01/05·AI workflow audit

AI workflow audit.

You have a process that takes too long or is too error-prone. In five working days I analyse what an agent could take over, what it should not, and what it costs to build. You get an implementation plan, an effort estimate, and the GDPR templates.

One workflow, five days, fixed price.

AUD-WF-1W

·

EUR 4,800 net

·

5 working days

·

remote, 1 person

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

Steps 0 to 2 in five days: journey, stage map, top findings.

Volle Methode ansehen →

02/05·Detail

// promise

After five working days you have a technically solid implementation plan for one specific workflow on the table, including a draft data processing agreement and a DPIA template.

// process

Day 1

Kickoff, workflow walkthrough, data sources mapped.

Day 2

Stakeholder interviews (max. 3), compliance requirements (GDPR, BFSG) captured.

Day 3

Architecture draft, risk register, MCP/skill sketch.

Day 4

Document finalised, internally reviewed (multi-role panel: UX, architect, risk).

Day 5

Handover session, recording, follow-up decision documented.

// deliverables

+Implementation plan PDF (12 to 18 pages, Janis brand): current state, target architecture, data flow diagram, risk register, effort estimate in hours.
+Draft data processing agreement (DPA) between client and Anthropic, pre-filled.
+DPIA template (data protection impact assessment) for the audited workflow.
+Recommendation matrix: pilot yes/no, build vs. buy, model selection (Sonnet vs. Opus).
+60-minute handover session with live Q&A.

exit criterion

Handover session held, PDF plus DPA plus DPIA delivered as email attachment, follow-up decision documented in writing.

// tech footprint

Claude (Sonnet 4.6 or higher), MCP, optionally the Anthropic SDK. The client's existing stack is mapped, not replaced.

03/05·Reference profiles

Wer das typischerweise kauft.

Tax firm with 60 to 120 employees

branche: Tax advisory, DATEV-centric, multi-site.
größe: 60 to 120 FTE, one to two IT leads, no in-house ML.
anlass: Management has rolled out M365 Copilot or ChatGPT Business; expected value in client onboarding, transcript extraction and reporting fails to materialise.
outcome (geschätzt): A recurring processing cycle reduced from an estimated 6 to 8 hours to 30 to 60 minutes per case after the pilot.

Insurance broker or HR consultancy, 50 to 200 employees

branche: B2B service with a high share of structured text work.
größe: 50 to 200 FTE, in-house IT of one to three people.
anlass: First ChatGPT pilot ran without structure; data protection officer blocks rollout because DPA and data flow are unclear.
outcome (geschätzt): Procurement clearance in 5 to 10 working days instead of 6 to 12 weeks; one prioritised workflow with clear ROI.

04/05·Testimonial

template, zur Freigabe durch Kunde

[TEMPLATE]

„We had already introduced Claude but did not know where to start concretely. Janis Tirtey selected one workflow within five days, documented it fully and delivered an implementation plan including a draft DPA. With that we closed the data protection sign-off within two weeks."
Managing director of a tax firm, approx. 80 employees.

05/05·What you do not get

Does not fit.

×Client wants an AI strategy across the whole company.
×Several workflows in parallel are to be reviewed (then two audits in sequence).
×Nobody on the client side has 4 hours per week for interviews.