Every week there's a new headline about AI transforming some industry. Every week there's another vendor pitching an AI feature that will change how your business operates. And somewhere in the middle of all of it, a leadership team is trying to work out what any of it actually means for them.
We work with businesses that are genuinely trying to answer this question, and the most useful thing we can offer isn't an answer. It's a set of better questions.
Start with the problem, not the technology
The worst AI implementations we've seen start with the technology. A business decides it needs to “do something with AI,” finds a tool, and builds a use case around it. Sometimes this works. More often it produces something that's technically functional but commercially useless — a solution looking for a problem.
The better starting point is almost boring: where does your team spend time on things that shouldn't require a human? Where does information get lost between systems? Where do delays or inconsistencies hurt the customer experience? Where is a human doing pattern recognition on a large dataset that a machine could do faster and more accurately?
These are not AI questions. But they're what leads you to AI applications that are actually worth building.
The questions worth asking
Before any AI investment, we'd encourage a leadership team to work through a handful of questions honestly.
Is this problem actually solved by intelligence, or just by better data and process? A lot of what gets pitched as AI is really just automation or data quality work in disguise. Both are valuable. Neither requires a language model. Getting clear on the distinction saves significant time and money.
What data does this need to work? AI systems are only as useful as the information they can access. If the relevant data doesn't exist, is incomplete, or lives in five different places with no connection between them, the AI can't help you. Sometimes the prerequisite work — cleaning the data, connecting the systems — is the real project.
Who is this actually for, and what does good look like for them? The best AI features are built around a specific person doing a specific job. A care worker who needs a summary before a client visit. A sales rep who needs context before a call. A manager who needs a forecast before a board meeting. When you can describe the user and the moment clearly, you can evaluate whether the output is actually useful — and design it to be.
What are the failure modes, and how bad are they? AI systems get things wrong. The question isn't whether they'll fail but how they'll fail, and what the consequences are. In some contexts a wrong answer is an inconvenience. In others it's a liability. Understanding the failure modes before you build — not after — shapes how the system should be designed.
What's the build versus buy decision? There are excellent off-the-shelf AI tools for many common use cases. There are also cases where the value is in something specific to your business — your data, your process, your context — that a generic tool can't access. Knowing which situation you're in determines whether you're choosing a subscription or commissioning a build.
The honest version
AI will create real value for most businesses. But the businesses that benefit most won't be the ones that move fastest. They'll be the ones that think most clearly about where the value actually is, and build accordingly.
The hype is real. The opportunity is real. The work of figuring out where they intersect for your specific business — that's the part that takes honest thinking and usually benefits from an outside perspective.
If you're in that conversation, we're happy to be part of it.
