consulting as a product

Bespoke AI Solutions: Consulting as a Product

Draft — February 25, 2026

the paradigm shift

Old Model vs. New Model

The old world of software consulting looked like this: a vendor builds a product over months or years, hosts it on their own servers, manages integrations, handles security, and sells subscriptions. The product itself is the moat — it took so long to build that customers can't replicate it. Scaling means onboarding more tenants onto the same platform, negotiating BAAs, passing security reviews, and managing multi-tenant data isolation. Sales cycles stretch to 6–12 months. The vendor owns the infrastructure, the data flows through their systems, and the client is locked in.

The new world is fundamentally different. AI-accelerated development means a skilled team can build a production-quality, bespoke application in days rather than months. But here's the critical insight: the bottleneck was never the code. It was always knowing what to build, having the data expertise to power it, and understanding the domain deeply enough to make it actually useful.

What changes:

Deployment flips inside-out. Instead of “send us your data and we'll host it,” the model becomes “we build it inside your walls.” The application lives in the client's own Azure tenant, runs on their Azure OpenAI instance, and their data never leaves their environment. This eliminates the entire security/compliance gatekeeping process that kills most health tech startups.
Speed becomes a feature, not a shortcut. A rapid build isn't a hack — it's a demonstration of deep expertise. We've already solved the architecture, data pipeline, and UX patterns. Each new deployment is a configured instance of proven infrastructure, not a from-scratch experiment.
The value shifts from software to intelligence. The application is the delivery vehicle. The real product is the domain expertise, data architecture, and ongoing intelligence layer that makes the tool actually work.

what we sell

What We're Actually Selling

It's tempting to think this is just “we can code faster than your team.” That's wrong. Here's what clients are actually paying for:

Domain Expertise + Data Architecture

We understand healthcare data — claims, provider ecosystems, CMS datasets, entity resolution across data sources. A cascading NPI match engine with specialty stemming and multi-source entity resolution isn't something a hospital IT team prompts into existence. They don't even know what to ask for. We do, because we've lived in this data.

The Blueprint, Not the Bricks

Anyone can lay bricks fast now. We're the architects who know what to build. Which data sources matter, how they connect, what questions the business development team actually needs answered, how physician liaisons work day-to-day. That knowledge comes from years in the field, not from a prompt.

Configuration as Product

The component libraries, data schemas, tool definitions, pipeline configs — that's reusable IP. Each client deployment is a configured instance of our architecture, not a ground-up build. This is why we can move in days: we're not starting from zero.

External Data & Intelligence Layer (Add-On)

For clients who want us to bring external datasets and manage ongoing data enrichment, this becomes a managed intelligence subscription. This can take multiple forms:

Public data pipelines — CMS Medicare data, NPPES provider registries, quality scores, Open Payments
Web intelligence — Google search crawls, web-scraped provider information, review aggregation
Third-party data — Trilliant Health, claims aggregators, or other commercial datasets we integrate on their behalf
Ongoing maintenance — Data refreshes (CMS releases quarterly), match algorithm improvements, new source integrations

This layer is the subscription engine. The data keeps flowing, the intelligence keeps improving, and the client doesn't need to manage any of it.

White-Glove Upsell: Bespoke Solutions

Beyond the core products, we offer custom-built solutions for specific needs:

Bespoke dashboards — Executive views, board-ready visualizations, operational dashboards tailored to their KPIs
LLM processing pipelines — Automated report generation, document analysis, unstructured-to-structured data workflows
Custom integrations — Connecting to their EHR, internal databases, or proprietary systems
Advanced analytics — Market analysis, referral pattern modeling, competitive intelligence

the products

What Our Clients Get: Three Tangible Products

Agent Data Catalog

The foundation layer. Before any tool works well, the AI needs to understand the client's world — their business rules, data sources, terminology, organizational structure, and domain knowledge.

We deliver this through a structured consulting engagement:

Discovery workshops with stakeholders across the organization
Documentation of business rules, data dictionaries, KPI definitions, and institutional knowledge
Structured knowledge repository built as an agentic-ready data catalog — not a static wiki, but a machine-readable foundation that any AI tool can query
Data source mapping — what lives where, how it connects, what's trustworthy, what's stale

This becomes the “brain” that powers everything else. Every bot, every dashboard, every pipeline reads from this catalog. It's also the consulting engagement that builds the relationship and surfaces the real needs.

Delivery: Workshops + repository + docsTimeline: 2–4 weeks

Conversational AI Bots (Teams / Web)

Standalone AI assistants deployed into the client's Microsoft Teams environment (or as web applications), each purpose-built for a specific function:

Data Reporting Bot — “Show me referral volumes for cardiology in Q4” → instant chart with drill-down. Natural language access to their data without needing SQL or BI tools.
Roster Management Bot — “Which providers in the West Valley accept Medicare?” → provider cards with contact info, network status, and recent activity.
Market Intelligence Bot — “What's our share of orthopedic referrals vs. UCLA?” → competitive analysis from claims data and public sources.
Operations Bot — Meeting prep, report generation, data quality alerts, automated briefings.

Each bot is a standalone product with its own capabilities and tool set. They share the Agent Data Catalog as their knowledge base but serve different users and use cases. New bots can be spun up as needs emerge.

Delivery: Client's Azure tenantTimeline: 1–2 weeks per bot

Dashboards & Bespoke Applications

Web applications hosted in the client's Azure environment — container apps that provide rich, interactive interfaces beyond what a chat bot can offer:

Analytics dashboards — Real-time operational views, executive summaries, trend analysis
Provider search & intelligence tools — Interactive maps, provider profiles, network visualization
Workflow applications — Custom tools for specific business processes (territory management, referral tracking, campaign planning)
Reporting portals — Automated report generation with scheduled delivery

These use modern generative UI frameworks where the AI dynamically selects and renders the right visualization based on the user's question. “Show me the data” becomes a chart. “Compare these providers” becomes a side-by-side card view. The interface adapts to the question.

Delivery: Azure Container AppsTimeline: 1–3 weeks

deployment architecture

Deployment Architecture

Client's Azure Tenant
├── Azure Container Apps
│   ├── Conversational Bots (Teams integration)
│   ├── Web Dashboards & Applications
│   ├── MCP Server (data tools & integrations)
│   └── Agent Data Catalog service
├── Azure OpenAI Service (their key, their usage, their data)
├── Azure SQL / Cosmos DB (their data stays here)
└── Azure AD (their auth, their access controls)

Our Infrastructure (optional managed service)
├── External data pipelines (CMS, NPPES, web crawl)
├── Data enrichment & matching engine
└── Scheduled updates pushed to client tenants

The security pitch to IT: “Nothing leaves your environment. The AI runs on your Azure OpenAI instance. Your data stays in your tenant. We deploy, configure, and maintain the intelligence layer.”

competitive advantage

Why This Is Better Than SaaS (For Healthcare)

Healthcare is allergic to sending data outside their environment. Every SaaS deal requires a BAA, security review, data governance approval, vendor risk assessment — the sales cycle alone can kill a startup.

Our model flips it: “We build it inside your walls.”

No BAA negotiation for data transit — the data never leaves
No vendor risk assessment for data hosting — we don't host their data
IT can inspect every container, every API call, every model interaction
They control the Azure OpenAI instance — they set the guardrails
Decommissioning is simple — it's their infrastructure

This isn't a limitation we're working around. It's a genuine competitive advantage. We skip the 6-month procurement gauntlet that every SaaS competitor has to endure.

defensibility

The Real Moat

If AI makes building fast and cheap, what prevents competition? It's the stack of five things together:

Domain knowledge

Healthcare BD, claims data, provider networks, how the business actually works on the ground.

Data architecture

CMS pipelines, entity resolution, match engines, multi-source intelligence. Built and proven, not theoretical.

Reusable component library

Tested schemas, UI patterns, bot configurations, pipeline templates. Each deployment makes the library stronger.

Speed

We've done this before. What takes a competitor 3 months to figure out, we deploy in days. That compounds.

Ongoing intelligence

The subscription layer. Data keeps flowing, algorithms improve, new sources get integrated. The tool gets smarter over time.

A competitor would need to replicate all five simultaneously. AI-assisted coding only gives them the code part — the easiest piece of the puzzle.

pricing framework

Pricing Framework

Don't price the time to build. Price the value of the capability and the cost of the alternative.

The client's alternative: hire a data analyst ($120K/yr), plus a developer ($150K/yr), plus 6 months to build something half as good, plus they still don't have the external data pipeline or the domain expertise. That's $135K+ in salary alone before they have anything working.

Component	Pricing	Recurrence
Agent Data Catalog Discovery + workshops + repository	$15,000 – $40,000	One-time setup
Conversational Bot Per bot deployment	$10,000 – $25,000	One-time + maintenance
Dashboard / Bespoke App	$15,000 – $50,000	One-time + maintenance
External Data & Intelligence	$2,000 – $5,000/mo	Subscription
Maintenance & Support	$1,500 – $3,000/mo	Subscription
White-Glove Custom Solutions	Scoped per engagement	Project-based

Typical first engagement: Data Catalog ($25K) + 2 Bots ($30K) + Intelligence Subscription ($3K/mo) = ~$55K setup + $3K/mo recurring

The subscription is the long-term play. Setup fees cover costs and build the relationship. Recurring revenue scales with the client base.

This framework positions us as domain experts who happen to build fast — not developers who happen to know healthcare. The AI is our accelerant, not our product.

← back to work·get in touch →