Astrodata  /  [Client]
Confidential proposal
Proposal · May 2026

Proposal for [Client] Hotels

Strategic data and AI advisory, with a path to build.

Advisory partnership
$XXXXX / month · 30 days notice
Build engagements
Scoped per project
Approach
Ongoing partnership, build alongside
01

The opportunity

In a partnership with Astrodata, [Client]’s 2026 technology strategy gets the architecture right. Data readiness is critical to the success of any AI model, and the unified guest profile is the foundation. Commercial execution (B2C and B2B) is the highest business priority, and every agent in the 2026 to 2028+ roadmap depends on a governed, scalable, and comprehensive data layer.

Astrodata is a partner who has already shipped the same architecture in production, who can compress the timeline from “directionally right” to “in market,” and who can stay long enough to make sure the build matches the vision.

02

Why Astrodata

We are a 30-person data + AI consultancy with a senior-to-principal team and active partnerships with Snowflake, Omni Analytics, and Anthropic. We have built the patterns [Client] is going to need:

Teladoc Health

A unified analytics platform powering agentic and conversational workloads at Teladoc Health, the largest virtual care provider in the United States.

Kyruus Health

An agentic search experience at Kyruus Health that lets 150 million health plan members find providers by asking questions in natural language, and gets answers back and powers appointment scheduling.

Decision Resources

An agent at Decision Resources that lets business process managers explore ERP data, enrich it with external context, and take action on forecasts and recommendations.

03

Conceptual Architecture

Astrodata proposes to integrate [Client]’s transactional systems across every key domain (sales, marketing, operations, etc.) in a cloud data warehouse to establish a Customer 360 and Enterprise 360 vantage point. A semantic layer will define contextual meaning and relationships between data elements, which is essential for both business and AI users to engage core datasets. Conversational and agentic AI will comprise the control plane allowing [Client] staff to analyze business opportunities and take action in transactional systems. Activation will be facilitated by a reverse ETL solution that allows [Client] staff, with AI support, to define data extracts (like customer segments or room pricing tables) and automate data movement back into transactional systems.

AI control plane over a unified customer data warehouse [Client] staff interact with an AI control plane (circle, left) that reads from a unified data warehouse (pill, bottom) to analyze, and writes to four transactional system groups (marketing, sales, revenue, operations) shown as circles in a row to act. The transactional systems feed the warehouse, completing the loop. Hover or tap each transactional system for the underlying tools. [Client] Staff AI Control Plane Act Analyze Transactional Systems Marketing systems Marketing Sales systems Sales Revenue systems Revenue Operations systems Operations Data Warehouse Customer 360 · Enterprise 360
AI control plane over a unified customer data warehouse
04

The engagement

We are proposing an ongoing advisory partnership alongside a flexible build capability that activates when [Client] is ready. The advisory relationship is the constant, and build work happens in parallel, scoped to what the business needs and when it needs it.

Advisory partnership · The constant

Ongoing strategic advisory

$XXXXX per month · 30 days notice from either party

A consistent Astrodata presence in [Client]’s working sessions, architecture decisions, and vendor conversations. We stay close enough to the business to know when the data is ready, when the organizational conditions are right, and when it’s time to move.

What we deliver, on an ongoing basis

  • Architecture guidance. A vendor-evaluated recommendation for the data warehouse layer (Snowflake, Microsoft Fabric, Databricks, BigQuery), the ingestion layer, the reverse ETL layer, and the agent runtime. Total cost of ownership modeled across three years, defended against the constraints that matter to [Client]: data residency in Mexico and the build roadmap ahead.
  • Unified guest profile data model. The semantic model for guest_uuid: which entities resolve to it, which source systems feed it, which downstream agents and tools read from it. Delivered as an entity relationship diagram plus a written specification and implementation recommendations, maintained as the business evolves.
  • Agent roadmap, sequenced and gated. Which agent activates when, gated on which data layer milestone. The Revenue Agent, Pre-Arrival Agent, Voice Reservations, and WhatsApp co-pilot each get a readiness checklist covering data dependencies, integration dependencies, policy and human-in-loop design, and success metrics.
  • Commercial scoreboard. The metrics framework that tells [Client] whether the platform is paying for itself: direct booking lift, cost per acquisition by channel, advisor-attributed revenue, group conversion rate, homes pipeline velocity, and guest LTV by segment. Defined early so every build decision is oriented toward measurable outcomes.

How we work

Weekly working sessions with operational owners. Async working documents shared in real time. Periodic on-site visits in Mexico City or at the property. A standing readout cadence with the leadership team so strategy stays connected to execution.

Advisory retainer

$XXXXX per month. Either party can close the engagement with 30 days’ notice. Travel billed as reimbursable expenses if approved in writing.

Build engagements

Scoped and contracted as the business is ready

Scoped per engagement

Advisory work surfaces the right moments to build. When the data is ready, the use case is clear, and the business priority is aligned, we propose a build engagement. Build work is scoped and contracted separately, in parallel with the advisory relationship, and sized to what [Client] needs at that point in time.

The work below represents some of the initial projects we expect to build together, activated in the order that makes sense for the business.

Initial projects

  • Foundation. Stand up the warehouse and ingestion layer. Land Opera Cloud, Synxis, dailypoint, Duetto, FareHarbor, and ROH into the chosen warehouse. Build the guest_uuid resolution layer. Replace Power BI silos with the unified semantic model. The first deliverable is a live dashboard the Revenue Manager actually opens every morning.
  • Revenue intelligence. Push LTV scores, segments, and preference tags back into dailypoint, WhatsApp, and Duetto. Ship the Revenue Agent in human-in-loop mode. The Revenue Manager sees a daily briefing with recommended rate moves and one-click apply to Duetto. We measure pace lift against control weeks.
  • Guest-facing agents. Ship the Pre-Arrival Agent (drafts personalized messages 7 days out, staff approves) and the WhatsApp AI co-pilot (drafts replies in real time, staff approves). Pilot Voice Reservations for after-hours capture.
  • B2B and autonomous operations. Expand to B2B: TA module intelligence, ROH group scoring, HubSpot homes pipeline enrichment. Move the most reliable agents from human-in-loop to autonomous-within-policy. Add the next data sources (Symphony POS, Book4Time, Quore) as the use cases demand.

Investment

Each build engagement is sized and contracted when we get there, informed by the advisory work and the conditions on the ground. Investment is scoped more accurately as each engagement is activated.

Staffing

Astrodata staff may include senior or principal level resources in the following roles:

  • Data architects
  • Data engineers
  • Analytics engineers
  • AI engineers
  • Other product and data specialists
05

Engagement economics

A pricing framework, not a price tag.

Most Astrodata clients start in one engagement model and end up using two or three across a year. That’s by design. The four models below are levers, not lanes, and they’re combinable. For [Client], we anticipate beginning with Strategic Advisory and layering in Project or Retainer work as build engagements come into focus.

Model 01 · Most popular

Strategic Advisory

$XXXXX per month · 30 days notice
  • Architecture, vendor selection, data modeling, and agent roadmap, on an ongoing monthly basis
  • Weekly working sessions with operational owners
  • Optional periodic on-site visits in Mexico City
Model 02

Fixed Bid Projects

Well-scoped point solutions
  • Greenfield builds, migrations, and integrations with clear acceptance criteria
  • Crisp scope, transparent deliverables, hard handoff with documentation
  • Indicative ranges in the sizing table below
Model 03

Time and Materials

Discovery-heavy work
  • For projects large or ambiguous enough to be discovered as we go
  • Principal architect $XXXXX/hr · Senior engineer $XXXXX/hr · Offshore engineers $XXXXX/hr
  • Transparent burn, weekly checkpoints, no minimum commitment
Model 04

Retainer Pods

Fixed monthly fee, flexible focus
  • Quarter-time, half-time, full-time, or multi-resource pods
  • Goals and timeline agreed up front; weekly direction can flex
  • Indicative monthly $XXXXX to $XXXXX per member, depending on pod composition

Project sizing, indicative ranges

Small

$XXXXX to $XXXXX

2 to 4 weeks

Discovery, architecture audits, vendor evaluations, data model design, ROI scoping sprints.

Medium

$XXXXX to $XXXXX

6 to 12 weeks

Data foundation build, semantic layer, one agent shipped in human-in-loop mode with measurable lift against control.

Large

$XXXXX to $XXXXX

3 to 6 months

Multi-agent revenue intelligence platform with reverse ETL, governed automation, and the commercial scoreboard live.

Combinable. The expected [Client] path is to start with Strategic Advisory, layer in a Fixed Bid Foundation project when the architecture is settled, then dial back to Advisory once execution is steady. Recommended sequencing for 2026 is in §06 below.

How we set these ranges. The four models are how Astrodata structures every client engagement, not custom built for [Client]. Hourly rates and retainer brackets reflect what comparable hospitality and SaaS clients have paid for similar work in the last 18 months. Project sizing brackets are calibrated against actual engagement scope: small is one or two senior architects for a discovery sprint, large is a multi-pod build with reverse ETL and an agent runtime in production. The board is welcome to push on any number here.

06

The proposed roadmap

Direction, not perfection. A working hypothesis we’ll refine together.

The shape of 2026 is three quarters of compounding work. Q2 is for architecture and the first measurable agent. Q3 is for the data foundation and the Revenue Agent in human-in-loop. Q4 is for guest-facing agents and the operational scoreboard. The pace is dictated by data readiness, not optimism.

Q2 2026 · Foundation

Architecture and the first lift

May to July
  • Vendor selection: warehouse, ingestion, reverse ETL, agent runtime
  • Guest profile data model, written specification
  • Land Opera Cloud, Synxis, and dailypoint into the chosen warehouse
  • First daily-revenue-briefing prototype to the Revenue Manager
Q3 2026 · Revenue intelligence

Revenue Agent in production

August to October
  • Unified guest profile (guest_uuid) resolution layer in production
  • Revenue Agent in human-in-loop with one-click apply to Duetto
  • Reverse ETL: segments and tags pushed back to dailypoint and WhatsApp
  • Commercial scoreboard live with baseline metrics captured
Q4 2026 · Guest-facing

Front-of-house intelligence

November to January
  • Pre-Arrival Agent and WhatsApp co-pilot in human-in-loop
  • Voice Reservations pilot for after-hours capture
  • First agent graduates from human-in-loop to governed automation
  • Year-end ROI readout against the commercial scoreboard
Medium term · 1 to 3 years

What the platform becomes

By the end of year two, every revenue and guest decision flows through a unified data layer with a control plane built for the people making them. The stack moves from recommend-and-approve to governed-automation-within-policy for the patterns that have proven safe. By year three, [Client] has the architecture to absorb 10+ properties without re-platforming.

  • Year 1. Foundation, Revenue Agent, and guest-facing co-pilots in human-in-loop. Scoreboard live.
  • Year 2. Governed automation for proven patterns. B2B integrations (TA module, group scoring, homes pipeline). Second property onboarded without re-architecting the data layer.
  • Year 3. The platform scales from one property to ten or more on the same architecture. Astrodata’s role shifts from builder to strategic partner.

How we sequenced this. The quarters unlock each other on data readiness, not calendar. Architecture has to be settled before the foundation can land, the foundation has to be live before the Revenue Agent can run, and the Revenue Agent has to prove itself in human-in-loop before guest-facing agents inherit the same governance. The medium term assumes [Client] adds properties at the cadence stated in your 2026 technology strategy. If the cadence changes, we re-sequence.

07

Outcomes and ROI

Methodology, not perfection. Directional ranges to anchor the conversation.

Joshua asked the board-facing question directly: what does a successful engagement look like? Below, the focus areas from §03 paired with the KPIs we would watch and the directional impact ranges we would target. These are informed guesses based on patterns we have seen in adjacent industries. We will refine them against [Client]’s actual baselines in the first 30 days of the advisory engagement.

Focus area 01

Revenue management

KPIs · Direct booking lift · RevPAR uplift · Pace vs. control

Daily revenue briefings, governed rate moves, and segmentation pushed back into Duetto. The Revenue Agent compounds across thousands of micro-decisions per week.

Directional impact, 12 months

  • RevPAR uplift: [3 to 7%] vs. control weeks
  • Direct booking share: +[2 to 5] percentage points
  • Time to revenue decision: from days to hours
Focus area 02

Operational efficiency

KPIs · Headcount-to-revenue · Manual workflow hours · Reporting cycle time

Replace Zapier sprawl with consolidated orchestration. Replace manual report-pulling with self-service. The same operations team runs a materially larger business.

Directional impact, 12 months

  • Manual workflow hours saved: [___] per week across ops and revenue functions
  • Headcount-to-revenue: hold flat while scaling from 1 to [N] properties
  • Reporting cycle time: from days to near real time
Focus area 03

Guest experience

KPIs · Pre-arrival engagement · WhatsApp response time · NPS · Repeat booking rate

Pre-Arrival Agent, WhatsApp co-pilot, and Voice Reservations bring the guest into a personalized conversation before, during, and after the stay, without growing front-office headcount.

Directional impact, 12 months

  • Pre-arrival engagement: [X]% of guests engage with personalized outreach
  • WhatsApp response time: from hours to under a minute for first reply
  • Repeat booking rate: +[1 to 3] percentage points
Focus area 04

Marketing and attribution

KPIs · Return on ad spend · Cost per acquisition by channel · Advisor-attributed revenue

With unified guest data and proper attribution, ad spend can be targeted against lifetime value instead of last click. JB’s prior work leading data at Better is the playbook.

Directional impact, 12 months

  • Return on ad spend: +[10 to 25%] from better targeting and attribution
  • Cost per acquisition: down [10 to 20%] on the best-performing channels
  • Advisor-attributed revenue: measurable for the first time

How we got to these numbers. The KPIs are drawn from JB’s playbook leading data at Better and from the agentic analytics work in flight at Teladoc Health, Kyruus Health, and Decision Resources. The directional ranges are conservative midpoints from comparable deployments, weighted toward the lower end where [Client]’s data foundation is not yet in place. They are guesses today. They become specific [Client] targets in the first 30 days, against actual baselines.

How we validate. Each build engagement is gated on its measurable contribution to the scoreboard. ROI for the platform is computed against cumulative lift, not against any single project, and reviewed quarterly with the leadership team.

08

Why advise first

The rationale.
Reason 01

The next 90 days set five years of cost and capability

The technology decisions [Client] makes in the next 90 days will determine the cost and capability of the platform for the next five years. Picking the wrong warehouse, the wrong agent runtime, or the wrong reverse ETL approach is a six-figure mistake that compounds. The advisory relationship exists to make those decisions deliberately, with vendor-neutral analysis, before any code is written.

Reason 02

Build projects are faster and cheaper when informed by business context through partnership

A scoped, sequenced plan with named dependencies cuts the discovery overhead out of the build.

Reason 03

[Client] stays in control

With any build plans Astrodata delivers, [Client] owns a costed plan and can take it to any builder, including us. We earn the build work by delivering the advisory well.

Reason 04

AI accelerates delivery only with the right architecture underneath

LLMs are good at writing code but bad at creating system architecture. Define a plan and project roadmap into smaller, integrated components to enable AI-accelerated delivery of a cohesive and scalable system.

09

What we need from [Client]

To get started.
10

Timeline and next steps

The advisory partnership begins with a 10 weeks discovery and design sprint focused on the foundational deliverables: architecture recommendation, guest profile data model, agent roadmap, and commercial scoreboard. Build engagements are scoped and activated from there, informed by what we learn.
Week Milestone
0 Countersignature and kickoff
1 to 3 Discovery, system access, stakeholder interviews, on-property visit (optional)
4 to 6 Architecture options analysis, data model first draft
7 to 8 Agent roadmap, commercial scoreboard, on-property visit (optional)
9 First build engagement scoping
10 Leadership readout and advisory cadence established
Appendix A

Selected case studies

Case 01
Teladoc Health

The largest virtual care provider in the United States. Astrodata embedded with Teladoc Health’s data and analytics teams to deliver a modern data platform purpose-built to support agentic workflows and conversational analytics on top. Snowflake architecture, dbt modeling of clinical, member, and revenue domains, Fivetran ingestion, and Omni Analytics enablement, all designed so that natural-language interfaces and AI agents read from the same governed data the BI team uses.

Stack

Snowflake · dbt · Fivetran · Omni Analytics

This is the foundational precedent for [Client]. A unified profile (member at Teladoc Health, guest at [Client]), a curated semantic layer, and AI workloads in scope from the architecture forward.
Case 02
Kyruus Health

Connects 425,000 providers and 150 million health plan members. Astrodata partnered with Kyruus Health to build an agentic provider search experience. Members ask questions in natural language (“I need an endocrinologist who treats Type 1 diabetes, takes my plan, and has evening availability near me”) and the agent returns matched recommendations grounded in unified provider data. Provider, specialty, condition, plan, and availability data indexed in Elasticsearch as the retrieval layer; Google Gemini doing the reasoning; React for the member-facing experience.

Stack

Elasticsearch · React · Google Gemini

This is the closest precedent to the Pre-Arrival Agent and Voice Reservations agent. Take a natural-language request, reason over a unified profile, return a grounded answer the human can act on.
Case 03
Decision Resources

40+ years in business, 500+ manufacturing clients, top US Infor CloudSuite Industrial (SyteLine) ERP partner. Astrodata is building a conversational analytics experience that lets process managers explore SyteLine data, enrich it with context from web sources, and take action on forecasts and operational recommendations. Airbyte ingestion into Snowflake, dbt models for orders, inventory, production, supplier, and customer entities, custom agents reasoning over the modeled layer with grounded retrieval.

Stack

Snowflake · dbt · Airbyte · Snowflake Cortex · Omni Analytics · React

This is the operational analog of the Revenue Agent. A non-technical operator asks a natural-language question against a system of record, gets a grounded answer with external context, and the agent can take action within policy.
Appendix B

Leadership team

Principal Consultant
David Stocker
David is a technology leader who brings customer focus and team empathy to the forefront of his strategic and hands-on contributions. His deep background in product strategy, design and engineering allows him to get clients to value faster through hard earned wisdom acquired by leading projects at numerous Fortune 500, mid-tier and start-up companies.
Principal Consultant
Spencer Taylor
Spencer has an extensive background in embedded analytics, helping businesses turn their data into strategic assets. He has a track record of leading teams through data infrastructure transformations to support data application design and development. He has led data modernization initiatives at global logistics, technology, and consumer products companies among many others.
Principal Consultant
Johnathan Brooks (JB)
JB brings deep expertise designing data and AI systems built for both speed and scale, combining the governance required in healthcare and data-sensitive environments with the agility of high-growth startups. He helps modern data teams move from strategy to execution, including experience deploying enterprise-scale data warehouse, near real-time ML, and agentic AI solutions.