A shop running on IBM i is a fine mechanical watch — decades of RPG and Db2 cut to fit, a movement that has kept perfect time for forty years. That is exactly why you can't strap agents onto the outside of it.
A mechanical watch keeps time because it is a closed, integrated whole. Every wheel and jewel was cut to fit the ones beside it; the movement runs because it was designed together, as one thing. That is its excellence — and its constraint. You cannot open the case and screw a new complication onto the outside of a sealed movement and expect it to keep better time. It will keep worse.
Your shop is that watch. Forty years of business logic encoded in RPG, decades of truth living in Db2, a sovereign core that has never missed a close. It runs the company because it was built as one integrated movement — below the application, beneath the noise, exactly where the real work happens.
So when agents arrive, the tempting instinct is the wrong one: strap them onto the outside. Bolt a copilot here, an agent there, hang them off the running movement and hope they mesh. They don't mesh. They interfere. Not because any one part is bad — because you added to a sealed whole from the outside, and a sealed whole has no outside to add to.
The bolt-on instinct isn't a strawman. It's the default enterprise play right now — strap governed agents onto the estate you already run, keep everything where it sits.
And the warning is already in print: build your agent workflows, governance, and orchestration entirely on the vendor's framework, and the lock-in simply relocates to the orchestration layer. Deploy without building internal muscle, and you end up reliant on their platform for every update and extension. You strapped something onto the outside — and now it's their movement, not yours.
— industry analysts on the wave of vendor "agentic advantage" services, early 2026
You would never crack open a watch that's keeping time and rebuild it live on the wrist. The business is still running on it. Instead, a watchmaker builds the reworked movement on the bench, beside the running one — proves it keeps time — and only then swaps it in.
That is the whole reason the first agentic workflow stands up outside the existing structure, reporting to the C-suite rather than inside the org it's meant to reshape. It isn't an org-chart nicety. It's the only way to rework an integrated whole without stopping the clock. You lift one flow out, rebuild it the new way in a protected space beside the running company, and let it prove itself in shadow — the override rate falling turn after turn until it's carrying real work on its own.
On IBM i, this has a name and a shape the community already trusts. The pilot runs beside the sovereign core, not through a rewrite of it. An agent is just another identity on the box — reached at the object level, journaled, gated before the write lands. The reworked movement is built beside the watch, cut to fit, then set running.
When the movement is reworked as a whole rather than burdened from the outside, the human shape of the company changes — and it changes upward. This is the part every stalled transformation skips, and it's the part that decides whether the rework holds.
That is what "woven in, not bolted on" actually means on an IBM i shop. Not agents hung off the estate. A company reworked as one movement — where the intelligence stays yours, because you rebuilt your own watch instead of strapping on someone else's.
~5 MIN · NO EMAIL · STAYS IN YOUR BROWSER
Rather start with one flow? Book a 30-minute 4i / Tech call →