Stop Prompt Engineering. Start Context Engineering.

If you've been using AI for more than five minutes, you've probably heard the term "prompt engineering." It's the art of crafting the perfect question or instruction to get better AI outputs. And sure, it matters.

But here's what most people miss: the prompt is only part of the equation. If you're obsessing over how you phrase your question while ignoring everything else you're feeding the model, you're leaving 80% of AI's potential on the table.

That's where context engineering comes in—and why Dharmesh Shah's CART framework is one of the most practical mental models for getting better results from AI.

What Is CART?

CART stands for Context, Archive, Resources, and Tools. It's a framework Shah, Co-Founder and CTO of HubSpot) introduced at INBOUND to help people understand that generative AI isn't just about the prompt—it's about what you've situated the model within.

Here's the breakdown:

Context: The information you give the AI in the moment. This is your prompt, instructions, and any relevant prior data. Think of the "context window" as everything the AI can "see" at once. Current models can handle up to a million tokens (roughly three-quarters of a word each), which Shah joked is "most of the Cheesecake Factory menu." Better context = better output.

Archive: Past interactions, chat history, logs. This is basically what allows continuity and memory in the interaction. This makes AI feel less like a one-off question-and-answer machine and more like a collaborator that remembers what you've been working on.

Resources: Supporting data you feed the model. This includes any documents, spreadsheets, images, other external references that you share with the AI. The more relevant resources you provide, the more grounded and accurate the AI's output becomes.

Tools: The connections the model has. This includes things like APIs, databases, other software. This enables AI to act, not just answer. It's the difference between an AI that writes a draft email and one that actually sends it, schedules the meeting, and updates your CRM.

Shah's point is simple but powerful: investing in context engineering (via CART) produces better outcomes than just tinkering with prompts alone.

Why CART Matters in a Human-Centered AI Future

CART isn't just a productivity hack. It's a philosophy that aligns with building AI systems that are collaborative, accountable, and human-centered. Here's why it matters:

It keeps humans in control. When you formalize what context looks like (archive, resources, tools) you're not handing over decision-making to a black box. You're defining the inputs, which means you retain responsibility and oversight. AI becomes a true collaborator, not a magic trick you don't understand.

It enables repeatability and scalability. Once you've structured your context (what data, what tools, what history), you can build reliable workflows. You're not starting from scratch every time. You're building systems that get smarter and more efficient over time.

It preserves human-centric design. CART ensures that AI is grounded in human-provided data, human-relevant resources, and human-structured context. The AI doesn't replace your thinking, it enhances it. As Shah put it:

"Use AI to test your thinking, clarify your thinking, and elevate your thinking, but don't use it to replace your thinking."

It shifts the conversation from fear to possibility. Instead of asking "how do I compete with AI?" CART reframes the question as "how do I build with AI?" It's not about battling the machine. It's about building with the machine.

From Prompt Engineering to Context Engineering

Most people approach AI like they're asking a really smart intern a question. They type a prompt, hit enter, and hope for the best. That's prompt engineering.

Context engineering is different. It's about designing the environment the AI operates in. It's about asking:

• What data does this model need to see?

• What history should it remember?

• What resources should it pull from?

• What tools should it have access to?

When you shift your focus from crafting the perfect prompt to engineering the perfect context, you unlock AI's real potential. You're not just getting better answers, you're building better systems.

How HIVE Strategy Uses CART to Multiply Growth

At HIVE Strategy, we don't just use AI, we build AI-powered ecosystems. And CART is at the heart of how we do it.

Context: We design custom HubSpot workflows that feed AI the right context at the right time, whether that's customer data, campaign performance, or sales pipeline insights. The AI doesn't guess. It knows.

Archive: We build systems that remember. Past interactions, campaign history, customer behavior; all of it feeds into smarter, more personalized automation. Your HubSpot-centered ecosystem doesn't start from zero every time. It learns.

Resources: We integrate external data sources like spreadsheets, APIs, third-party platforms (Supered, Ask Elephant, and n8n being some of our favorites) so your AI has access to the full picture. Not just what's in HubSpot, but everything that matters to your business.

Tools: We connect HubSpot to any platform with an accessible API. That means your AI doesn't just analyze data, it acts on it. It updates records, triggers workflows, sends alerts, and drives real business outcomes. Siloed data is wasted data.

This is what we mean when we say we're not a marketing agency, we're a growth consultancy. We're not here to run campaigns. We're here to build the infrastructure that makes your entire revenue engine smarter, faster, and more effective.

The CART Mindset: Try It First with AI

Shah left INBOUND25 attendees with a simple rule of thumb:

"Try it first with AI. Don't overthink it. Just start."

He encouraged everyone to adopt a mindset of "this doesn't work, yet." Because what fails today might work brilliantly three months from now as the tech evolves.

His pro tip? Set calendar reminders to revisit and retry failed AI experiments every few months. What once felt impossible could become a no-brainer with just one model update down the road.

That's the CART mindset in action. It's not about perfection. It's about iteration, experimentation, and building systems that get better over time.

The Beautiful Paradox

Here's the thing about AI: it hasn't lived life. It's never tried to fold a fitted sheet, resolve a family feud, or negotiate with a sleep-deprived toddler (though I desperately need it to figure that one out and my budget to get that done is everything I own). It doesn't understand nuance, empathy, or culture in the way that we do.

That emotional intelligence along with judgment, humor, and lived experience is irreplaceable. Those are our greatest advantages.

So instead of fearing what AI will take from us, Shah urged us to imagine what it can give us back in time, energy, and focus. The stuff we actually want to spend our days on.

AI should replace the repetitive parts of our jobs so we can double down on the creative, strategic, and joyful parts. And the companies that figure this out soonest will be the ones leading the charge over the next decade.

Your AI Journey Starts with CART

You don't need to know everything. You just need to be willing to try, iterate, and build.

Start with CART:

• What context does your AI need to do its job well?

• What archive (history, memory) should it have access to?

• What resources (data, documents, references) should it pull from?

• What tools (APIs, integrations, actions) should it be able to use?

When you answer those questions, you're not just using AI, you're building with it. You're creating systems that multiply your capacity to grow.

And that's the future. Not artificial intelligence replacing us. Augmented intelligence amplifying us.

Ready to build AI-powered ecosystems that multiply growth? Let's talk about how HIVE Strategy can help you engineer context, not just prompts. Schedule time to chat with one of our experts today!