The short answer: build in-house when you have sustained, year-round AI and automation work and can afford a $120K to $160K loaded hire. Bring in outside help when you need specific systems built well and soon. And know that these are not the only two options, because the best answer for most B2B companies under $50M is a mix: outside help to build, your team to own.
Here is how to make the call properly.
What each path actually costs
The in-house hire. A GTM engineer, automation specialist, or ops engineer runs roughly $100K to $140K base, $120K to $160K loaded, plus recruiting time and a ramp measured in months. Under 1% of companies in the $1M to $50M range have this seat filled today, which tells you two things: you are not behind your peers if you lack it, and the talent pool for it is thin, because everyone is fishing in the same small pond.
The consultant or agency. Fixed projects roughly $2,000 to $25,000 per system, retainers roughly $1,500 to $10,000 a month, hourly $100 to $300. Full numbers and the questions that expose a padded quote are in our cost guide.
The secret third path most companies actually take: nobody. The founder or a motivated ops person automates around the edges with off-the-shelf tools. This is fine, and it has a ceiling. It typically produces a handful of disconnected automations, no monitoring, and one person who is a single point of failure for all of it.
The four-question test
1. Is the work a project or a function? Count the systems you would build in the next 12 months. Three to five named systems is a project stream: outside help wins on speed and unit cost. A continuous, growing backlog across departments is a function: that justifies a seat.
2. Who will own it in month six?An automation nobody owns internally dies quietly: tools change, the workflow breaks, nobody notices until the pipeline does. If you hire in-house, ownership is solved by definition. If you bring in outside help, ownership must be designed in: a named person on your team, runbooks in your hands, training included. If a consultant’s proposal has no answer to “who on my team owns this when you leave,” that is the whole answer.
3. Is your foundation ready for either? This is the question both paths skip. AI automation built on dirty CRM data and undocumented process fails identically whether an employee or a consultant builds it. The debt you could tolerate for years now decides whether AI works at all. Whoever you choose, the first work is the same: fix the data, agree the definitions, then automate. An honest outside partner prices this in. A good hire spends their first quarter on it. Anyone who says you can skip it is selling you the failure mode.
4. Can you evaluate what you are buying? Hiring in-house means interviewing for a skill set you may not be able to assess. Hiring an agency means judging proposals you may not be able to compare. In both cases the fix is the same: pay for a diagnostic first. A paid audit with specific findings, a prioritized map, and fixed prices attached is cheaper than either a mis-hire or a mis-scoped project, and it converts both decisions from faith to evidence.
The honest comparison
Side by side
The honest comparison
| In-house hire | Consultant / agency | The mixed path | |
|---|---|---|---|
| Cash cost, year one | $120K to $160K loaded | $10K to $40K for 3 to 5 systems | Audit + builds, then your team owns |
| Speed to first system | Months (recruit + ramp) | Weeks | Weeks |
| Ownership in month six | Built in | Must be designed in; often is not | Designed in by definition |
| Breadth of experience | One person's history | Patterns from many clients | Both |
| Risk | Mis-hire in a thin talent pool | Dependency on the vendor | Smallest of the three, and it is the one most comparisons leave out |
In-house hire
- Cash cost, year one
- $120K to $160K loaded
- Speed to first system
- Months (recruit + ramp)
- Ownership in month six
- Built in
- Breadth of experience
- One person's history
- Risk
- Mis-hire in a thin talent pool
Consultant / agency
- Cash cost, year one
- $10K to $40K for 3 to 5 systems
- Speed to first system
- Weeks
- Ownership in month six
- Must be designed in; often is not
- Breadth of experience
- Patterns from many clients
- Risk
- Dependency on the vendor
The mixed path
- Cash cost, year one
- Audit + builds, then your team owns
- Speed to first system
- Weeks
- Ownership in month six
- Designed in by definition
- Breadth of experience
- Both
- Risk
- Smallest of the three, and it is the one most comparisons leave out
The mixed path is the fourth column most comparison pages do not have: outside help builds the systems at fixed prices, and every build ships with the ownership transfer built in: a named owner on your team, the runbook, the training. You get agency speed without agency dependency, and when the automation backlog eventually justifies a full-time seat, you hire into a working, documented system instead of a greenfield.
That column is our model, so weigh the source accordingly. It is also the model we would want as a buyer, which is why it is the one we sell.
When to definitely hire in-house
To be useful to the buyers it does not fit, the honest list:
- You are past roughly $50M or past the point where automation work spans every department continuously.
- You already employ someone with proven automation judgment; give that person the mandate before paying anyone outside.
- The work is core product, not operations. If AI IS the product, that is an engineering hire, never a consultant.