-By Geoff McDonald, CEO, Ambassador
This week Adam Turner at Postscript published something worth reading. His team rewrote an eight year old codebase from scratch, in Rust, with AI doing most of the writing. He says feature delivery is running 23.4 times faster than their previous platform, and he expects AI native rebuilds to become the normal path for legacy software.
I think he is right about the direction. I also think the post is missing the part that matters most, and I can say that without any smugness, because we made the same bet a year earlier and I have the receipts for what actually happens.
Ambassador 3.0 launched in October 2025. Same decision, same reasoning, same nerve required. We are now twelve months past it.
So here is the year that nobody posts about.
For the first half of 2025, before the rebuild shipped, our engineering team closed somewhere around 250 pieces of work a month. That was the steady state.
After the October 2025 launch, that number went down. Not for a week. For roughly five months.
January 2026 was the bottom. 184 items closed, the slowest month we had recorded in two years.
Nothing was wrong. That is the important part. Nobody was slacking, the architecture was not a mistake, and the decision was not wrong. It is just that a rebuild moves all of your accumulated knowledge from your codebase into your team's heads, and for a few months the team is the only place it lives. Every task carries a tax. Patterns that used to be obvious have to be re-decided. The tooling that made the old system fast has to be rebuilt for the new one.
You do not see this in launch posts, because launch posts get written in month one and the trough arrives in month three.
If you are a client who felt like things were quiet in that window, that is what you were feeling. I would rather tell you that plainly than pretend the line went up and to the right the whole way.
March 2026 is where it changed, and it has not stopped since. Every single month from March forward has shipped more than the month before it.
September, with five days still to go as I write this, is already the largest shipping month in the history of this company. More than four times what we closed in January.
I am not going to give you a single headline multiple, because I think those numbers are usually doing more marketing than measuring. What I will tell you is the shape, because the shape is the actual lesson.
The rebuild does not make you fast. It makes you compoundable.
What you buy with twelve hard months is a system where each new thing is cheaper to build than the last one, because the foundation stops fighting you. That compounding is the return. The speed is just what compounding looks like from the outside once it has been running long enough.
Adam is at month one. If the curve holds the way ours did, month three will be uncomfortable, and month twelve will be worth it.
We did not rebuild the platform because the code was old. We rebuilt it because what clients were asking us for had stopped matching what the product was shaped like.
Ambassador started as a referral and advocacy tool. A very good one. But a point solution, which means it did one job inside a customer relationship that has many jobs.
What we kept hearing, especially from larger clients, was some version of the same thing. The data is old. The data is siloed. It has no relevance to how the business actually runs today. And underneath that, a sharper point: nobody is trying to buy data management. They are trying to buy visibility into what is happening and outcomes they can defend in a budget meeting.
You cannot deliver that from a point solution. A point solution can tell you what happened inside its own box. It cannot tell you what a customer is worth across their whole life with you, because it never sees the whole life.
That is the actual reason for the rebuild, and it is why the product is now The Customer Lifecycle Operating System rather than a referral tool with more features bolted on. Acquisition, retention, and proof running on one record instead of three disconnected ones.
I want to be honest about where that sits. The architecture is built. Filling it out is the work of the next several years, and we are earlier in that than the positioning implies. We are further along than we were, and less far than we intend to be. Both of those are true.
Here is the concrete part, because a velocity curve that does not turn into things you can use is just a chart.
Custom domains for widgets. Your widgets now serve from your own domain rather than ours. Identical experience for the customer, but the third-party signature that ad blockers look for is gone. Referral clicks that were being silently dropped now get recorded. Nothing changed about what your program produced. What changed is whether you can prove it.
Health score v2 across every surface. Dashboard, reporting, and exports now calculate it the same way, so the number you quote in a meeting is the number the other person is looking at.
Expired balance in the Storefront report. Expiration date and expired balance are now shown directly, and available balance excludes expired cash instead of carrying it. Your number got smaller. It also became defensible.
Money that reports its real state. Remittance funding shows as settling until Stripe actually clears it. Advocates redeeming to a bank account see expected ACH timing on the confirmation.
Cleaner data going in. Contact import requires a customer status rather than defaulting everything to Enrolled. Concurrent identify calls no longer orphan device fingerprints. The Shopify integration got better observability and a cleaner setup flow.
Three things are built and in final review. Email suppression visibility, so you can see why a message stopped going out. A redesigned email editor. And auto-remediation for down landing pages, which repoints traffic to your site root and notifies you instead of letting clicks hit nothing.
In active build: attribution model as a company setting, so you choose first click or last click and the ledger recomputes against your choice. Reward status history with actor and timestamp. HubSpot rebuilt on our current integration framework. Stripe rewards enablement. Role based access control with hierarchy.
No dates on any of it. I am not going to hand you one I might miss.
The rebuild is the entry fee. It is not the achievement.
Plenty of companies are going to do what Adam did over the next eighteen months, and a number of them will post a chart in month one and go quiet by month four. The ones that matter will be the ones still publishing their curve at month twelve, trough included.
Speed is what everyone posts. The slope is what decides it.
Grow. Keep. Prove.
Ambassador is The Customer Lifecycle Operating System, orchestrated by HiroAI. Velocity figures are counts of work items closed on our own engineering board; September is a partial month. See how it works.