Treating it as a feature wish list
Accelerate is data-driven optimization. If the client starts submitting 'nice-to-have' feature requests, the phase is being misused and the system will bloat.
III.Flame
05 · Blaze · 60 to 90 days
Accelerate
Stop installing. Start optimizing.
Accelerate is what happens after your new operating system has been live for 30 days. We stop installing and start optimizing. We find the two or three adjustments that will produce the next step-function of improvement, and we ship them. This is also where net-new SI agents get layered in — once the system has found its rhythm.
What it is
A 60–90 day phase of measured optimization against a live system. We run the numbers, interview the operators using the system, and identify two or three high-leverage adjustments. Then we ship those adjustments and add net-new SI surfaces — agents, copilots, predictive layers — that weren't appropriate to install earlier.
What it is not
Accelerate is deliberately not:
- Not a second Build. If Architect and Build were done well, Accelerate is tuning, not reconstruction.
- Not continuous feature drift. Every Accelerate change is tied to an observed performance signal from the live system.
- Not open-ended. 60–90 days, fixed scope, then we either move to Sustain or close the engagement cleanly.
You leave with
A measurably faster, smarter operating system
Two to three measured performance improvements on the live system, each tied to a specific observed signal. One to three net-new SI surfaces added on top of working workflows. A written performance report comparing the post-Accelerate baseline to the end-of-Build baseline. A go/no-go recommendation on Sustain.
Why we wait for Accelerate to add SI agents
Every consultant wants to install the SI agents in Phase 3. We don't, and the reason is simple: adding an intelligence layer to a system that hasn't yet found its operational rhythm creates confusion, not leverage. You have to see the system run against real load before you know which workflow benefits from an agent and which would be better left as a rule-based automation.
By day 30 of live operation, the signal is clear. That's when Accelerate starts. That's when the agents go in.
What 'the two or three adjustments' actually look like
In practice, a typical Accelerate surfaces adjustments like: a misconfigured pipeline stage that's dropping lead follow-ups; an automation threshold that's too tight and firing false positives; a human-in-the-loop checkpoint that's redundant for 95% of cases and should be removed; an inbound classifier whose categories are too broad. None of these are visible in Build. All of them are obvious after 30 days of real load.
When Accelerate moves into Sustain
Accelerate is bounded — 60 to 90 days. At the end, you either move into Phase 6 Sustain (a monthly retainer that keeps the system alive, upgraded, and expanding), or we run a formal handoff with full documentation and a 90-day advisory window. Both are legitimate endings. The one ending we refuse is open-ended engagement drift.
Where it fails
Adding SI surfaces before the baseline is stable
If a workflow is still being tuned in Accelerate, that is not the time to layer an LLM on top of it. Stabilize first, then add intelligence.
Never closing the phase
Accelerate must end. Either it moves to Sustain (retainer) or to handoff. The worst outcome is perpetual 'we're still in Accelerate' after six months — that's just project drift with a consultant invoice attached.
Phase 5 Accelerate — FAQ
Why isn't SI agent work part of Phase 3 Build?
Because adding an intelligence layer to a system that hasn't yet found its rhythm creates confusion. By day 30 of live operation the signal is clear. Agents go in during Accelerate, not during Build.
How long does Accelerate run?
Sixty to ninety days. It is bounded — Accelerate that runs six months is project drift, not optimization. At day 90 you're either in Sustain or you've handed off.
Can we skip Accelerate and go straight to handoff?
Yes, if the system is performing to plan at end-of-Amplify. Roughly a third of our engagements skip Accelerate. The other two-thirds find enough measurable gain in 30 days of real load to justify it.
System live but not yet humming?
If your system has been running for 30+ days and you suspect there's measurable gain left on the table, Accelerate is the right scope.
Get your time back.
Nicholas runs this studio on it. Mobile speed went from 49 to 92. We run on it too.