CRM, Automation & AI

Why "which is best?" is the wrong question

OpenAI's GPT models, Alibaba's Qwen models and Anthropic's Claude are all capable of answering customer questions, drafting copy and summarising documents. The differences that matter for a business show up only on your own content: your product names, your tone, your edge cases and your customers' spelling.

Rankings published online measure general ability. They rarely predict how a model handles a question like "pwede po ba magpa-reserve for Saturday?" against your actual booking rules.

How to test the models properly

• Collect 20 to 30 real questions or tasks, including awkward ones
• Write down what a good answer looks like for each
• Run every example through each model with the same instructions and your content
• Score accuracy, tone and whether it admits when it doesn't know
• Record the response time and the cost per answer

What to compare

• Accuracy on your content: does it stay grounded in your documents, or invent details?
• Language: how well it handles English, Filipino and Taglish if your customers mix them
• Speed: a chat widget needs answers in a couple of seconds; a nightly report does not
• Cost: priced per volume of text processed; small models are often a fraction of the cost of large ones
• Data terms: check each provider's current policy on whether API data is used for training

Avoid lock-in from day one

Whichever model wins your test today may not win next year. I build integrations with a provider-agnostic layer, so switching models is a configuration change rather than a rebuild, and a second model can take over automatically if the first one is down.

AI features start from ₱18,000, and an AI assistant grounded in your own content starts from ₱30,000 (US$1,200 internationally). Model usage is billed by the provider directly to your account. See AI integration for what's included.

A sensible way to start

Pick one narrow job with a clear measure, such as answering the ten most common enquiries or drafting quotation emails. Prove the value there, then expand. A small, well-tested assistant earns more trust than an ambitious one that occasionally makes things up.