Most businesses we talk to start in the same place: “We need an AI strategy.” It’s the wrong first question. Nobody needed an electricity strategy. They needed a factory that ran longer and cost less, and electricity was how they got there.
AI is the same kind of shift. What it changes is the cost of routine thinking work: reading, sorting, drafting, checking. When that cost falls close to zero, some bottlenecks in your business simply disappear. The useful question is which ones.
What this looks like in a real business
Picture a manufacturer that sends a price update to every customer, every morning. Each customer has its own products, prices, packing and delivery terms, and they want it on different channels: some by email, some on WhatsApp. One person builds and sends all of it by hand, every day, and one wrong number goes straight to a customer.
Nobody there needs “AI”. They need that time back, and they need a wrong price never to go out. That’s the right brief. The AI part is small: read the price sheet, draft each customer’s message, flag anything that looks off. The real work is finding where each customer’s price actually lives, and making sure a person still presses send.
Four things separate the businesses that get value from the ones that get demos.
1. Treat AI as a capability, not a project
A chatbot on the homepage is a project. It launches, nobody owns it, and it quietly dies. A capability is different: it sits inside a workflow your people already run, and makes it faster or more accurate every day.
- Help your people before you try to replace them. The quickest returns come from giving the team that already does the work a tireless assistant: first drafts, summaries, checks.
- Fix the process, then add AI. AI on top of a broken process gets you the same mess, faster. Sometimes the biggest win is a cleaner spreadsheet and one clear owner.
2. Your edge is your data, not the model
Every business can rent the same models from Google, OpenAI or Anthropic. The model isn’t your advantage, because your competitor has it too.
- What they don’t have is your customer history, your pricing logic, the way you work and what your best people know.
- The blocker is rarely the technology. It’s data spread across inboxes, spreadsheets and people’s heads. Finding it and cleaning it up is usually the hardest part, and the most valuable.
3. Value this quarter beats perfection next year
Models improve every few months. Waiting for the perfect one means waiting forever.
- Start where “mostly right, checked by a person” already saves real time: routine emails, summaries of long documents, sorting incoming orders or support requests.
- Keep a person on anything that matters. For prices, money, hiring or health, AI prepares the options and a person decides. In the price example above, nothing reaches a customer until someone presses send.
4. Govern the new risks from day one
AI brings risks most businesses haven’t had to manage before.
- Data leaving the building. Staff paste customer data or contracts into public AI tools. Give them an approved tool and a short rule, not a ban they’ll work around.
- Confident mistakes. AI can state wrong things fluently. Anything customer-facing or regulated needs a check built into the workflow, not left to goodwill.
- Access. When AI agents connect to your systems, treat them like a new hire: only the access they need, and every action logged.
The hype trap, and where the value is
| The hype trap | Where the value is |
|---|---|
| “We need an AI strategy.” | “We need a business strategy, and AI is one of the tools.” |
| Buying tools because competitors have them | Buying tools that cut a cost or a delay you can measure |
| Expecting AI to fix a culture or process problem | Scaling a process that already works by hand |
| Using AI to cut headcount | Using AI to get more done with the team you have |
Where to start
Pick one workflow that runs every day, costs real hours and has a clear right answer. Map where its data lives. Put AI in the middle and a person at the end. Measure the hours saved. Then do the next one.
That’s how our AI readiness assessment works: one workflow, real numbers, and a plan you can act on whether or not we build it.