Operator-heritage vs agency: why we run our own SaaS
Most agencies that build software don’t run any. We’ve taken the opposite approach since 2003 — four production platforms with paying customers, on-call engineers, and the long tail of operational reality.
Most agencies that build software don’t run any. They ship a project, hand over the credentials, and the next time they look at the codebase is when something’s broken and the client’s furious. Whatever institutional knowledge existed lives in two engineers’ heads, and one of them left in March.
We’ve taken the opposite approach since 2003. Team Bison sits inside the Bison Grid Ltd group, and the group runs four production platforms — not as portfolio pieces or marketing fronts, but as live systems with paying customers, on-call engineers, and the long tail of operational reality that comes with both.
Those four platforms are:
- BisonGrid — vehicle logistics QC software in production for global shipping operators including Siem Car Carriers, TOTE Maritime, and the Wallenius Wilhelmsen partner network. Ships as web SaaS and as native iOS and Android apps (the Bison App on the Apple App Store and Google Play) used by inspectors and drivers in the field.
- BisonPress — an AI marketing automation WordPress plugin live with three external customers: Herd Group, New Team Services, and Siem.
- Bison Exchange — a production OEM EDI booking platform we’ve run for Siem Car Carriers in finished vehicle logistics for over ten years.
- Bison Insights — a modular operational insights platform, with Bison Track (Vessel Tracking Module) live at Siem, a Claude-powered AI chat agent in production, and a published seven-module roadmap.
Siem Car Carriers runs all four. One customer, four production systems, same engineering team. That’s not a portfolio. That’s an operating relationship.
This matters to buyers, but not always for the reason you’d expect.
We’ve made every mistake at least twice
The most useful thing about running your own software is that you make every architectural mistake, every deployment mistake, every “we’ll fix that later” mistake, every “the contractor said it was fine” mistake — and you’re the one who has to live with them. You can’t hand the codebase back to a client and walk away. You have to fix it, at midnight, with the customer on the phone.
By the time we quote on a comparable build for a customer, we’ve already paid the tuition. We know what breaks. We know what scales. We know what the cleanup costs look like three years on. An agency that hasn’t lived inside a long-running production system is genuinely guessing on those questions, however senior the team.
We’re calibrated on what reliability actually costs
The cheapest version of any software is the one that ships once. The most expensive version is the one that ships repeatedly, in production, while customers depend on it, for ten years.
That gap — between cheap-to-ship and cheap-to-run-for-a-decade — is where most agency builds quietly fail. You inherit a system whose original cost was £40k but whose annual maintenance cost is £35k because it was built without observability, without separation of environments, without sensible deployment pipelines, by a team that priced for the build and not the life of the thing.
We’re not better people than other agencies. We’re calibrated differently because we still own and run our oldest production systems. When we quote on a build, the number reflects what we know it will cost to keep that build running for the next five years.
We don’t need to bluff on AI
Half the agency market in 2026 is repositioning around AI. “AI-led”, “AI-first”, “AI-native” — pick a deck and you’ll find the language. Some of those agencies are very good. A lot of them are leading with positioning ahead of evidence.
We’ve got a Claude-powered AI chat agent running in production today against live operational data at a Tier 1 maritime customer. That’s not a roadmap item. It’s not a pilot. It’s not a marketing demo. It’s a deployed AI system in a regulated operating environment, with named tools, structured guardrails, and real users asking real questions about real vessels.
When we propose AI work to a customer, we’re proposing something we’ve already done at least once for ourselves. That changes the conversation in discovery.
What this isn’t
It isn’t a claim that running your own SaaS makes you a better partner by default. There are excellent agencies that have never shipped a multi-tenant product, and there are SaaS-running agencies whose services arm is mediocre. Operator-heritage is necessary but not sufficient for the kind of work we do. You still need the discovery to be right, the engineering to be honest, and the team you’d be working with on day one to be the team you actually get on day twenty.
What it does mean is this: every claim we make about how to build, run, and maintain operational software is checkable against four production systems with named customers. Most of our competitors don’t have that. The ones that do are usually three to five times our size, with the cost base that goes with it.
The line we use
We build like operators because we are operators. Twenty-three years of engineering history, four production platforms, with a global shipping operator running all four.
That’s the differentiator. Everything else is implementation.
Team Bison is the software, AI and operations consultancy within the Bison Grid Ltd group. We’ve been building production software since 2003. If operator-heritage matters to your evaluation, book a discovery call — we’d be happy to walk you through any of the four production platforms on a live or Siem-approved demo environment.