By Ariel Vallejos, Head of Cloud Solutions, Migration & Modernization at Strata Analytics Group
Whether your company runs on AWS, Google Cloud, Azure, Snowflake, Databricks or its own data center doesn’t matter. What matters is knowing where you’re going, taking firm steps, and never getting locked in along the way.
Almost every company is experimenting with artificial intelligence today. A pilot here, a chatbot there, a new tool every month. And a fair question comes up: what if we choose wrong? What if tomorrow we want to change cloud or vendor?
The good news is there’s a way to start without that fear. It comes down to four ideas: think big, start small, build on open standards, and never stop observing.
1. Think big: a 2-year roadmap
Before choosing technology, agree on the destination. We co-create a two-year roadmap that covers four mutually reinforcing fronts:
- Data: Trusted information, in one logical place, with one shared language across the company.
- Governance: Clear rules on who accesses what, with quality, security and traceability.
- AI agents: Assistants that work on that data and carry out real business tasks.
- Observability: Seeing at all times what each piece does, what it costs and what value it creates.
The roadmap is the compass: it lets you say “yes” to what adds value and “not yet” to what distracts. And it holds true on any cloud.
How is it built? With a short but thorough assessment
A good roadmap doesn’t come from a template: it comes from understanding the company. That’s why we start with an assessment that takes just a few weeks, focused on what really matters for making decisions. The goal isn’t to inventory every detail, but to get a clear, shared picture.
- The AS-IS, unfiltered: What data exists, where it lives, which processes depend on it, what hurts today and what it costs.
- Strengths: What the company already does well: teams, systems, know-how and data that are a real advantage.
- Context: Business goals, industry regulation, budget, available talent and level of urgency.
- What leaders are doing: How the companies leading change in the industry are using data and AI, and the results they’re getting.
With that picture, the roadmap is built in three layers, from the ground up:
- Foundations first. There’s a base every company needs, regardless of industry or cloud: trusted, well-organized data, governance and security, identity and access, and observability of costs and operations. It’s not up for debate: it gets done, and done right from the start.
- Quick wins on that foundation. Use cases prioritized by business impact and effort, showing value while the foundation takes shape.
- Growth bets. More advanced agents, data products and new revenue streams, arriving once the foundation and the organization are ready.
The result is a plan that’s easy to read, with clear priorities, owners and success metrics. A path everyone in the company understands and can trust.
2. Start small: quick wins people notice
Not everything is solved with AI
AI is a powerful tool, but it isn’t the answer to everything. Sometimes the greatest value lies in redesigning a process, cleaning up and organizing data, automating a simple task, or building a good dashboard everyone actually uses.
So the question isn’t “where do we put AI?” but “what does this process need to work better?”. We figure it out together, and AI comes in only where it adds real value to the process.
With the direction set, we start with what delivers visible results in weeks, not years. We look for cases that meet at least one of these conditions:
- Cut costs: fewer hours on manual work, fewer duplicate licenses, less idle infrastructure.
- Speed things up: reports that take days are ready in minutes; customers get instant answers.
- Bring order: one version of the numbers and shared business definitions.
- Reduce risk: regulatory compliance, controlled access and audit-ready from day one.
One highly effective quick win, on any cloud, is an AI-assisted architecture review. The major providers have best-practice frameworks (AWS Well-Architected, Google Cloud Architecture Framework, Azure Well-Architected Framework) and are starting to offer agents that review your environment automatically, prioritize findings by business impact and even estimate savings in dollars. In a few weeks you get a map of savings, security and resilience, and an expert team validates and executes the changes.
3. Freedom to move: open standards
The biggest risk of a data platform isn’t picking one cloud over another. It’s getting locked in, with your data in a format only that vendor understands. When that happens, migrating becomes expensive, slow and risky.
The fix is simple: build on open standards. That way your data and business logic travel with you to any cloud, or even back to your own data center.
| What you need | Open standard | What it gives you |
|---|---|---|
| Store data | Apache Parquet | A file format any tool can read; efficient and low-cost. |
| Organize tables | Apache Iceberg (or Delta Lake) | Reliable tables with history and versioning that Snowflake, Databricks, Spark or each cloud’s native engines can all use at the same time. |
| Process data | Apache Spark and SQL | The same code runs on any cloud or on your own servers. |
| Infrastructure | Infrastructure as code (Terraform / OpenTofu) and containers (Kubernetes) | Environments you can rebuild elsewhere in hours, not months. |
| Connect AI agents | MCP (Model Context Protocol) | Agents plug into systems the same way, regardless of model or vendor. |
| See and measure | OpenTelemetry and OpenLineage | Traces, metrics and data lineage in a common language any monitoring tool understands. |
On this foundation, switching clouds, adding a second one or going back on-premise stops being a problem. It becomes a business decision, not a rescue project. And the company keeps its negotiating power with vendors.
4. The platform, explained simply
Every modern data and AI platform, on any technology, follows the same path: the journey a piece of data takes from the moment it’s created until it becomes a decision.
① Sources: where information lives today
ERP, CRM, contact center, HR, databases, files, other clouds.
For the business: no need to replace what already works. We start from what you have.
② Ingestion and connectivity: bringing data in securely
Standard connectors, private connections and a single corporate identity.
For the business: security from day one and fewer custom integrations to maintain.
③ Data platform: bringing order and meaning
Data moves through stages (received, cleaned, validated, ready to use) in a lakehouse built on open formats. On top sit a semantic layer, which defines what each metric means, and an ontology, which describes how customers, products, contracts and processes relate. It’s the company’s business dictionary.
For the business: order and quality. Without this, AI answers fast… but wrong.
④ AI agents: intelligence at work
Agents with memory, permissions and continuous evaluation, for internal use (analysis, productivity, natural-language questions) and for customers (web, voice, WhatsApp).
For the business: productivity and a better customer experience, always with human validation.
⑤ Consumption: value reaches people
Dashboards, self-service, AI-powered analysis, portals and APIs for customers and partners.
For the business: faster decisions and more autonomy for every team.
⑥ Data products: share and monetize
Data packaged as products with an owner, quality standards and usage rules.
For the business: data stops being a cost and becomes new revenue.
⑦ Governance, security and operations: what runs through everything
Catalog, quality, identity, encryption, agent monitoring and cost control (FinOps).
For the business: compliance, lower risk and costs under control.
5. Speed, yes, but always with visibility
Moving fast without seeing is the most expensive way to make mistakes. Some AI projects have run several times over budget for months without anyone noticing. That’s why observability isn’t an extra: it’s what lets you accelerate with peace of mind.
Observing means being able to answer three questions at any moment: what is each agent doing?, what is it costing? and what value is it creating?
6. From SaaS to RaaS: paying for results
For years, the model was buying licenses (SaaS) or paying for consulting hours. Either way, the risk that value never arrives stays with the client.
We believe in a different model: Results as a Service (RaaS). Instead of selling tools or hours, we commit to measurable outcomes: hours saved, costs reduced, faster processes, business metrics that actually move.
- SaaS / hours: You pay for access or time. The outcome is up to you.
- RaaS: You pay for value delivered. Your partner shares the risk and the goal.
AI agents make this shift possible: they speed up execution, and our specialists validate and stand behind every deliverable. What gets measured is the result, not the effort.
Anyone can write a prompt. Auditing the result requires professional expertise.
7. At Strata, we know where to go. Let’s go together.
In uncertain times, relying on a partner with experience across industries (banking, telecom, healthcare, manufacturing, public sector) helps you avoid mistakes others have already made. We’ve worked across all the major clouds and platforms, so we can recommend what’s best for each case without tying you to any of them.
But experience alone isn’t enough. A good partner gets involved in understanding your business, stays through the whole journey and doesn’t disappear after the first delivery. And it shares what it knows: we believe sharing knowledge is the best way to grow, for your team and for us.
- We set the direction. Through a short, thorough assessment we understand your business and co-create the 2-year roadmap across data, governance, agents and observability.
- We deliver the first quick wins. Focused cases that cut costs or speed up processes, with clear metrics from the start.
- We scale on open standards. We add sources, semantics, ontology and new agents without locking you into any vendor.
- We transfer knowledge and measure results. Your team grows with us, and value is measured in business outcomes.
Trusted data + open standards + agents + people = results
No giant projects up front, no blind bets, no strings attached to anyone. Firm, safe steps on a clear path.
Take the first step with us.
Let’s talk about your priorities and find the first quick win for your company.
About the author
Ariel Vallejos is Head of Cloud Solutions, Migration & Modernization at Strata Analytics Group, a data analytics and artificial intelligence company that helps Latin American organizations develop accessible, context-aware, and culturally adaptable AI systems across the telco, finance, retail, and healthcare sectors.
