soo, things have escalated. welcome to the new site.
Last time I checked in I had just built a business in a day. What I didn’t mention was that the building didn’t stop there. The month since has been heads down, and some of what came out of it is in this post. Some of it is still taking shape. oliviakeiter.com is where all of it is going to live, personal, professional, the consulting direction I’m working toward, all of it in one place.
And here we are. First post on oliviakeiter.com. It feels right to start with the thing that pushed me from builder to… well. That’s what this post is about.
I walked into an interview cold. No background in the industry, no prior exposure to how the work gets done. Just a mini analysis they wanted me to run to show I understood their issue well enough to actually help and not just drop new tools no one will use. Evaluate four CRMs against their needs and recommend one, along with automation and AI use case recommendations for their firm.
So I did exactly what was asked and analyzed the four CRMs. Then I ran into a problem. The more I looked at the data the more obvious it became that their issue wasn’t “what CRM does what we need?” The issue was “none of these CRMs do what we need, and even stacked there are huge gaps.”
At that point I stopped evaluating and decided to build something instead. A system that worked on top of the SaaS products they already paid for, with a roadmap to phase that out with their own internal tooling over time. What I built to close that gap is really what this post is about. A working production system, built on real data, with Claude Code, from assignment to completion in 25 hours.
I called it Strata. The name is doing some work, the system is layered and the layers matter.
what was actually broken
This is an accounting firm that serves doctors and dentists. Their clients aren’t simple. A single client household might include a medical practice, a personal account, a spouse’s business, a trust, and a handful of other entities all managed under one relationship. The firm’s unit of work isn’t a contact. It’s a household.
Every CRM I evaluated modeled people as the primary object and tried to bolt relationships on top. That’s backwards for this kind of firm, and you can’t fix it with configuration. You either model it correctly from the start or every workflow you build on top of it spends the rest of its life fighting the data structure.
The other thing I kept seeing was knowledge living in people’s heads. Process, context, history, judgment calls, all of it stored in the team’s collective memory and a handful of spreadsheets nobody loves. That’s not a documentation problem. That’s what happens when no system was ever designed for how the firm actually works, so the only place the work can live is in the humans doing it.
what i built
The structural decisions that mattered most:
Client households as first-class objects. Not contacts with a household tag. Households are the parent object, entities live under them, individuals live under entities. The data model matches how the firm actually thinks about clients, which means every workflow built on top of it gets to be simple instead of clever.
A full agent fleet with a mandatory human approval queue. Agents draft, summarize, flag, route. Nothing goes out, nothing gets filed, nothing touches a client without a human approving it first. The queue isn’t a nice-to-have, it’s a structural requirement. Agents handle work that is fast and consistent and tireless. They are not doing work that requires judgment. That line has to be drawn in the architecture, not in policy.
🚨 If you work in a regulated, federal, or enterprise environment: a human approval queue is not optional. I am certain your compliance team has opinions about agents taking autonomous action on client work, and they are correct.
Designed to scale toward 2,000 households without proportional headcount growth. This was the actual business problem underneath the CRM question. The firm has a growth ceiling that isn’t about demand, it’s about how much manual work each new client adds. The system is built so that adding clients doesn’t add proportional human hours, because the agents absorb the routine work and the humans stay focused on the parts that require judgment. The point wasn’t to replace anyone. It was to make the existing team’s ceiling go up.
I am not a software engineer. I have ten years in technical operations and program leadership and I build with AI tools because they let me close the gap between seeing a problem and solving it. 25 hours from “this isn’t a CRM problem” to a working production system is not a thing I could have done two years ago. It’s barely a thing I could have done last year. The hard part was never the code. The hard part was seeing the problem clearly enough to know what to build.
the part i can’t stop thinking about
The interview was the spark. It’s not why I’m writing this post.
I’m writing this post because once I saw this gap I couldn’t unsee it. And the more I think about it the more I think this firm is not unusual. There is an entire category of niche businesses running on either spreadsheets or a duct-taped stack of three different SaaS products that don’t talk to each other, with hundreds of thousands of dollars a year in subscriptions, none of it branded to them, all of it sitting between them and their clients as a wall of logins and authentications and technical friction they never asked for.
They are not running on spreadsheets because they are behind. They are running on spreadsheets because the systems that exist were not built for them. Buying off the shelf means renting three tools that almost fit and paying a tax for the gap every single month. Building custom has historically meant six figures and a year and an engineering team they don’t have.
So the work lives in someone’s head. The growth ceiling stays where it is. The team works harder instead of working differently, because there has never been a real third option.
I think there is a third option now. I don’t think it’s a product. I think it’s the kind of work I just did, applied to more firms, by someone who has spent ten years watching this exact pattern play out in enterprise environments and now has the tools to actually do something about it at the small business level too.
what i actually want to do
I want to do AI enablement and consulting for the businesses sitting in this gap. The professional services firms, the specialty practices, the niche operators whose work doesn’t fit the shape of anything on the market and who have been quietly absorbing that cost for years.
Not as a pitch. I am genuinely not trying to sell anyone anything right now. What I want is to understand what is actually broken, because the way you build something that works is by talking to the people doing the work and listening for the parts that don’t.
So if you run a niche business, or you work inside one, or you know someone who does, and there is a part of the operation where the answer to “how does that work” is “well, [name] knows how to do that” or “we have a spreadsheet for that” or “we use a few different tools and a lot of copying and pasting,” I want to hear about it.
Tell me what’s broken. Tell me what nobody has built for you. Tell me where the work lives that shouldn’t have to live there.
That’s where I’m headed. Come with me. More soon.
find me on LinkedIn if the vibes feel right: linkedin.com/in/oliviakeiter