You don’t need the best model. You need the right problem.

Completely, 100% written by a human.

Whether your org is chasing the latest highly ranked models (anyone lamenting the current loss of Fable 5?) OR if your org is still watching AI from the sidelines, then I’ve got a reality check for you. It’s good news.

You don’t need the latest and greatest models to transform your organization with AI.

Particularly in the impact space, where so much fantastic well-intentioned work is bottlenecked by bureaucracy, data siloes, administrative tasks or other things that take an extraordinary amount of time…you have ample opportunity to make lasting and impactful changes, specific models aside.

The Problem with Model Chasing

By “model chasing”, I mean embracing the idea that your org must use the most highly ranked models or LLM vendors, in order for any implementation to be worth its weight.

The implicit thinking here is: the best models do the best work.

To a certain extent, the logic tracks. Model rankings, typically formed from a variety of expert perspectives, take into account many factors that we should all appreciate as users of AI. Think accuracy, factual correctness, mathematics, writing quality, speed, reliability, etc…

But when you’re too focused on going after the best of what’s available, the result is that you often lose sight of the other pieces that matter: like how to build consistency and scalability around your team’s AI use. And while flexibility is key for all of us working in AI right now, there’s a huge difference between augmenting as the landscape changes vs always looking for where the AI grass is greener.

That’s why it’s worth remembering three key points:

• Access to the highest-performing models won’t guarantee your org’s success. How you define your use case and ultimately deploy AI are the biggest determinants of your org’s ability to leverage its potential. Put another way, AI tech alone won’t save your org. It’s how people use it that drives transformation.
  
• Access to the highest-performing models isn’t necessary for your org to see gains from AI. This is your PSA that a top-knotch AI strategy, coupled with even a basic model from one of the “greats” (Anthropic, OpenAI, Google etc…) will take you much further than a half-baked approach reliant on the most eloquent model. Because again, it’s the brains behind the technology (the human ones!) driving true transformation here.
  
• This space is still new and changing. If you’ve chosen to go all in with AI, then for better or worse, you’re pretty much along for the ride with the rest of us! New models, features, terms and regulations are constantly dropping. And while some of us may find that exciting, operationally, it sounds exhausting to attempt to keep your org’s AI implementation in lock step.

Best case, model-chasing requires your org to expend additional effort where it isn’t needed. Worst case, it’s a distraction for where your time & energy should go.

Why Problem-Chasing is a Better Use of Your Time

If you close your browser with no other takeaways 🙏 – let it be this point that I raised in the last section:

How you define your use case and ultimately deploy AI are the biggest determinants of your org’s ability to leverage its potential

It’s the unsexy reality of AI implementation. When you chase AI with no plan – because it sounds cool, or because everyone else is doing it, or because it’s so powerful so it therefore must be impactful – then you are chasing tech for tech’s sake.

And historically, the gains with that approach are slim & slow. Chasing tech for tech’s sake usually looks like:

• Projects that start strong, but eventually lose momentum because the vision is no longer clear
  
• Project work that’s significant in load and feels productive, but ultimately amounts to busy work
  
• Projects that make it to completion, but then suffer from lack of adoption because the use case wasn’t properly scoped/the team won’t use it
  
• Heavy investment – time, money or even excitement – into tech that doesn’t yield an ROI (return on investment) or it’s unclear, because no one ever decided how to actually measure that return
  

Conversely, when you pinpoint specific areas that can benefit from an AI assist – be it with your org’s programming, fundraising, operations, or wherever else – you’re not limiting AI’s potential. You’re actually expanding what’s possible for your organization, in two key ways:

• You’re honing in on hot spots where there are real logjams. You’re not taking a broad stroke to your program management approach to see where AI sticks. You’re clear on which area(s) present a unique challenge to your staff and what the ideal scenario looks like. This not only streamlines your execution, but makes the universe of AI intervention far more accessible than it would otherwise be (because perhaps counterintuitively, nebulous possibility is just tactically limiting)
  
• You’re expanding the range of potential best-fit solutions. AI is a means, not the end. With a proper requirements discovery, you may find the best way to address that problem isn’t actually just by turning on Claude. Maybe it’s a series of prompts or AI “skills” that can be replicated across your org. Maybe it’s actually AI coupled with a set of integrations, or python code that can help write data where you need it to go. My point being: the clearer you are on your use case, the easier it is to grasp at the right possibilities

⚡️ The Action Item: How To Start Being an Effective Problem Chaser

This is probably the only context where I’ll tell you that it pays to chase your problems.

Discovering your org’s problems – and translating that to a technology use case – is part art, part science. Definitely way more than one blog post can effectively cover. But that doesn’t mean we won’t try.

At the end of the day, with any tech project, just remember that your org always wants to know:

• what it is you’re solving for, and ideally
• how you’ll know that it works

So before you get too deep in the AI weeds – or if you’re already there, hello – take a step back with each of your teams and carve space to ask:

1) What even IS our process right now for doing xyz? Are we aligned on what that looks like?

💡Tip: if your org is shaky around process clarity, map this out visually on paper/with post its/however you need. It helps if you all work from the same shared understanding of what your process looks like.

2) What are the biggest bottlenecks right now in our process? Why are they this way?

3) What would it look like to our team & constituents if those bottlenecks were reduced or eliminated?

4) (If your team has AI familiarity) How do we hope to see AI address this challenge? Where have we seen the potential for it to alleviate some of this pain?

5) Once this project is done – what should we look at to determine if AI helped us reach our desired outcomes?

p.s. If you have a hard time coming up with answers to these questions, that’s a pretty good indicator that your org may need to pump the brakes on your AI execution and spend just a biiiit more time on the planning.

Wrapping Up

The questions in that last section are big questions, for which you may not have straightforward answers.

But I promise, time spent addressing those questions – and what we’re trying to accomplish broadly with AI – is a far better use of our time than trying to keep pace with [insert-current-famed-model-here].

p.s. If you’re trying to get started with AI, check out my free e-guide – written for leaders who want to move forward with intention & a plan.