The most important AI investment isn’t choosing the right tool. It’s building an organization that can keep learning as the tools inevitably change.

In today’s climate, it’s increasingly difficult to know who to trust.

On one hand, we have the technology providers themselves. Armed with millions of dollars in venture funding and their entire future business on the line, they are aggressively pushing the adoption of their specific services.

From startups like Replit and Lovable (who are rapidly growing their creative community growth teams) to giants like OpenAI and Anthropic, every company is investing heavily in education, community, and enterprise adoption, with Anthropic’s Claude Corps as the latest example of what a super-funded institutional adoption campaign looks like.

It’s the exact same playbook that Google Classroom used to capture distribution in nearly every school in America, wrapped in a narrative that echoes the structure of Teach for America (which, as we’ve learned, works well in some contexts, but really not well in others).

The problem? Any training designed by a single vendor inherently ignores the broader, rapidly shifting technological landscape. This blind spot is exactly why the market for independent, AI-first consultants is booming.

On the other hand, there’s a surging wave of boutique agencies, independent experts, and agentic middleware startups are flooding the market, positioning their AI know-how as the ultimate competitive moat. As a result, they are commanding massive premiums, with quotes ranging from tens of thousands for basic builds to millions for annual retainers.

Which leaves behind business leaders and their teams, who are completely overwhelmed. They are caught in a dizzying cycle of decision fatigue, trying to figure out which tool won’t be obsolete by next quarter without being taken for a ride from a consultant, all while their employees are still stuck using AI as a glorified Google search, rather than a tool to actually build things.

Every company wants to become your AI standard. Every consultant wants to become your AI expert. But neither is a substitute for building an organization that can learn.

Like it or not, every company in the world is becoming a school for learning AI.

What these companies are doing is pure provocation. They are injecting the market with a massive dose of FOMO to make us believe that a more perfect organization, a higher-performing team, or deeper market penetration is just one software purchase away.

But before you invest in your internal training (or tooling) options, you need to understand three critical realities.

Last year, every single company Build First worked with asked us teach AI using ChatGPT or Gemini. This year, it’s nearly all Claude. Next year? Maybe open source and locally hosted models will be in vogue. The fact is: Tools change. Don’t commit to anything forever.

This isn’t just happening with the large LLMs. Earlier stage VC-backed founder friends of mine tell me that their user retention has never been worse; people jump tools faster than ever. Given the novelty factor and the shifting social landscape of which tools are in favor by the broader population, it’s inherently risky to anchor yourself too deeply to anyone provider. Humans are too fickle, and tech move too fast. In that sense, the only way to build long-term technological agency is through resiliency of adaptability and tools. Learning how to learn a new tool is the meta exercise we all need to adopt.

There is a very popular idea when cloud services first came out for software developers called microservices. That’s to say, you didn’t have to store all of your data in one single hosting provider. Companies started using Amazon Web Services for some parts of their codebase and Google Cloud for others.

I’m noticing the same thing is starting to happen with AI tools. Some of the best companies that I’m seeing are ones that are actually looking end to end at the problems they’re trying to solve where the data lives today and picking the tool based on the problem, not the other way around.

Instead of forcing your team into a single, restrictive ecosystem, consider a different goal: Interoperability. Build a flexible stack where different AI models talk to each other to solve specific business needs. This approach also invites you to inherit the best practice of developing "sandboxed" environments with distinct security access and permissions. More importantly, it builds long-term technological resiliency for your organization. If a specific model falls out of favor, or a software contract renewal suddenly gets too expensive, you maintain your optionality and keep switching costs low.

Should we buy a new tool…or should we build it ourselves?

Engineers have been asking this question for ages, but only until recently have my non-engineering friends begun to ask of themselves. Like a lot of things, the conclusion is often, “It depends.”

At its core, this is a philosophical conversation that needs to happen at the leadership level and be largely informed by the makeup and capabilities of your current team. To make a long term investment in any enterprise grade tool means you must make some assumptions about what you think your employees are going to be doing.

I think we are under-estimating the massive paradigm shift that follows suit from this AI transformation. In this sense, it doesn’t matter where you start. It matters that you start teaching everyone how to stop being users of software and start thinking like makers of software.

The good news is that you can do this with any tool on the market today. You can build Copilot agents or a Gemini Gems or Custom GPTs or Claude projects or artifacts. The workflow is the same. From there, you’ll get a much better sense at who in your organization has the technological chops to think like an engineer and start asking them what problems they need to solve. This does not have to be a decision made from the top. It should be made informed by grassroots behavior of real things at your company.

The goal isn't to pick the perfect AI tool. It's to build an organization that can keep learning as the tools evolve. If you're working through that challenge, Build First would be happy to help. You can reach out directly at bethany@buildfirst.ai.