The skills you actually need to run an AI agency

The most common question from people considering this is "do I need to be technical?" It is the wrong question, asked in good faith, and answering it properly takes longer than yes or no.

The short version: you do not need to write code. You do need a set of skills that are harder to acquire than code, because nobody sells a weekend course in them.

The skill that decides everything: talking to small business owners

This is first because it is not close.

You will spend most of your working life in conversation with people who are busy, sceptical, mid-job and not interested in technology. A roofer taking your call from a van. A salon owner between clients with a colour developing. A driving instructor in a lay-by with fifteen minutes.

The specific skill is getting to the point without being abrupt, and listening long enough to hear what actually annoys them. Not what you assumed annoys them.

An example of the difference. You approach a garage assuming their problem is missed calls. Their actual problem is that people book MOTs and do not turn up, and each no-show is a wasted ramp. If you spend the meeting talking about call answering, you lose. If you hear the no-show complaint and say "we can send a reminder the day before and the morning of," you are now selling something they asked for.

You cannot do that if you arrived with a script. You can only do it if you asked a question and then shut up.

Signs you have this skill already: people tell you things. Signs you do not: you finish sentences, you explain features unprompted, and you find silence uncomfortable enough to fill.

The skill nobody mentions: being methodical about setup

Not technical. Methodical. There is a difference and it matters.

Configuring a voice agent for a client involves getting perhaps forty small things right. Opening hours including bank holidays. The areas they cover and the ones they do not. Whether they do emergencies. What they charge for a call-out, or whether they refuse to say. How to pronounce the business name. Which mobile the notification goes to. What happens if the caller asks something nobody anticipated.

None of these is difficult. All of them are easy to miss. And a system that gets thirty-nine right and tells callers the wrong opening hours is worse than no system, because now the client is angry and the mistake is in writing.

The skill is having a checklist and using it every time, including the twentieth time when you are sure you remember. People who are good at this are often people who have done a trade, run a kitchen, or worked somewhere with procedures. People who are bad at it are frequently the most technically capable in the room, because they trust their memory.

The skill that separates month three from month twelve: follow-up discipline

Most sales in this business do not happen in the first conversation. They happen in the fourth, six weeks later, after the owner has missed a run of calls on a busy week and remembered you.

Which means the actual determinant of whether you have an agency in a year is whether you keep a list and work it. Not enthusiasm. Not talent. A list.

Spoke to a landscaper in March, said he would think about it, ring him in May. That is the job. It is dull, it is entirely unglamorous, and the people who do it consistently do dramatically better than the people who send a brilliant proposal once and wait.

If you have never maintained a follow-up system in your life, be realistic that this is the skill you will need to build, and build it deliberately with a spreadsheet or a CRM rather than hoping you will remember.

The skill you need less of than you think: technical ability

Here is what the technical side actually involves for most people running this kind of agency:

• Writing and editing instructions for a voice agent in plain English - essentially writing a very careful briefing document for a new receptionist.
• Connecting accounts to each other by pasting keys between settings screens.
• Editing text and images on a website template.
• Building automations in a visual builder - this happens, then that happens.
• Pointing a domain at a host, which is three fields on a form.

None of that is programming. It is closer to being competent with software in general - the sort of person who sets up their own router rather than ringing someone. If you can manage a Facebook page, a bank app and a spreadsheet without help, you have enough.

What genuine technical ability buys you is speed and confidence when something breaks at an awkward moment. That is worth having. It is not the constraint.

The setup process is deliberately built to be followed rather than invented, which is the point - assembling a stack from scratch requires real technical depth, running one that already exists does not.

The skill that is genuinely hard to teach: judgement about who to walk away from

Some prospects will waste months of your life. Learning to spot them early is worth more than any sales technique.

Common patterns: the owner who wants endless changes before signing anything. The one who needs to "run it past" three people who never appear. The one who haggles the setup fee to nothing and will haggle everything else too. The one who is genuinely too small - no staff, no marketing, a handful of jobs a month - where nothing you sell will change anything.

Walking away from a possible sale feels wrong when you have three clients. It is nevertheless correct, and the sooner you can do it politely, the better your year goes.

The skill you will need in month two: saying "I do not know, I will find out"

Clients ask things you cannot answer. Can it transfer a call to my mobile if it is urgent? Can it book straight into my diary? Can it text them the quote?

Some of those are possible depending on the provider and how it is configured. Some are not. Some are possible in principle and awkward in practice.

The wrong answer is a confident yes. You will have built a promise you cannot keep, and the client will remember the promise more clearly than you do. The right answer is "I think so - let me check and confirm tomorrow", followed by actually checking and actually confirming.

This sounds like humility advice. It is really risk management. Every overclaim becomes an unpaid obligation. Knowing where the boundaries are before you are asked is the cheaper version.

The skill nobody enjoys: tolerating rejection at volume

You will be ignored. Repeatedly. By people who are not being rude, just busy.

The emotional skill is not confidence. It is indifference - the ability to treat a no as information rather than a verdict. The people who last at this are usually not the ones with the best pitch. They are the ones for whom being brushed off on Tuesday does not affect what they do on Wednesday.

If you know from experience that rejection lands heavily on you, that is genuinely worth weighing. It does not mean you cannot do this. It means the finding-clients part will cost you more than it costs someone else, and you should plan around that rather than discover it.

What you can buy and what you cannot

You can buy the stack. Voice agent, chat widget, automations, client sites, billing, the templates and the checklists - all of that is a purchasing decision, and what a complete setup contains is a fair thing to compare against the months it takes to build yourself.

You cannot buy the conversation with the roofer. You cannot buy the follow-up in week six. You cannot buy the judgement to walk away from a bad prospect or the discipline to check the opening hours twice.

That is the honest division. The technology is the solved part. The skills above are the unsolved part, they are yours to bring, and no system removes them. If that reads as discouraging, it is meant to read as accurate - this is a reasonable test of whether the shape fits you.