AI is useful when it removes friction from real work. It becomes expensive noise when the business starts chasing tools before it understands the process, the data, the risk and the outcome it wants.
01
The problem is not AI. It is distraction.
Most businesses do not need an AI strategy that starts with a big presentation and a long list of tools. They need a clearer view of where time is being wasted, where decisions are slow, where data is messy and where people are repeating work that software could help with.
AI can be genuinely useful. It can summarise, classify, search, draft, route, check, monitor and help people make sense of information faster. But if you start with the tool instead of the problem, you can burn a lot of time and money building something impressive that nobody trusts or uses.
That is the bit I care about. Not whether a business is using AI for the sake of sounding modern, but whether it is using it to make operations clearer, delivery faster and decisions better.
02
Start with boring workflows
The best AI opportunities are often boring. Support emails. Call notes. Document checks. CRM updates. Report summaries. Internal search. Lead qualification. Risk flags. Repetitive admin around finance, compliance, sales or delivery.
That might not sound as exciting as building a full AI product, but it is where the value usually appears first. If a team spends hours every week finding information, rewriting the same message, checking documents or moving data between systems, AI may be able to reduce that drag.
The question is simple: where is the business paying skilled people to do repetitive information work that could be made faster, safer or more consistent?
03
Do not automate what you do not understand
Before adding AI to a workflow, you need to understand the workflow without AI. Who does the work now? What information do they need? What decisions are they making? What happens when the answer is wrong? Who checks it? Where does the data come from?
If that process is already messy, AI will not magically fix it. It may just make the mess faster. You can end up with confident-looking outputs from weak data, unclear ownership and no proper way to know whether the system is helping or causing risk.
This is why technical leadership matters. Someone has to slow the business down just enough to map the process, understand the risk and decide where AI should assist, where it should not touch anything, and where a simpler system improvement would be better.
04
Use AI as support, not invisible authority
For most businesses, the safest starting point is AI-assisted work rather than AI-owned decisions. Let it draft, summarise, suggest, classify or highlight. Keep a person responsible for the final decision where risk, money, compliance or customer trust is involved.
That does not make AI less valuable. It makes it more usable. People are more likely to trust a system that helps them move faster while keeping them in control than a black box that suddenly starts making decisions nobody can explain.
A good AI workflow should make the human better informed, not remove accountability from the business.
05
Data quality is the real foundation
A lot of AI conversations skip over the unglamorous part: the quality of the data. If customer records are duplicated, documents are stored inconsistently, CRM fields are unreliable and decisions live in people's heads, AI has very little stable ground to work from.
Before investing heavily in AI, I would usually look at the data model, permissions, integrations, audit trails and how information moves through the company. That foundation decides whether AI can be useful or whether it will constantly need manual rescue.
Sometimes the most valuable AI project starts with cleaning up the systems around it.
06
Measure the outcome, not the novelty
AI work needs a practical measure. Did it save time? Did it reduce mistakes? Did it help the team respond faster? Did it improve customer experience? Did it make risk easier to see? Did it reduce dependency on one person knowing where everything is?
If the only success measure is that the business is now "using AI", the project is already drifting. The technology should be judged by whether it improves something that matters commercially.
That might be hours saved, faster turnaround, fewer manual checks, better reporting, cleaner handovers or more consistent customer communication. Pick the outcome before picking the tool.
07
A sensible way to start
I would start small. Pick one workflow with visible pain, decent data and a clear owner. Map how it works now, decide where AI could help, define the checks and build a controlled version that proves the value before rolling it wider.
Keep the scope tight. Avoid connecting everything to everything on day one. Make sure the output can be reviewed, logged and improved. If the first version saves time and people trust it, you have a foundation to build on.
That is how AI becomes part of the business properly: not as a shiny side project, but as a practical layer inside systems people already use.
Useful questions
Before spending money on AI, ask:
- What business problem are we actually trying to improve?
- Is the current workflow understood well enough to automate or assist?
- Is the data reliable, accessible and safe to use?
- What should AI suggest, and what must still be decided by a person?
- How will we measure whether this saves time, reduces risk or improves delivery?
