We use AI every day. It drafts, it sorts, it spots patterns across accounts faster than any human could. A lot of what used to take our team hours now takes minutes. So when we say AI is not the answer to everything, it is not coming from someone who is scared of it. It is coming from people who use it constantly and have watched where it helps and where it quietly costs you.
Lately we keep having the same conversation. Someone has decided AI can replace a thing it cannot replace, and they are about to spend real time and money finding that out the hard way.
Two recent ones stuck with us.
One business told us they could build their own custom system with AI and replace HubSpot entirely. Another held off on a proposal because, in their words, AI could probably do it.
Both are understandable. Both are missing the same thing.
The question is not "can AI do this"
AI can do a startling amount. That is exactly why "can it" is the wrong question. The better one is: should it, and what breaks if it does.
Take the idea of replacing a CRM with something AI builds for you. AI can absolutely generate the code for a contact database. That part is not the hard part. It never was. The hard part is everything around it. The data model that holds up when you have ten thousand records instead of ten. The integrations that keep talking to each other when one of them updates. Reliability when it is 9am and your sales team needs it to work. Support when it breaks. The year of maintenance nobody pictures when they are looking at a shiny demo.
HubSpot is not expensive because it stores contacts. It is the choice it is because someone else carries all of that weight, forever, so you do not have to. Replace it with a custom build and you have not removed the cost. You have moved it onto yourself, and you will feel it the first time something goes wrong and there is no one to call.
The proposal one is the same shape. When you hire us, you are not paying for the task. You are paying for the judgment about which task is even the right one. You are paying for someone to own the outcome and to be accountable when the numbers need to move. AI will happily produce a plan. It will not stake anything on whether the plan is correct, it will not notice the thing you did not think to ask, and it will not answer the phone when something needs to change. Confidence is not the same as being right, and AI is very good at sounding confident.
Where AI is the right call
We reach for it without hesitation when the work is:
- High volume and repeatable. Sorting search terms, tidying data, first passes across hundreds of rows.
- A first draft. Copy, outlines, structure you are going to shape anyway.
- Pattern spotting. Surfacing what is worth a human look across a lot of data.
- Speed on the known. Tasks where the right answer is clear and you just need it faster.
In all of these, AI is doing what it is genuinely best at. And here is the part people miss. This is not about doing less work. The time AI frees up does not get pocketed, it gets reinvested. We use it to go deeper on the accounts we run and to cover more ground than we ever could by hand. More analysis, more testing, more of the thinking that actually moves the numbers. AI takes the grind off the table so our people spend their hours on the work that needs a person. The client gets more for it, not less.
Where we tell people to slow down
We pump the brakes when the work involves:
- Judgment under ambiguity. When there is no clean right answer and the call depends on context AI does not have.
- Accountability. When someone needs to own the result, not just produce a draft of it.
- Strategy. Deciding what to do, in what order, and what to ignore. Not executing a decision already made.
- Trust and relationships. The things clients and customers stay for.
- Anything where being confidently wrong is expensive. The higher the cost of a quiet mistake, the more a human needs to be in the loop.
The pattern is simple once you see it. AI is excellent at doing the work. It is not built to decide whether the work is the right work, or to carry the consequences when it is not.
The line we draw
We are not the agency that will tell you to use AI for everything, and we are not the one that will tell you to avoid it. Both are lazy. The skill now, the thing that actually saves you money, is knowing the line.
Use AI to move faster on the things it is good at. Keep a person on the things where judgment, accountability, and trust are the whole point. Most of the time the strongest setup is not AI or a human. It is AI doing the heavy lifting and a person deciding what matters and standing behind it.
That is how we run our own agency. It is also why, when a client tells us they are about to replace something important with AI alone, we are honest with them even when it is not what they want to hear. We would rather lose the easy yes than watch someone spend a year learning a lesson we could have saved them in a single conversation.
If you are weighing up where AI fits in your own setup, that is a conversation worth having properly. It tends to save a lot more than it costs.