AI is everywhere right now, and most of what we see are demos, such as a chatbot that summarizes a meeting, drafts an email, or even suggests code in seconds. Those are useful, but they’re only one part of the story. What leaders really want to know is how these large language models (LLMs) transform plain-English requests into actionable work that actually moves within their systems, CRMs, ERPs, automation steps, and data platforms. As Omar del Rio, Chief Strategy Officer, at Definity, puts it: ´Generative AI is only a piece of the puzzle.´
Generative AI (LLMs like GPT, Claude, Gemini) is the foundation. But, the real business impact becomes apparent when we connect that base to tools that actually do things- the systems that fetch, file, move, and make.
LLMs are great at turning messy ideas into clear language, such as summarizing a meeting, drafting an email, restructuring a plan, or proposing a first pass at code. Think of them as a language interface for intent, they help us say what we want in plain words.
We need more than language alone to ship outcomes, though. An LLM won't update your CRM, reconcile an invoice, or create a ticket...unless it can embed in the right tool, with the right parameters, and the right permissions. That's where the value in simple AI tools changes completely.
By tools, we are referring to APIs, business applications, data platforms, RPA/automation steps, code runners, and MCP servers. You can embed an LLM into those tools (safely), and your ´chatbot´ turns into an operator.
´Most people struggle to understand the importance of language in translating user intent into actions.´ - Omar del Rio, CSO at Definity.
The language captures what we mean, and the tools execute it. The LLMs are the language brain, and the tools are the hands. Business value occurs when these two elements work together in harmony with security.
Practical business patterns you can pilot now:
A bigger model does not mean a bigger impact, there is a recurring narrative that the next model will just ´solve it all´. Helpful improvements do happen, but bigger models alone won't transform all your operations - integration will. As Omar mentions, ´AGI will not come directly from just iterating over what is a token predictor with a simulated chain of thought process.´ LLMs predict words, they simulate reasoning well enough to be useful, but the business step that becomes the differentiator comes from tooling expertise, safe execution, reliable handoffs, and feedback loops tied to your data and policies.
At Definity, this is the approach we are looking into. We are building for the intent-to-action layer that embeds LLMs to the tools our clients already use. This means we are implementing guardrails by design, operational visibility, and measurable outcomes.
As Omar says, ´We are working toward completing the puzzle with tools that help you connect the impressive language capabilities of LLMs with the actual work that needs to be done.´
If you're exploring this space, here's a way you can begin:
Pick one workflow that’s frequent and measurable (e.g., hot-lead follow-ups, ticket triage, invoice matching).
Establish success criteria and guardrails (target metric, data scope/permissions, audit logging, human-in-the-loop).
Map 3–5 intents into one system (e.g., “follow up on high-probability leads”, then CRM actions).
Run in two phases: shadow mode (where AI proposes), followed by action mode for low-risk steps.
Measure & expand: compare baseline vs. pilot, add intents once quality is stable.
How Definity helps: we design the guardrails, connect LLMs to your systems of record, and stand up a secure, measurable pilot.
Generative AI started the conversation, and tools finish it. AI, with the right implementation, can understand what you mean and take action accordingly.
Let Definity amplify what you can do by giving AI the ability to take meaningful, secure, and measurable actions.