04/05·Migration sprint

Migration sprint.

You use GPT or OpenAI and want to switch to Claude; for cost reasons, EU data residency, or because Claude works better for your use case. In ten working days I migrate your workflows, with a before-and-after eval comparison and a rollback plan in case anything breaks.

OpenAI to Claude, no breakage.

MIG-CLD-2W

·

EUR 7,500 net (6 to 9k)

·

10 working days

·

remote, one weekly sync

0X/05·Methodology anchor

Where in the Six-Step-Plan map.

[00]

method.step0Title

method.step0Sub

[01]

method.step1Title

method.step1Sub

[02]

method.step2Title

method.step2Sub

[03]

method.step3Title

method.step3Sub

[04] ●

method.step4Title

method.step4Sub

[05]

method.step5Title

method.step5Sub

Movement between models, eval-backed, no breakage.

Volle Methode ansehen →

02/05·Detail

// promise

OpenAI, workflow or model-version migration in two weeks, with eval comparison before and after and a rollback path.

// process

Week 1

Inventory the existing system, define gold-set, open migration branch, run first migration pass, run eval.

Week 2

Close gaps, test rollback, finalise cost delta, handover.

// deliverables

+Migrated codebase (PR against client repo) using the Anthropic SDK instead of the OpenAI SDK.
+Eval comparison: 20 plus gold-set examples, before/after score, latency, token cost.
+Rollback runbook with feature flag strategy.
+Cost delta report: monthly inference cost before/after.
+Short documentation (8 to 12 pages) with architecture diff.

exit criterion

Migration branch green on the eval, rollback path successfully tested once, cost delta report delivered. Client merges after internal QA.

// tech footprint

Anthropic SDK, OpenAI SDK (to be replaced), Claude skills/MCP, eval framework (custom or Promptfoo/Inspect), GitHub Actions or Fly.io for deploy.

03/05·Reference profiles

Wer das typischerweise kauft.

OpenAI to Claude migration in B2B SaaS

branche: B2B software or consultancy with a GPT pilot, compliance pressure towards Claude.
größe: 80 to 300 FTE.
anlass: DPO demands EU residency and lower hallucination rate; existing GPT implementation should remain in place.
outcome (geschätzt): Three workflows migrated in two weeks, eval score on Claude slightly above GPT, cost saving from prompt caching approx. 30 percent.

Model version migration (4.6 to 4.7)

branche: Existing Claude usage; SOPs to be packaged as skills.
größe: 50 to 500 FTE.
anlass: Anthropic announces a new model version; existing prompts must be reviewed, evals re-run.
outcome (geschätzt): Controlled model switch with documented eval delta. For SOP migration: three to five skills in the client repo instead of PDFs in SharePoint.

04/05·Testimonial

template, zur Freigabe durch Kunde

[TEMPLATE]

„The migration sprint saved us two weeks of in-house work and delivered a clean eval comparison. We could justify the model migration to our management with a date, numbers and a diff."
Engineering Manager, B2B SaaS, approx. 120 employees.

05/05·What you do not get

Does not fit.

×Migration plus new build in one sprint (fixed price breaks).
×Client has no gold-set and no willingness to build one.
×Live traffic must not deviate but eval is not yet set up (hardening first).