The incumbent broker was already in production carrying integration traffic, owned by an application group with its own development server, release process and support queue.
The question an architecture function has to ask about that is not whether it works for the integrations already built. It is whether it is the right platform for the integration the organization has not done yet.
2014: the evaluation that was won on price and lost on architecture
A structured proof of value ran against the incumbent in 2014, and the comparison came down to cost — at which point the incumbent manufacturer moved to protect its installed base by offering the City upgrades at no charge. That removed the cost case, and the evaluation went quiet.
A vendor defending an incumbent position can always discount faster than a challenger can justify a migration. So a platform comparison reduced to the next three years of licensing will not be won by the platform that is architecturally right; it will be won by whoever has the most to lose. What the City got was free upgrades. What it did not get was the platform the evaluation pointed to — and the question was held open rather than closed as settled.
How that evaluation was run
As a proof of value against the City’s own interfaces, not as a demonstration. Written use cases came back from the vendor before anything was installed. The trial software was downloaded in advance, so the on-site install would not consume the engagement time.
Integration platforms demonstrate well and implement badly, and the only useful test is one run against the interfaces you actually have.
When the vendor offered a reference visit to a nearby institution, the question back was whether they had worked with a particular person there — because the intention was to check independently. A vendor-arranged reference and a reference you call yourself are not the same evidence.
The decision was made by the licensing position, not the architecture
Three things, none of them technical, and all of them on the record:
- An existing entitlement the City already held was surfaced to the Chief Information Officer as the licensing route to the challenger platform, with its maintenance renewal expiring at the end of the year.
- A reseller quoting the incumbent’s next version replied that there were no upgrade licenses on the state contract — new licenses only.
- Procurement ruled that because the City was piggybacking another jurisdiction’s solicitation on a customer relationship project, it could not acquire third-party middleware at all under that vehicle.
Those are the mechanics of how an architecture decision actually gets made in a public body: by what is already licensed, what the contract vehicle permits, and which pot of money the work can be charged to.
An architect who does not read procurement rulings is watching somebody else choose the platform.
2022: making interoperability a design requirement before anything was installed
As enterprise architect on the new Medical Examiner’s facility, a four-hour working session was called in September 2022 with the agenda set here: how all the technology components would interact and function together.
In the room were the project architect and the interior designer, the mechanical and electrical engineers, the Public Buildings facility manager who would operate the building, and the City’s security engineering lead.
On a normal construction project each technology trade is procured, installed and commissioned separately, and whether the systems talk to each other is discovered at handover. Asking the question at design stage, with the designers and the eventual operators in the same room, is what turns interoperability into something the drawings and the specifications have to answer.
The interface matrix as the artifact that drove the platform choice
What came out of that session was a vetted technical interface matrix, worked system by system: which interfaces actually had to exist, between which systems, carrying what. It was still being worked through the following summer, and the next step logged against it was to meet the integration vendor.
The project’s own weekly task list then put the integration platform at item one — above the nine building subsystems it was to bind, each from a different manufacturer with its own installer and its own commissioning.
Putting integration at the top of the list rather than at the end of it is the whole argument. It was a workstream with its own hours from the start, not a reconciliation exercise discovered at handover.
Why a greenfield building was the right proving ground
A new building is the only place in a government estate where every system arrives at once with no legacy, under one construction budget, against one set of design drawings, with a commissioning date that forces the question.
It also answers the problem that killed the first attempt. In 2014 the platform had to be justified on its own against an incumbent being upgraded for nothing, which is an argument that cannot be won. Attached to a capital project, the platform is paid for as part of delivering the building and proves itself against live subsystems at the same time — and what comes out the other end is the reference implementation a wider rollout needs.
The scope put together in 2023 reflects that intent rather than one building: a cloud tenant with three buildings and five hundred devices, hybrid across software as a service, on-premises and a disaster recovery site. The device estate being sized against it ran to more than eleven hundred lights and two thousand telephones.
The licensing model was consumption-based — an annual committed transaction volume of several hundred thousand integration services and millions of application programming interface calls a month, with overage above it.
Which is a familiar shape from the other side of the desk. Eleven budget cycles had already gone into a chargeback model that allocated the City’s own services by measured consumption at published unit rates that moved with usage. Buying integration the same way is the same argument running in the opposite direction, and it is easier to sign a consumption commitment when you have spent a decade defending one.
Outcome
The platform the 2014 evaluation pointed to, bought in 2023 — attached to a capital project rather than argued for on its own.