Primer

The real cost of SaaS AI: the line item your vendor leaves off the deck.

Forty nine dollars per seat per month. Sounds reasonable. Multiply by your team in eighteen months. Now add the costs you are not seeing. The vendor's pricing slide shows one of four. Here are the other three.

The hook

Where this shows up.

A SaaS AI vendor publishes a pricing page. It says $49 per seat per month, $99 per seat per month for the premium tier. You multiply by your headcount and put a number in the budget. That number is wrong by a factor of two to four.

The other three quarters of the cost are real, predictable, and never on the pricing page. Here they are with a worked example.

The misconception

That the per seat price is the cost.

What most people believe

The per seat price is one component of total cost of ownership. The other three components show up at predictable points in the relationship and are larger than most teams budget for.

The vendor's incentive is to make the entry price look small and the exit price look invisible. Your job is to put the exit price on the page before you sign.

The better model

Four costs. Model all four before you sign.

What actually works

Total cost of ownership for a SaaS AI tool over a two year horizon decomposes into four pieces. Most pricing comparisons only model the first one.

01
Per seat cost (compounded by growth).
The visible cost. Multiplied by team size at month eighteen, not month one. A team that grows from 25 to 40 over that period pays for the larger team's worth of access for most of the contract.
02
Lock in cost (data export and migration time).
When you decide to switch, your data needs to come out. Some vendors charge. Most make the export incomplete or in a format that is hard to work with. Plan for engineering time even when there is no fee.
03
Switching cost (rebuilding workflows on a new tool).
Workflows you build inside a SaaS tool do not transfer. Templates, prompt libraries, integration points, training, all rebuild from scratch on the next tool. Estimate this number when you sign, not when you leave.
04
Customization debt (paying for what you do not use, missing what you need).
You are paying for the entire feature suite. The vendor optimizes for the median customer, not for you. If your needs diverge, you absorb the gap as ongoing workarounds.

Three decisions

Make these calls differently this week.

Always model the 24 month cost, not the launch month cost.

Worked example. A 25 person team starts at $49 per seat per month, total $14,700 per year. Grow to 40 seats and year two is $23,520. Add $5,000 in eventual export consulting. Add $30,000 to rebuild the workflows on the next tool. Total two year cost: roughly $58,000. The launch month number was $1,225.

Always test the export before you commit.

Ask the vendor to demonstrate a full export against a sample dataset. Look at the file format, the completeness, and the time to receive it. A tool that hides its export path is a tool that has decided you are not leaving.

Always price the build it yourself option, even if you decide against it.

The exercise of pricing the custom path forces clarity on what features you actually need. Often you will find that the SaaS tool is the right call, but at a tier two below the one the vendor pitched.

How ByteWorthy uses this

What this looks like in our work.

Custom AI you own. Not SaaS you rent. The thesis is not anti SaaS. SaaS is the right call for plenty of workflows, especially for small teams with non sensitive data and a short time horizon.

The thesis is anti undisclosed cost. When we work with a client, we put the twenty four month total cost of ownership on the page before any contract is signed. SaaS vs custom is one of the calls that math drives.

Read the build versus buy primer

Keep reading