.png)
–
4 min read
Is your organization falling behind with AI?
"But wait," you might be thinking, "aren't we supposed to move fast and break things?" Not when it comes to AI.
The digital landscape is littered with the remnants of ambitious "moonshot" AI initiatives that promised the stars but delivered cosmic disappointment.
A recent PwC survey of 4,701 CEOs across 109 countries confirms what the successful tortoises already know: confidence in AI is sky-high (75% of Philippine-based CEOs trust AI integration compared to 67% globally), but the path to that confidence isn't through overnight transformation. 1
It's through something decidedly less sexy but infinitely more effective: measured, methodical implementation.
Reid Hoffman, the co-founder of LinkedIn and host of the Masters of Scale podcast, observes that the most effective AI strategies begin with a crawl-walk-run approach—starting with a clearly defined problem where AI can demonstrably improve outcomes, learning from that success, and methodically expanding.2
According to McKinsey & Company's research on AI adoption, AI high performers are achieving greater scale and see both higher revenue increases and greater cost decreases than other companies that use AI.3
The business case is clear: when implementing AI, patience and measurement, not speed and scale, deliver the most substantial results.
If you've already begun evaluating AI use cases and establishing metrics, congratulations—you're on the right track.
The most successful organizations aren't those with the biggest AI budgets or the flashiest technology; they're the ones methodically building capabilities based on measured outcomes.
By focusing on incremental progress and rigorous evaluation, you're positioning your organization to extract real value from AI rather than chasing technological mirages.
Picking the right initial AI project is the most critical decision in your implementation journey.
Its success will create momentum or stall your AI transformation before it begins.
Rather than chasing flashy applications, focus on these strategic considerations:
Customer service inquiries, internal knowledge base searches, and standard document processing are consistent starting points for AI implementation, as they combine natural language understanding with concrete, measurable outcomes.
Vague aspirations like "improved experience" aren't sufficient—you need quantifiable metrics such as time saved, error reduction rates, or satisfaction scores that demonstrate concrete value.
Your first project should have clear boundaries and reasonable implementation timelines measured in weeks rather than months.
When a process spans multiple departments or requires integration with numerous systems, it's likely too complex for an initial effort.
AI needs quality information to learn from, so prioritize areas where you already have structured data about how the process works—historical customer interactions, documented procedures, or organized datasets that can train your models effectively.
Your first AI project should be meaningful enough that success will be recognized across the organization, but not so mission-critical that any hiccups would cause significant business disruption.
A well-chosen first project builds confidence, demonstrates value, and creates momentum for future AI initiatives.
Starting too big risks expensive failure that sets your AI efforts back by months if not years.
At Swarm, we've developed a maturity framework tailored to the Filipino business environment that guides organizations through progressive stages of AI implementation:
The foundation layer: Basic prompting with fundamental infrastructure setup to support usage (API accounts, authentication, and basic integration).
Philippine business example: Cebu Pacific has deployed a generative AI customer support agent in partnership with Ada Support Inc.
The AI tool handles flight bookings, itinerary changes, and travel documentation with 24/7 real-time support, marking an initial step in the airline's broader AI strategy.4
The enhancement layer: Makes extensive use of prompt engineering techniques grounded in business understanding.
This includes few-shot prompting, chain-of-thought reasoning, and tailored instructions that align with specific business processes.
Philippine business example
UnionBank is integrating AI and machine learning technologies, including large language models, to drive operational improvements and business growth.
The bank sees these technologies as strategic tools to enhance its competitive edge and customer experience.5
The customization layer: Leverages your organization's proprietary data to find answers and solve problems. This stage connects LLMs to your internal knowledge bases, documents, and databases.
Philippine business example
Anycase.ai is an AI-powered legal research tool for lawyers and law students. The platform reduces the time it takes to complete accurate legal research and draft legal documents by 75%.
Using Retrieval-Augmented Generation (RAG) built on top of a legal database of 90,000+ local case laws, statutes and government rules and regulations, Anycase.ai lets users conduct their research in natural language, and provides relevant AI-generated answers, summaries and analyses complete with legal citations.
The orchestration layer: Creates sequences of prompts and steps to achieve complex goals. These workflows connect multiple AI capabilities to handle multi-step processes.
Philippine business example
Unlike rigid chatbots that rely on static scripts, ChatGenie’s Execution Graph approach ensures seamless, intelligent conversations. Multiple AI agents collaborate—analyzing intent, filtering inquiries, and refining responses—to deliver fast, accurate, and secure interactions.
This means fewer chatbot errors and a better customer experience. Whether in Filipino, English, or local dialects, ChatGenie adapts to real customer inquiries, making it a reliable AI assistant that helps businesses engage and convert more effectively.
The transformation layer: Independent drivers of business operations that understand business and customer goals, and plan actions accordingly. These agents can innovate new approaches when existing business processes fall short.
Philippine business example: Expedock's Freya is an AI agent for ocean freight that automates complex logistics tasks, including email management, documentation, and coordination with service partners.
As Expedock's Cofounder and CPO, Jig Young emphasizes the importance of deploying AI where it delivers immediate ROI with minimal disruption.
“We deployed our AI agent to our customers’ Outlook email exchange so we could impact their business operations immediately with little change management. Instead of changing their workflow, we supercharged it.”
Their first AI use case—classification of 1,000+ emails per day per freight operator—delivered enough value to build trust, paving the way for automating responses and actioning tasks. This iterative approach helped end users see what’s possible, allowing Expedock to scale automation where it drives the highest impact.
Expedock’s Freya is a true autonomous AI agent, actively managing freight operations—not just responding to tasks, but coordinating, optimizing, and executing workflows with minimal human input. Trained on millions of freight documents and thousands of business rules, Freya triages emails, tracks shipments, extracts key data, handles invoicing, and resolves logistics exceptions in real time.
Unlike basic automation, Freya remembers business rules, adapts to changes, and collaborates with human Freight Operators to ensure 100% accuracy. As it gains access to new automation tools, it continuously improves, making freight logistics faster, smarter, and more cost-efficient.
Trained on 5 years of ocean forwarding data, Freya can handle processes across freight milestones while working alongside human operators to ensure accuracy.6
Even with a measured approach, organizations face several predictable hurdles when implementing AI:
Rather than attempting wholesale replacement, successful organizations create middleware layers that allow AI to interface with existing infrastructure while preserving critical business logic.
Address this by involving key stakeholders early, focusing initial applications on removing frustrating tasks rather than replacing roles, and celebrating wins that demonstrate how AI amplifies human capabilities.
Start with targeted data cleanup in areas supporting your first AI use case rather than attempting enterprise-wide data transformation.
Consider a hybrid approach: partner with specialized providers like Swarm for implementation while developing internal capabilities through targeted hiring and upskilling programs.
Different business needs require different agent architectures. Several effective AI agent design patterns support a measured approach:
This works well for processes with clear, fixed stages—like generating content in one step and checking it against guidelines in another.
Financial services companies use this approach to sort customer inquiries, sending simple account questions to basic agents while routing complex investment inquiries to specialized systems or human advisors.
This approach allows for quality control and continuous improvement without complex implementation.
The architecture should match both your business requirements and risk tolerance.
For critical functions, maintain more human oversight. For lower-risk areas, more autonomous approaches may be appropriate.
The cornerstone of the measured approach is rigorous evaluation.
In their analysis of AI-powered organizations, HBR notes that leaders need to clarify the business case for AI systems and carefully track both quantitative measures, such as reduced costs and increased revenues, and qualitative ones, such as improved stakeholder experiences.7
After demonstrating success in your initial implementation, you should consider expanding to additional use cases or enhancing capabilities.
This disciplined expansion ensures that your organization builds on proven success rather than continuously chasing new possibilities.
Even the most measured approach requires appropriate governance.
Start with frameworks that can evolve as your agent capabilities expand:
These governance elements should grow in sophistication as your agent capabilities expand, ensuring that increased autonomy is matched with appropriate oversight.
A measured approach also involves gradually building understanding and capability throughout your organization.
When business leaders across functions understand AI capabilities and limitations, they identify more valuable use cases and set more realistic expectations.
This literacy-building effort should track with your implementation timeline, ensuring that the right stakeholders have the right knowledge at each stage of expansion.
The greatest paradox in AI implementation is that slowing down actually gets you there faster.
The companies making headlines with their AI successes aren't the ones who attempted to transform overnight—they're the ones who methodically built foundations, measured what matters, and expanded with purpose.
Before launching your AI initiative, assess where your company stands:
Which repetitive, data-heavy tasks consume disproportionate resources?
Do you have the structured information needed to train effective models?
Where are your AI knowledge gaps, and how will you address them?
Choose a bounded project with clear metrics that demonstrates value.
Product thinking principles should guide your AI deployment decisions.
As Marty Cagan notes in his Silicon Valley Product Group article on AI Product Management: "The real point is that generative AI is a technology, not a product. And a product—even an AI-powered product—requires us to solve for the users, solve for the business, and solve for the broader ecosystem impacts."
This means identifying specific Philippine market needs that AI can address, not just implementing AI simply because it's available.8
As you begin your AI journey, remember that you're not just implementing technology—you're cultivating organizational capability.
And in the end, they don't just finish the race; they change how the race is run.
If you need expertise to scope out AI use cases for your proof of concept, or need help defining an AI strategy — it's time to talk to Swarm.
Endnotes
