relationships, rules, actions — built on Power BI semantic models.
Snowflake Cortex
SQL-native viewsSchema-level business concepts inside Snowflake. Autopilot auto-generates and maintains views.
Databricks Unity Catalog
Open metric viewsCentralized, SQL-addressable metric definitions, reusable across dashboards, AI agents, notebooks.
Google BigQuery
AI-native platformGemini-powered agents, Dataplex governance, Looker semantics, autonomous embeddings.
The uncomfortable truth
No platform builds the semantic layer for you.
The model is replaceable.
The semantic layer
is your moat.
Think of it like onboarding a new hire.
Before they answer questions on behalf of your company, they have to learn:
- Which "revenue" number to use for which audience.
- What "urgent" actually means in your culture.
- Who to loop in before certain decisions.
- How departments define the same term differently.
Your AI deserves the same onboarding. The semantic layer is how you deliver it.
How we build agents
The 7-Layer Agentic Framework.
What "good" looks like for a payer
A maturity arc — most payers sit between Crawl and Walk.
Stage 01
Crawl
Foundation
- Cloud data platform stood up (Fabric, Snowflake, Databricks).
- Governed source-of-truth domains — claims, members, providers, plans.
- AI policy and approved tooling defined.
- First production use case proves the pattern.
Stage 02
Walk
Scale
- Semantic layer formalized — agreed definitions for plan, member, episode, benefit.
- Multiple production use cases sharing one foundation.
- Self-service analytics with governance guardrails.
- AI / agent operations team — monitoring, evals, drift, escalation.
Stage 03
Run
Differentiation
- Agentic workflows across claims, prior auth, member service.
- Real-time decisioning at the point of action.
- Continuous learning loops refine vocabulary and models.
- AI capability becomes a hiring magnet and board-level differentiator.
Where AI creates value for payers
Six categories paying back fastest in PBM-adjacent businesses.
Claims & adjudication intelligence
Anomaly detection, automated edits, denial prediction, root-cause analysis on the claims pipeline.
Member 360 & eligibility assist
Conversational interface for member services and benefits verification — pulls from multiple systems live.
Drug pricing & cost analytics
Natural-language analysis of medication costs, manufacturer trends, and plan-level cost optimization.
Plan performance & compliance
Automated reporting, compliance monitoring, audit-ready summaries — frees analysts from deck-building.
Prior auth & document intake
Document classification, extraction, and prioritization — reduces cycle time on highest-volume workflows.
Manufacturer & PBM reporting
Audit-defensible reporting packages — replaces hours of manual deck construction with minutes.
Proof points
Patterns that translate directly to payer operations.
AstraZeneca
Global pharmaceutical
10 brands
× 72 markets
Outcome
Automated 200-slide manual market analysis presentations across global commercial teams.
Field teams shifted from report consumers to active data explorers.
Reckitt
Global hygiene, health, nutrition
450% ROI
24,000 hrs/yr
Outcome
Business users finally answering driver questions on top of an existing Power BI investment.
24,000 hours per year saved on global performance reporting.
Eversana
Life sciences commercial services
Beat the
competition.
Outcome
NL-driven pricing analytics differentiated against direct competitors.
Democratized access for global pricing strategists.
Haleon
Global consumer healthcare
24/7
self-service
Outcome
Eliminated manual reporting overhead on a complex harmonized global dataset.
Enabled 24/7 self-service insights across categories.
Path forward
Where this could lead — if it's useful.
Phase 0
Assessment & roadmap
4 weeks
- Architecture assessment of current Azure / Fabric environment.
- Future-state reference architecture.
- Gap analysis with prioritized recommendations.
- 12–18 month roadmap with phased budget.
Phase 1
Platform hardening
6–8 weeks
- Pipeline architecture standardization.
- Observability & monitoring framework.
- Fabric optimization & cost controls.
- Security & compliance hardening.
Phase 2
Use case development
8–10 weeks
- Highest-priority use case from Phase 0.
- Semantic layer & data foundations for that case.
- Agent prototype with governance guardrails.
- Production rollout plan and team handoff.
Or — nothing at all. No rush on our end. Better questions are a fine outcome.
answerrocket.com