Hire, Outsource, or Build?

Yesterday morning an offshore accounting firm reached out to me about a partnership. I want to tell you what they said, because their argument was a good one and it deserves a real answer rather than a dismissal.

Their pitch went roughly like this. AI removes a lot of friction from a firm. But as firms grow, they still need reliable people behind the workflows to actually absorb the additional client work. So their delivery teams and my agents could complement each other: the AI makes the firm operate smarter, their people provide the scalable execution capacity, and the firm grows without rebuilding its internal team every time volume increases.

That is not a silly argument. It is the standard one, and most of the profession believes some version of it.

But it points at something worth sitting with, because it describes a fork that every firm owner reaches eventually. You are at capacity. The work is there. You cannot take it on. What do you do?

There are three doors, and most owners only ever seriously consider two of them.

Door one: hire

The instinct. Someone leaves at five o'clock buried, so you post a job.

Hiring is the most expensive option and the slowest, and it is also sometimes exactly right. A good hire brings judgment, takes ownership, builds client relationships, and grows into more than the role you hired for. Nothing replaces that.

But be honest about the full cost. Salary is the smallest part. There is recruiting time, three to six months before they are genuinely productive, your own time spent supervising, the review burden while they learn, and the risk they leave in eighteen months and take the knowledge with them. And the capacity you bought is fixed. When the busy season ends, you are still paying for it.

Door two: outsource

Faster and cheaper per hour. A team is already trained, they can start next month, and you can scale them up or down with volume. For a lot of firms this genuinely works, and the good providers are very good.

But look closely at what you are actually buying, because it is the same thing you were buying in door one. You are buying hours. Somebody else's hours, at a better rate, with less management overhead. The work still gets done by a person, one item at a time, and you still review it. Your cost still rises roughly in line with your volume. You have improved the price per unit. You have not changed the shape of the equation.

There is a second thing worth naming. If you hand over a process that is a mess, you now pay somebody every month, forever, to perform that mess accurately. Outsourcing does not clean up a workflow. It puts more hands on it.

Door three: build

The one most owners never seriously consider, usually because they assume it requires a developer and a budget.

Building means creating agents that do the repetitive core of the work and hand it back to you finished, for your review. The categorizing. The reconciling. The monthly client report. The receipt chase. The same client question answered for the fortieth time this year.

Here is the difference that matters, and it is not a small one. Hiring and outsourcing both buy hours. Building removes the need for them.

An agent you build once runs on the eleventh client the same way it ran on the first. There is no recruiting, no ramp, no per-hour cost, and no ceiling that arrives at five o'clock. Your capacity stops being a function of how many people you can afford. That is not a better version of the same equation. It is a different equation.

In my own practice it is the reason I doubled my client load with the same staff. Not by grinding harder, not by hiring, not by cutting corners. I built the work once and let it run, and I have built over two hundred agents across what we do.

The honest part

Building is not free, and anyone who tells you otherwise is selling something.

The cost is not money and it is not technical skill. The cost is that you have to describe how the work actually gets done — clearly enough that something else could follow it. Most firms have never written that down. It lives in one person's head, and it has never been examined.

That is the real barrier, and it is also the real prize. Firms that go through it usually discover their process was more tangled than they thought, and they end up with something cleaner regardless of what the agent does. Nobody enjoys that week. Everybody is glad afterwards.

And no, you do not need to write code. If you can explain a task the way you would explain it to a new hire, you can build an agent to do it. That is genuinely the whole skill.

When the other two doors are right

I would not want to pretend building answers everything, because it does not.

Hire when the work needs judgment, ownership and a relationship. Hire when you need someone who can be wrong in an interesting way and tell you so. No agent does that.

Outsource when you have a genuine surge, when the work is truly commodity and already clean, or when you need capacity faster than you can build it. That is a legitimate use of somebody else's team.

But notice that both of those are decisions about the work that is left over. Which brings me to the actual point.

The order is what matters

The mistake is not choosing outsourcing. The mistake is choosing it first.

If you build first and staff second, you hire or outsource for a smaller, cleaner, higher-value remainder, and every dollar you spend on people goes toward work that actually needs a person. If you staff first and build later, you have locked in a permanent cost against work that never needed a human at all, and you will find it very hard to unwind. People are not a line item you cancel casually.

So the offshore firm and I probably do not disagree as much as it looks. Their model and mine can coexist. But the sequence is not neutral, and almost nobody talks about the sequence.

Build first. Then decide who you actually need.

Where to start, if this is your fork right now

Take the free agent, run it on one real piece of your work this week, and judge it yourself. It costs nothing and nothing renews: https://aiaccounting.legacysbc.com/claim-agent

If you would rather learn to build them properly, we run a free week of hands-on training. One hour a day, live at 3 PM CST, every session recorded, no card. The next weeks are September 7, September 14, and September 28: https://humansplusagents.ai/legacy-sbc-aia

And if you would rather just buy the ones already built, a single agent is $25, any five are $99, and a full Pro Bundle for your role or industry is $249: https://aiaccounting.legacysbc.com/pro-bundles

The free community is open to any practitioner, no cost, no pitch: https://aiaccounting.legacysbc.com/community

AI does the typing. You make the judgment calls. And you decide which door you walk through first.

— Yvonne Razo, Co-Founder, AI for Accounting