From what I see, business owners are increasingly adept at using AI for more than drafting emails. Many are now using it agentically and are wondering how much further to take it. I have pushed AI into my workflow a long way, mostly into the parts of the job I enjoy least.
Here are some operating principles I have picked up along the way. It’s not an exhaustive list, and none of them is about which tool to use; they apply at whatever level you are.
1. Be patient. Correct consistently.
It is easy to get frustrated when an AI doesn’t give you what you wanted the first time, and just as easy to decide it doesn’t work.
When I first set one up to sort my inbox, the categories were wrong. Over a few rounds, I added one, found that two others did the same job, dropped one, and then asked whether the less important categories could be archived automatically while anything that needed action stayed unread in the inbox. None of that was the AI making mistakes. I was working out what I needed, and each answer was only as good as the question I had asked.
These models are very good at taking feedback, as long as the feedback is good. What makes the difference is where you put the correction. If you write it down somewhere the AI reads every time it starts work - a set of rules, or even a second brain (I’ll write more about that in a different article) - you stop making the same correction over and over. I correct far less now than I did at the start, and that’s because that second brain records my decisions and goals as they evolve, more than because the models have got better.
2. Type less, read more
AI saves you time and energy on typing. A first draft of a proposal now comes together from the meeting transcript and my notes, with AI enriching it with what it can find about the client and their sector, and I spend my time editing rather than writing.
That time saving pays dividends in more time to read.
Think of a newspaper editor, which may not even exist these days, but indulge me. They do not write the articles. Their job is to decide which stories run, read what the reporters file, and send a story back with direction when it is not strong enough. That is the role to adopt: be the editor.
I learned this by getting it wrong. A proposal for a client went to a colleague for review before I had read it properly myself. I had built up a lot of trust in the rules I had written for it and assumed it would be fine. The verdict came back that it read as cringe and a bit AI, and when I read it, I agreed. When I asked, the AI said it had not checked its own rules. Great.
I did not trust it any less afterwards. The lesson was that if something matters, and a proposal where tone of voice can cost you the work does, you have to read it before it goes out. Trust builds, the habit of checking slips, and that is exactly when you need to keep it.
3. Just because you can does not mean you should
Writing was the first skill AI made cheap for most people. Since late 2022, anyone can produce a page of competent copy in seconds, and most people can now spot it. Unedited AI writing has tells, and when a reader notices them, they assume no one put in any effort, which undercuts the message’s value. The same is now happening with software, images and video. Audiences are savvy and get used to it quickly, so whatever impact it has wears off.
The clearest evidence I have is from our own sales. Here is what has brought us work, best first:
- Events and meeting people face to face.
- Our website, a carefully worked proposition in a brand we care about.
- One-to-one approaches I make myself on LinkedIn.
- Automated email outreach.
That order is the reverse of how easily each one can be automated, and the reverse of how many people each one reaches. Automated email reaches hundreds, LinkedIn dozens, an event a handful, and the handful usually wins.
We built our website with AI, in that an AI wrote the code. People with a lot of experience directing every step created the brand, the layout, the proposition, and how it is put together. An agent could copy it now, but it could not have come up with it. I will stand by that statement.
AI doesn’t take the effort out of the work; it just shifts where the effort goes, to directing the tool to deliver better quality and results. Sometimes the right answer is not to use it at all.
4. Look to the mundane
The best value I have got from AI is in the jobs I like least: the inbox, sifting what comes in through the website, financial reports, forecasts, and keeping our records up to date. None of them needs much critical thinking, and all of them take a share of the week out of proportion to what they are worth. Setting one up is a small piece of work that pays back every day afterwards.
Someone I saw at an event recently had automated the jobs that used to eat his Saturday, and now he has the day back.
It is nearly always in the areas you least expect. If you are looking for somewhere to start, list the tasks you put off.
5. Learn by playing
Once I could build things quickly, I built a LOT: a couple of apps, prototypes of other people’s ideas, a first version of a site before it was needed, and our own website, which was worth it.
Not all of it needed building, but it was the best education I have had. I know far more about how these tools work, and where they fall down, than I did a few months ago, and understanding how things are built makes me better at selling them.
I have come to think curiosity matters more than technical ability. Try things without worrying about getting them wrong, because the mistakes show you what the right path looks like and what the wrong one costs.
6. Remember why you started
When you have a big hammer, everything looks like a nail. And it’s hard not to use that hammer. All that building cost me; I set a number of personal outreach messages each week and didn’t hit it every week. I went to fewer events than I meant to, and a conference we run did not get the attention it needed. It is fair to say the building pulled my focus away from the commercial tasks I had set out to do.
What pulled me back on course was a single tracker of what matters most and what to do next, kept by an AI assistant that brings it up at the start of every working session. I haven’t stuck to it perfectly, because I am far from perfect (discipline), but without it I would be far off course.
I need the help as much as anyone.
7. The pie is not finite
A common worry is that if the work gets faster, it gets ‘finished’ and there will be no more to do. I do not see it that way. No fixed number of problems waits to be solved, and faster tools mean we solve more in less time, leaving more time and resources to tackle problems that we never knew existed, or that were never worth the time, attention and cost before.
You can see it in this article. Nobody would ever pay someone to sort my inbox or keep our records tidy, and the prototypes in point 5 would never have been built. No one took that work from anyone; it didn’t exist until it became cost-effective enough to do it.
I believe in infinite pie.
If you have used AI a bit and would like to talk through where it could take the dull work off your plate, I offer a free 30-minute call with no obligation attached. Fill in the form on our contact page.
