When your voice platform goes enterprise-only

Here is a thing that happens to software you depend on, and almost nobody plans for it.

On 15 June 2026 we screenshotted Synthflow’s pricing page as part of a routine sweep. It showed what it had always shown: a voice engine at $0.09 a minute, per-model AI charges, managed telephony at $0.02 a minute, and white-label at $2,000 a month. Stack those up for a realistic agent and you landed around $0.15 to $0.24 a minute. Ordinary usage pricing, the kind you can budget from without talking to anyone.

On 11 July we screenshotted the same URL. Every one of those numbers was gone. In their place, one card:

Contracts start at $30,000 annually.

One button. Contact sales. We have re-captured it on 30 July and again on 16 August, and it has not moved.

Nobody’s agent stopped working. That is what makes this worth writing about, because a platform that goes dark on you is an obvious problem and everyone already knows to fear it. A platform that quietly changes who it is for is the one that catches you, and it looks like nothing at all until you go to renew.

First, the part most coverage gets wrong

This is not a company in trouble.

It is tempting to read a deleted pricing page as a distress signal, and I want to be careful here, because that reading would be both unfair and useless to you. The public evidence points the other way.

On 23 June, inside the same window in which the prices vanished, Synthflow announced a partnership with AVANT, a distributor that sells through technology advisers into enterprise accounts. That is not a channel you build when you are winding down. It is a channel you build when you have decided your buyer is now a procurement department rather than a developer with a company card. Earlier in the year, Synthflow had already put SOC 2 and HIPAA compliance in place, announced on its own changelog. Compliance paperwork is expensive, slow, and bought almost exclusively by large organisations. Nobody does it to please solo builders.

Line those three up in order and the story is coherent: get the certifications, sign the distribution partner, then remove the self-serve pricing that was pulling in customers you no longer want to serve.

So the honest framing is not “Synthflow is failing”. It is “Synthflow has decided you are not the customer”, and if you are a small team, that is worse for you personally while being better for the company.

Now the maths, because “enterprise-only” does not tell you whether you are hurt

A $30,000 floor sounds enormous or trivial depending entirely on your volume, and you can work out which one it is for you in about ten seconds.

Take the old all-in range of $0.15 to $0.24 a minute. Divide $30,000 by each end:

So the break-even sits somewhere around 10,000 to 17,000 minutes a month. Under that, the new floor is a price rise for you, and the less you use the steeper it gets. A team running 2,000 minutes a month was paying roughly $300 to $480 a month before. The floor asks $2,500 a month. That is a five-fold increase for identical usage.

Over that break-even, the floor is not really the issue and you should be negotiating the per-minute rate inside the contract instead.

Worth saying plainly: the $30,000 is the starting figure, scoped around call volume, so a high-volume buyer pays more than the floor. It is a minimum, not a price.

What to do if you are on it right now

Work out your real monthly minutes before you talk to anyone. Not your plan, your actual usage. Our calculator will run your volume against every platform in the directory, including ones that still publish rates. Walk into a sales conversation knowing your own number or you will be negotiating blind against someone who does this daily.

Ask for the per-minute rate inside the contract, in writing. The floor is the headline. The rate is what you actually pay. A $30,000 commitment at a good rate can beat pay-as-you-go; the same commitment at a bad one is just a bigger bill with a longer lock-in.

Get your renewal date and your data-export path confirmed now, while you are still a customer they want to keep. Both are much harder to extract once you have said you are leaving.

Do not assume grandfathering, and do not assume it lasts. If you are on legacy pricing, ask directly how long it holds. We have seen no evidence either way, and an unanswered question here is itself an answer.

Price the switch honestly. Rebuilding an agent elsewhere is not free. Count the build time, the testing, the number porting and the risk of a worse first month. If you are near the break-even, staying may genuinely be cheaper than moving.

If you are moving, what you give up

Here is the uncomfortable bit, and I would rather say it than pretend the replacement is clean.

Synthflow still holds the highest ease-of-use score in our entire directory, a 10 out of 10, and we did not lower it when the pricing went. That would have been dishonest scoring: the drag-and-drop builder did not get worse, it got more expensive to reach. It is genuinely the easiest way to build a working voice agent that we have looked at.

The two realistic replacements you can still buy on a card both ask more of you:

Retell is the closer swap for most people: it publishes its rates, scores highest of the three on our universal weighting, and sits at a 7 on ease of use, so you lose some of the no-code comfort but not all of it. Vapi is the developer’s answer, more capable and considerably less forgiving, at a 5.

We keep a fuller list on alternatives to Synthflow. One thing to note before you click: our scores moved on 16 August, when we re-rated every platform in the directory in a single pass, so the numbers above are current as of that date rather than carried over from spring.

How to see the next one coming

Synthflow is the case study, not the moral. Every platform in this directory could do the same thing next quarter, and a few of them probably will. The tells are visible in advance if you know to watch for them.

Compliance certifications arriving before you asked for them. SOC 2 and HIPAA are enterprise procurement requirements. When a self-serve tool starts announcing them, it is telling you who it is courting.

A distribution or channel partnership. Resellers, technology advisers, marketplace listings aimed at large accounts. Self-serve companies sell direct. Enterprise companies sell through channels, and building one takes months, so the announcement is a lagging indicator of a decision already made.

The free tier or entry plan quietly shrinking. We watched exactly this pattern in the same fortnight covered by our latest changelog, where two platforms cut what you get for nothing while leaving their headline rates untouched. Trimming the bottom of the funnel usually comes before removing it.

“Talk to sales” replacing a number anywhere on the pricing page. Often one tier at a time, starting at the top. By the time it reaches the bottom tier, the decision is old.

None of these individually means anything. Two or three together, within a quarter, is a pattern.

The broader point, which is not really about Synthflow

There is a version of this article that treats the pricing change as a betrayal. I do not think that holds up. A company is allowed to decide which customers it wants, and the evidence says Synthflow made a deliberate, well-executed strategic move rather than an act of carelessness.

What I will criticise is the silence. We found this by re-photographing a page on a schedule. There was no announcement we could find, no changelog entry marking it, no notice on the pricing page acknowledging that anything had changed. Customers who were not watching found out when they went looking for a number that was no longer there. A dated line saying “we have moved to enterprise contracts, existing customers keep their rates until X” would have cost nothing and would have been the difference between a strategy change and an unpleasant surprise.

That is the actual lesson, and it is why this site takes dated screenshots of pricing pages instead of trusting what a vendor said last quarter. Prices are claims, and claims need a date on them. Our price index carries every current rate in the directory with the date we captured it, and where a vendor has stopped publishing, it says so rather than quoting a figure that has quietly expired.