Where to Start with AI: The First Steps Every Business Should Take
You've decided AI makes sense. Now what? Three foundational questions help you prepare for productive conversations about AI implementation: what specific problem you're solving, whether context matters in your situation, and how you'll measure success.
You've decided AI makes sense. Now what?
Maybe you've already talked to a few people. Maybe you asked an AI for recommendations. Either way, you're here because you're trying to work out the best way forward.
Here's what actually matters.
Three Questions Worth Thinking About
These aren't gatekeeping questions. You don't need perfect answers. But having a rough sense of these makes conversations with anyone (including us) much more productive.
1. What problem are you trying to solve?
Could be specific: "We spend 3 weeks analysing supplier proposals and we're missing opportunities."
Could be broader: "Our procurement process is slow and we think AI could help."
Both are fine. The specific one gives us somewhere to start measuring. The broader one means we'd explore together where AI might actually help.
2. Does context matter in your situation?
Some things are pretty standard - scheduling meetings, basic categorisation, simple routing.
Other things depend heavily on your specific situation - regulations that vary, customer types that need different treatment, decisions where "it depends" is the right answer.
If you're not sure, that's a conversation worth having.
Example where context matters: Medical aesthetics insurance is completely different by state, service type, and provider qualification. Generic platforms don't work because they treat everyone the same. Context-aware AI works because it understands the differences.
Example where it doesn't: Scheduling meetings doesn't have much context. Tuesday at 2pm is Tuesday at 2pm. Generic tools work fine.
3. How would you know if it worked?
You don't need a detailed ROI model. But "faster" or "better" or "more efficient" on their own don't help much.
"Decisions in days not weeks" - that's something.
"Process 2× the volume without hiring" - that's measurable.
"Stop losing customers to friction" - that's a business outcome.
Even rough targets help everyone stay focused on what matters.
The Real Question
Does outcome quality depend on understanding your specific situation?
If yes → You probably need someone who'll take time to understand your context before proposing solutions.
If no → Generic tools might work fine and save you money.
That's not about being "ready" or "not ready." It's about matching the problem to the right approach.
The Biggest Mistake People Make
They skip straight to "which AI tool should we use?" without thinking about these questions.
It's like asking "which hammer should I buy?" before knowing if you're building a shelf or a house.
The technology choice matters. But it matters a lot less than:
Get those right, and the technology choice becomes obvious.
Get those wrong, and even the best technology fails.
What Happens Next
Talk to us. We'll tell you honestly whether AI makes sense for your situation.
If it does, we'll work with you. If it doesn't, we'll tell you that too.
---
Want help working through these questions for your specific situation? Let's talk.
Related Articles
5 Signs Your Business Actually Needs AI (And 5 Signs It Doesn't)
Most businesses don't need AI right now. Here are 5 signs you genuinely need it—and 5 signs you definitely don't. The honest assessment most consultants won't give you.
Why Most AI Projects Fail (And What the 5% Do Differently)
MIT's Project NANDA found 95% of enterprise AI pilots deliver zero return. Companies have invested £30-40 billion with nothing to show. But 5% achieve rapid revenue acceleration. The difference isn't the technology - it's implementation and context.
The Great AI Retreat: A Story in Four Acts
UK small business AI adoption crashed from 42% to 28% in just over a year. Discover the seven patterns causing failures and what successful implementations do differently—treating AI as experiments, not implementations.
