Start with the work, not the AI
If I was looking at how AI could help your business, I wouldn't start by giving you a list of tools.
I'd start by looking at how the business actually works.
What does your team spend time doing every week?
What's repetitive?
Where are people constantly copying information from one place to another?
What takes far longer than it should?
Where are mistakes happening?
What information is difficult to find?
Which jobs keep getting pushed down the list because nobody really wants to do them?
Those are much more interesting questions than:
“Which AI tools should we be using?”
Because once you've identified the friction, you have something useful to solve.
AI might be the answer.
Automation might be the answer.
A better process might be the answer.
Sometimes the answer might simply be getting rid of something nobody needs to be doing in the first place.
Automating a bad process doesn't make it a good one
This is probably one of the easiest mistakes to make.
You have an inefficient process, so you automate it.
Now you have an inefficient process that happens faster.
Before automating anything, I'd want to understand why the process exists and whether it still makes sense.
Businesses naturally accumulate processes over time.
Someone creates a spreadsheet.
Someone else starts producing a weekly report.
Another person adds a step because something went wrong once.
Five years later, everybody is following a process nobody would design that way today.
Putting AI on top doesn't necessarily fix it.
Sometimes the best improvement is simplifying the process first and then deciding which parts are worth automating.
Look for the repetitive work
Where AI becomes interesting is when it starts removing repetitive work that doesn't require somebody's full attention.
That could mean helping organise information.
Summarising large amounts of material.
Producing an initial piece of research.
Categorising enquiries.
Extracting information from documents.
Turning meeting notes into actions.
Preparing the first version of a recurring report.
Helping teams find information buried across different systems.
None of those examples involves replacing an entire department.
They're simply ways of reducing the amount of time good people spend doing work that doesn't make particularly good use of them.
Save somebody 20 minutes once and it doesn't matter much.
Save five people 20 minutes every working day and suddenly it's worth paying attention to.
Keep people where judgement matters
I'm much less interested in using AI to remove people from decisions where experience, judgement or relationships actually matter.
AI can help you research.
It can help you analyse.
It can help you organise.
It can give you a useful starting point.
But there are plenty of situations where I'd still want a person making the final decision.
Marketing is a good example.
AI can help analyse competitors, explore a market, organise keyword data, generate ideas and speed up parts of the content process.
That's useful.
But I'd be very reluctant to hand over the entire marketing strategy to an automated system and assume the answer it produces is right for the business.
The same applies to customer relationships.
Making your team faster is useful.
Making your customers feel like they're trapped in an automated system isn't.
Efficiency shouldn't come at the expense of common sense.
More tools can create less efficiency
There's also a slightly ridiculous side to the current AI boom.
A tool saves you time.
So you buy it.
Then another tool promises to save even more time.
So you buy that too.
Then you need another system to connect them.
Another dashboard to monitor them.
Another subscription.
Another login.
Another workflow somebody has to maintain.
Eventually, the technology that's supposed to simplify the business becomes another layer of complexity.
That's why I'd be quite ruthless about adding new tools.
What does it replace?
What does it improve?
How much time does it genuinely save?
Who is responsible for it?
Does it connect sensibly with what you already use?
If you can't answer those questions, perhaps you don't need it yet.
Small improvements can be more valuable than big AI projects
When people talk about introducing AI into a business, it can sound like a huge transformation project.
It doesn't have to be.
I'd often rather find three small processes we can improve this month than spend six months designing an enormous AI strategy nobody actually uses.
Take something repetitive.
Improve it.
Measure whether it helped.
Learn from it.
Then move onto the next thing.
Those improvements start adding up.
And because you're solving real problems rather than implementing technology for the sake of it, it's much easier to tell whether the investment was worthwhile.
AI should make the business simpler
For me, this is ultimately the test.
If you're introducing AI into the business, something should become better.
A task gets faster.
A repetitive job disappears.
Your team gets better information.
Customers get quicker answers.
Mistakes reduce.
A decision becomes easier.
People get more time to concentrate on work where they're actually valuable.
If none of those things is happening, it doesn't really matter how impressive the technology is.
You don't need an AI strategy because everybody else has one.
You need to understand where the business could work better.
Then use the right tools to improve it.
Sometimes that will involve AI.
Sometimes it won't.
The technology isn't the strategy. The improvement is.