Finding your AI North Star

Why your strategy can't be 'Don't miss the boat.'

Mounting pressure

Your company may feel under pressure to “build something with AI” so they don’t fall behind in the race to efficiency. In the scramble to feel productive, organizational leaders look for the easiest fruit to squeeze, driven by the promise of juice.

Unless clearer heads prevail, this may lead to expensive chatbots that nobody wants to use, nonsensical vector databases, and teams that will be conditioned to bristle at the suggestion of working on AI projects. But the problem isn’t the technology. Companies are treating AI as a destination instead of a tool. You need a North Star: the problem you’re solving that justifies a project.

The scramble

Your executive leadership team – maybe you – just came back from an AI summit, and are positively charged up. The board has been asking why you’re not doing AI yet. Chatbots are springing up on your competitors’ websites, and industry publications are breathlessly reporting about how Acme Corp reduced their operational expenses by 80% after implementing an agentic observability tool. The golden goose has been set loose in your enterprise, and it’s fit to be chased. You’ve seen this play out so often it could be a sitcom trope.

Six months later, your crack product managers and engineers have designed a chatbot too, but it’s hidden in your UI because your confidence is shaky. Few people use it, and those who do complain that it’s actually made their experience worse. You’re still trying to find a proper consultant who can explain and drive agentic AI adoption, or a legitimate use case for it in your operation.

If your entire AI strategy is chasing that goose, you’re gonna have a bad time. You’re not just building solutions: you’re teaching your organization that panic is a valid planning methodology.

What’s actually broken

The issue isn’t the technology. Models work. Vector databases work. The core technology is ready. The problem is that companies are asking the wrong question: “how do we do AI?” instead of “what problem are we trying to solve?”

This is how you get:

• Chatbots that waste more time than they save
• RAG systems with fragmented sources returning irrelevant info
• Agentic solutions that technically work but solve nothing important

The biggest failure isn’t technical. It’s imaginative – and it is coming from the top. Your product managers are delivering what they were asked, and your engineers are sprinting. The failure is that leadership chases first-mover advantage without having the clarity moment that transforms AI from experiment to imperative.

Finding purpose in friction

Your AI North Star isn’t technology, it’s the answer to one question:

“What specific problem are we solving that we genuinely can’t solve any other way?”

“Our competitors are doing it” is not a problem. “We should probably have an AI strategy” isn’t a problem, either. This is reactionary behavior, and smells like panic.

The real answer is hiding in the friction. You know what I’m talking about, because you experience it personally and observe it every day. This is the stuff that makes your engineers curse and your support team apologize. Friction makes your workflows grind to halt. But friction can wait, right? That golden goose is getting away…

Getting your bearings

Back up, and pay attention to that friction.

Documentation hell

Your solutions architect spent 20 minutes grepping through a 500-page PDF for one command flag. They’re muttering “why can’t this just tell me the answer?”

There it is, the North Star. Not “we need better docs” - you’ve been saying that for five years. Instead, “what if documentation could answer questions instead of just existing?" or “what if the software could predict the flags?”

An AI assistant that lives in the IDE, or in the terminal, watching what users are trying to do. It surfaces the relevant snippet from your entire docs ecosystem before they even ask, or suggests the right flags as they type based on context. It’s not a chatbot or a search box with extra steps. It’s an actual answer, in context, right now.

Meeting overload

You just sat through your fourth status update meeting this week. Thirty minutes to learn what could’ve been three bullet points in Slack. You’re thinking, “someone on my team definitely already wrote this down somewhere. Why am I hearing it out loud?”

That’s your North Star. Not “meeting hygiene training” - this isn’t your team’s fault. No… the insight is “what if an agent could synthesize status updates from Slack, emails, and project tools into a brief you actually read before the meeting even gets scheduled?”

An agentic AI layer that watches your communication channels through MCP, identifies what you actually need to know, and provides a clean and properly contextualized executive summary. It doesn’t replace the one-on-ones that matter. It eliminates the update meetings that don’t. It’s not another dashboard, or meeting notes with extra steps. It’s the synthesis you’d get from a perfect chief of staff, automatically.

Click fatigue

A customer service agent navigates seven screens and clicks 23 times to process a simple return. They groan, “That was too many clicks. This should’ve been a phone call. This experience sucks.”

There’s your North Star. Not “streamline the return process” - that’s been on the backlog for ages. Rather, your insight might be “what if the system could watch what the agent is trying to do, and just do it?”

An intelligent agent that observes intent and automates the workflow. It analyzes fields, populates them contextually, progressively discloses only what’s actually needed. It reduces human error while turning a 5-minute task into 30 seconds. Not a workflow automation tool. Not a macro. Intelligence that understands the goal and handles the busywork.

Same trap, different customer

I’ve participated in meetings with companies that had impressive technology but struggled when I’d question “sure, but why you, specifically?” They’d pivoted to AI because the market demanded it. They could clearly articulate their novel technology stack, but when pressed on what made them better than alternatives, they simply couldn’t tell the story.

The problem wasn’t their engineering, it was their thinking. They were so focused on being in the AI conversation that they forgot to define what problem they uniquely solved.

It’s the same as the internal pattern I’ve been showcasing in earlier examples. Teams can, and will, build AI tools if their leadership demands it. That anyone articulates a clear problem in the market is not necessarily needed. And this may at times be a manifestation of a more serious concern. Maybe the problem is known to some, but the vision isn’t clear to those building or selling it. Solutions suffer when the people building them can’t empathize with the user.

Pick a vertical. Find a real use case where your technology does something competitors can’t. Get your teams on the same page. Build a demo that puts a face on the problem, solves it, and makes your target consumer react: “Oh shit, I need that." That is North Star work.

You don’t “do AI.” You solve a specific problem uniquely well, and using AI tools, you do it better.

When it actually works

I’ve also seen this done right. Companies that start by asking their own people, “where are the manual tasks eating your day? Where’s the friction that makes you want to throw your laptop out the window?”

They listen. And here’s the part that may surprise you: they often discover they don’t need to build anything new. The tools already exist. Off-the-shelf solutions, existing platforms, APIs that already do the hard part. The innovation isn’t in the technology stack, it’s in applying it to the specific problem that actually matters to their people. The results don’t take long to be transformative.

After a few months, they follow up with their people and check the data. Which solutions do people use? What is sitting idle? They kill the idle ones.

Whether you’re building internal tools or shipping products, the test is the same: if you can’t describe the point of your solution in one sentence – in a way that a non-technical person immediately understands the value – you probably don’t have a North Star. You may have a nothing burger dressed in too much ketchup.

Examples of good one-sentence pitches:

• “It reads your logs and tells you which service is failing.”
• “It knows all our documentation and gives correct answers to plain questions.”
• “It turns your meeting transcript into action items automatically.”

If your opening pitch requires a whiteboard or a slide deck, ten minutes of context and a room full of engineers to understand, you’re not solving a clear problem. You’re solving a complicated one that may best be left under the hood.

Your move

Following your competitors (or colleagues!) into AI is like navigating with their map. You’ll always be a step behind, heading toward a destination they chose.

The next game-changing idea isn’t in chasing someone else’s headline. That’s reactionary panic, not strategy. Your North Star is hidden in the frustrations your customers and employees have gotten so used to, they’ve stopped mentioning them out loud.

The move is to find those. The AI part is easy.